Sie sind auf Seite 1von 35

EJEAFChe

Electronic Journal of Environmental, Agricultural and Food Chemistry ISSN: 1579-4377 A CRITICAL REVIEW OF TOTAL VOLATILE BASES AND TRIMETHYLAMINE AS INDICES OF FRESHNESS OF FISH. Part 1. Determination.

Peter Howgate 26 Lavender Row, Stedham, Midhurst, West Sussex GU29 0NS, UK e-mail phowgate@clara.co.uk ABSTRACT. The large literature on the formation of volatile amines, and of their aggregation as Total Volatile Base (TVB), during spoilage of vertebrate, teleostean fish in ice, and their used as indices of freshness is critically reviewed. This part reviews analytical procedures. The thermodynamics of the distillation process in determining TVB is reviewed and optimum conditions for the distillation are considered. Distillation other than at temperatures below about 501C is accompanied by decomposition of nitrogen-containing compounds with the formation of ammonia which additional to the intrinsic ammonia in the sample. The amount of decomposition differs among the various procedures used for determining TVB. Rapid distillation in automated distillation units produces the smallest amount of decomposition. Trimethylamine (TMA) is often determined by the picrate method, but this is not completely specific for TMA and some dimethylamine is included in the result depending of the alkali and the amount of formaldehyde used. Other analytical techniques such as GLC and Flow Injection Analysis are specific and are to be preferred for critical work. There are differences among procedures for determining TVB and TMA in the way concentrations in the sample are calculated when using protein-free extracts which leads to biases between methods. KEYWORDS: TVB, TMA, DMA, ammonia, distillation, analysis, picrate method

I. INTRODUCTION Fish are notorious for spoiling rapidly even when stored under chill conditions. The storage life to end of good quality for typical demersal, white-fleshed species is around 7 days in ice, and by about 14 days the fish is unfit for consumption. Equivalent storage lives for pelagic fish are even less. Measurement of freshness is therefor an important part of quality assurance of chill-stored fish, and though sensory methods are generally considered to be the most appropriate for measuring freshness of fish, there is a role for non sensory methods (Botta, 1995; lafsdttir et al., 1997). The chemical test with the longest history of use as an indicator of freshness is measurement of the amount of basic compounds recovered by distilling fish muscle, or extracts of fish muscle, under alkaline conditions. The amount of Total Volatile Bases (TVB) is almost invariable expressed on a nitrogen basis, each component of the mixture containing one nitrogen atom per molecule, as Total Volatile Base Nitrogen (TVBN). The former term will be used in this review. The formation of TVB and its constituent amines in spoiling fish muscle and their relevance as measures of spoilage have been studied for almost a century now and consequently there is now a considerable literature on the subject. Though this literature runs to some hundreds of publications including reports of research, summaries, and discussion papers there are few extensive reviews. Hebard et al. (1982) wrote a comprehensive review of the subject of methylamines in fish emphasising the chemistry and microbiology of the formation of amines in chill stored and frozen stored fish and in fishery products, but also discussing the analysis of amines and TVB and their suitability as indices of quality. The paper cites 500 or so references. Oehlenschlger (1997) is a bibliography of methylamines and TVB as measures of spoilage, with a brief review, which cites 133 references and Botta (1995) has a short discussion of analytical methods for determination of TVB and its use as a measure of freshness. The objective of this review is to critically review analytical procedures for measurement of TVB and amines in fish muscle, their formation in fish, predominately vertebrate fish, during iced storage, and the effectiveness of TVB and individual methylamines as measures of spoilage. Part 1 will provide a general introduction to the topic and a detailed review of methods for their determination, Part 2 a review of formation of volatile bases in spoiling fishery products and applications in quality assurance. II. AN HISTORICAL PERSPECTIVE It seems to have been known in the mid 19th century at least that amines are present in fishery products. Winkles (1855) isolated amines from spent pickled herring liquor by a rather heroic preparation starting with 26 gallons, (120 litres), of liquor from which the amines were recovered by a series of distillations under alkaline conditions with the most volatile amine being identified as trimethylamine (TMA). (Readers of the paper will note that the empirical formula of the methyl group is given as C2H3 because at that time the atomic weight of carbon was considered to be 6; it was revised to 12 in Cannizzaro's revision of atomic weights in 1860). Winkles' (1855) isolation of TMA was part of a study on amines as chemical entities and not particularly of amines as components of fishery products, and systematic investigations of the spoilage of fish and the formation of TVB and amines during spoilage began in the following century. Anderson (1908) described changes in sensory properties of salmon during spoilage in ice and made some microbiological observations, but he did not carry out any chemical analyses. The earliest paper that the reviewer can find on TVB in the context of spoiling fish is Tillmans and Mildner (1916) who measured TVB by distillation of fish muscle in the 2

presence of magnesium oxide. At about the same time, Clark and Almy (1917) considered some chemical analyses that might possibly be used to monitor spoilage of fish and tested some candidate procedures for interference by constituents of fish muscle. The authors measured TVB by an aeration procedure and obtained complete recoveries of ammonia and amines from test mixtures, but did not apply this or any of the other tests to samples of fish. They recommended 'That a study be made of the value of volatile distillation products like amines, indol, etc. in detecting incipient decomposition in fish.', but did not cite Tillmans and Midler (1916). Weber and Wilson (1919) used the aeration procedure to determine TVB in canned sardines and measured the ammonia and methylamines contents in the distillates. A table in the paper shows the quantities of 'Volatile Alkaline Material' in canned sardines, and, for some samples, concentrations of ammonia, mono-, di- and tri- methylamines as well, all on a nitrogen basis. This study was concerned with changes in composition of the canned products during subsequent storage rather than with spoilage in the raw material, but is interesting for being the earliest published report of concentrations of methylamines in a fish product. Tillmans and Otto (1924) examined some chemical tests, including determination of TVB, for their potential to indicate the onset of spoilage. The text of the paper reports that the authors compared various distilling procedures with and without addition of baryta and selected vacuum distillation without baryta as the most stable. This is the only report encountered by the reviewer of distillation without the addition of an alkalising agent. The authors do not refer to the procedure as measuring TVB, but as measuring ammonia, ('ammoniak'), concentration. On the basis of the storage of some species of demersal fish at 12-151C the authors concluded that an 'ammonia' concentration of 30 mg (100g)-1 indicated the beginning of spoilage. The 1930's was a fruitful decade for studies of volatile amines in fish. In 1936 two papers were published that are frequently cited in the literature of TVB in fish. Lcke and Geidel (1935) measured TVB by two procedures: distillation of an aqueous extract of fish muscle using magnesium oxide as the alkalising agent, and distillation of a sample of minced muscle suspended in water with magnesium oxide. On the basis of storage of several species of fish in ice and analysed for TVB by the latter procedure the authors considered the fish no longer completely unobjectionable, ('einwandfrei'), when the TVB content exceeded 30 mg (100g)-1. The latter TVB procedure, distillation of muscle tissue directly with magnesium oxide, or slight variants of it, has been used to determine TVB in many studies of spoilage of fish and is often referred to as the Lcke and Geidel (L&G) method, (with variations of the spelling). Boury and Schvinte (1935) briefly reviewed various sensory, physical and chemical procedures for measuring freshness of fish, but concentrated on a more detailed study of the measurement of TVB. The authors remark that it was well known that heating muscle tissue with alkalis cause formation of ammonia, notably from amides, and they examined various combinations of distillation conditions and alkalis for the extent of ammonia production during distillation. They concluded that distillation at 37-381C under reduced pressure with lithium carbonate as the alkalising agent gave the lowest amount of decomposition ammonia, and that distillation at normal pressure with magnesium oxide gave TVB values close to lithium carbonate at reduced pressure. Beatty and Gibbons (1937) briefly discussed chemical methods for measuring the spoilage of fish and presented data on TVB contents measured by distillation of muscle tissue with sodium borate under reduced pressure. However, that paper is more significant for studies on TVB because of it introduced the Conway distillation cell for measurement of TVB and TMA. Prior to Beatty and Gibbons' (1937) method for measuring TMA, individual amines in the TVB were measured by cumbersome procedures involving direct measurement of ammonia content and selective decomposition of the methylamines by nitrous acid (Boury and Schvinte 1935; Reay, 1937). Reay (1937) also describes a colorimetric method for determination of dimethylamine (DMA) in fish muscle 3

extracts as its copper dithiocarbamate complex. Poller and Linneweh (1926) had cited publications on the presence of trimethylamine oxide in dogfish and in cephalopods, and also isolated it from herring muscle, but the significance of their paper for the history of TVB and volatile amines in fish is in their reporting of studies in another laboratory in Germany, (not referenced in the paper), showing that spoilage bacteria reduced TMAO to TMA. Further studies in the 1930's on TMAO content of fish and its reduction to TMA by bacteria during spoilage confirmed these observations (Beatty, 1937, 1938, 1939; Brocklesby and Riddell, 1937; Cook, 1931; Tarr, 1939; Watson, 1939)1 .

The title of the paper by Cook (1931) is misleading in that the paper is about TMAO, not TMA 4

By the end of the 1930's the main features of the measurement of volatile amines in fish and their formation during spoilage had been established: the amount of TVB determined is affected by the analytical procedure because of decomposition of nitrogen-containing substances during the distillation step in addition to the volatile amines present, the more severe the distillation conditions in terms of strength of alkali and temperature of distillation, the greater the degree of decomposition; the increase in TVB during spoilage is almost entirely accounted for by the increase of TMA until the fish is very spoiled, (at least in the species studies up to that time); the TMA is formed by reduction of TMAO by some species of the bacterial flora of spoiling fish; TMAO appears to be present in the flesh of all species of marine fish, but not present consistently in flesh of freshwater fish; TMA is not formed during storage of sterile muscle; TMA content of spoiling fish muscle increases approximately exponentially with storage time and is highly correlated with overall bacterial count; the TVB content of typical marine demersal fish at the limit of acceptability as a result of spoilage is around 30 mg nitrogen (100g)-1 of flesh. In the following decade a colorimetric method for determination of TMA as its picrate salt (Dyer, 1945; Dyer and Mounsey, 1945) was developed which considerably simplified the analysis of this amine in fish muscle and made a significant contribution to subsequent studies on TMA in stored fish and fishery products and its use as measures of spoilage. There is now a very large literature on volatile amines in spoiling fish which has contributed to our knowledge of the effect of species and storage conditions on production of these amines in fish and fishery products though not significantly adding to or modifying the main features established during the 1930's and listd above. There have been developments in analytical procedures since the early days of studies in volatile amines of which the application of gas liquid chromatography (GLC) has been the most significant, but judging from the literature, and the reviewer's own experience, the analytical procedures for TVB and TMA most commonly used nowadays in research and industry have been, and still are, based on those developed in the 1930's and 40's. III. ANALYTICAL PROCEDURES A. TVB 1. General Principles TVB content of fish muscle is an attribute of the composition of the muscle for which the analytical methodology is implied in its name, not by the components. By definition it must be determined by a procedure which volatilises the bases which are recovered and quantitatively determined. The procedure determines the sum of the bases without discrimination among them, but analysis of the components of the TVB of spoiling muscle show it to contain predominately ammonia and TMA with, in some instances, a small contribution of DMA; monomethylamine (MMA) and higher amines are not present, or if so, only in trace amounts. The component volatile bases could be determined separately and summed to give an estimate the TVB, but this nullifies the simplicity of measurement of TVB with regard to both concept and analytical methodology. The difficulty with analytical procedures for measuring TVB is that fish muscle and extracts of fish muscle contain compounds that decompose under some distillation conditions to produce ammonia which is estimated along with the ammonia and amines already present in the sample. The effect has been recognised from the earliest days of measurement of TVB in fish muscle, and has been noted in measurement of ammonia by volatilisation in other biological materials (Nichols and Foote, 1931). Boury and Schvinte (1935) in their extensive study of measurement of TVB remark that it is well known, ('bien connu' in the original), that ammonia, (notably from amides), is formed by hydrolysis during distillation. The completeness of distillation of the 5

volatile bases increases with alkalinity of the solution or muscle suspension being distilled and with the temperature of distillation, but these are also the conditions that increase the extent of decomposition of nitrogen-containing compounds, and practical analytical procedures attempt to reconcile these competing considerations. This has led to a variety of procedures being used for determination of TVB, but most fall into one of three major classes: distillation of a suspension of muscle tissue in water with a moderately mild alkali such as magnesium oxide (MgO), often referred to as the L&G method when MgO is used; steam distillation of a protein-free extract of muscle tissue with a strong alkali such as sodium hydroxide; and diffusion of the bases at room temperature or a little above in a Conway cell. One factor that affects the rate of distillation of the bases from a distilling mix is the proportion of the base that is in the unionised form at the pH and temperature of the mix. The pKas of ammonia, DMA and TMA at 251C are given reference books as 9.25, 10.73, and 9.81 respectively. pKs are temperature dependent and Emerson et al. (1975) derived an expression for the effect of temperature on the pKa of ammonia in the range 0-501C based on published information: pKa = 0.09018 + 2729.2 T-1, (1)

T in kelvins. The reviewer could not find a corresponding equation for ammonia above 501C, nor for DMA and TMA at any temperature range, but extrapolating this equation to1001C for all of the bases gives pKas of 7.41, 8.89, and 7.97 for ammonia, DMA and TMA respectively. The pH of a suspension of MgO and fish muscle does not seem to have been recorded, but an estimate can be made from its solubility in water. A search of the Internet produced several values of the solubility product of magnesium hydroxide, but the most frequently listed value was 5.6x10-11 at 181, equating to a solubility of 6.7x10-4M, 0.04 g l-1. Assuming all of the dissolved magnesium hydroxide is ionised the pH of a saturated solution of MgO should be about 10.7 at room temperature. The solubility of the MgO will be greater at the temperature of distillation, which will raise the pH, but the fish muscle will tend to buffer the mixture and lower the pH so it is not easy to anticipate the pH of the mix during distillation. Assuming a pH 10.5 the proportion of the bases in the unionized form are 0.98 in the case of DMA and more than 0.99 in the cases of ammonia and TMA. Though it would be unwise to take these extrapolations as accurate predictions of the dissociation of these bases at 1001C they show the at least the direction and possible size of the effect, and it would appear that incomplete association of the bases into the free form will not be a factor in the measurement of TVB by simple distillation using MgO or stronger alkalis as the alkalising agent. Steam distillation procedures usually specify use of strong alkalis and in these case the bases will be almost completely in the unionised form. Weaker alkalis giving lower pHs in the distilling mix can be used, but at the expense of longer distilling times to effect complete recovery (Nichols and Foote, 1931). The Conway procedure is carried out at room temperature or a little above, but again strong alkalis are used and the proportion of TVB bases in the unionized form will be greater than 0.99. Another factor that affects the volume of distillate required for complete recovery of a base, and hence the distillation time, is its volatility, a property measured by the Henry's law constant for the base. Egnr and Johansson (1938) made a theoretical study of the distillation of ammonia as performed in the final step in the Kjeldahl determination of nitrogen, and for simple distillation of volatile substances from an aqueous solution they derived an expression for relating the proportion of a volatile substance distilled over to the proportion of the original distillation solution distilled over. Rearranging the equation as given in the original paper, and using different symbols, a form of their equation is: 6

(1-Pm) = (1-Pv)Q

(2)

where Pm is the proportion of the original mass of volatile substance present in the distillate and Pv is the proportion of the original volume of distilling solution distilled over. Q is a volatility coefficient equal to H/k where H is the dimensionless Henry's law constant, the ratio of concentration of the substance in the vapour phase to that in the liquid phase, and k is the ratio of the specific volume of water to that of saturated steam at the temperature of distillation. Values of k can be obtained from standard steam tables and at 1001C is 6.24x104 . Egnr and Johansson (1938) estimated a value of 80x10-4 for H for ammonia at 1001C from published data giving a value of Q of 12.8. They studied the distillation of ammonium solutions in water using MgO as the alkaliser and obtained an average value for Q of 14.4, equating to a value of 90x10-4 for H, in reasonable agreement with the assumed value. Sander (1999) has compiled data on the Henry's law constants for very many chemicals and lists 15 values of the constant for ammonia. According to information in the tabulation some of these values are secondary data, that is, appear in reviews or similar, and some have not been checked by the compiler. Four values are shown as measured values and their mean is 6.04x10-4 at 251C when converted to the dimensionless constant from the units used in the compilation. The compilation reports that tabulated values, which are given for 251C, can be converted to values at other temperatures using the van't Hoff relationship and tabulates appropriate coefficients for the equation. Using the mean of three measured values of the coefficient for ammonia the Henry's law constant for ammonia at 1001C is calculated to be 83.7x10-4, equating to a value of 13.4 for Q compared with the observed value of 14.4. (There are reasons, discussed below, to expect the value of Q, and hence H, observed in simple distillation experiments to be greater than the true value depending on the conditions of distillation). Using the expected value of Q, 13.4, in equation (2) it can be calculated that 99% of the ammonia in a solution will be distilled over when the distillate volume is 29% of the volume of liquid originally in the distilling flask. Sander (2007) lists single, measured, values of H for TMA and DMA of 42.6x10-4 and 7.47x10-4 at 251C respectively when converted to the dimensionless constant. The value of H for TMA is very much greater than the corresponding value for ammonia and all of the TMA would be recovered in the volume of distillate required for recovery of ammonia. The values of H for ammonia and DMA are very similar at 251C and whether or not the volume of distillate required for recovery of ammonia would suffice for recovery of DMA will depend on the relative sizes of the temperature coefficient in the van't Hoff equation, but unfortunately, measured values are not included in the Sander (2007) compilation. 2. Direct Distillation of Fish Muscle In the procedure described in Lcke and Geidel (1935), the usual citation for the determination of TVB by direct, (as distinct from steam), distillation of fish muscle, 5-10g of sample is suspended in 300ml of water and distilled with MgO. Using the calculations above this starting volume would require a distillate volume of 87ml for recovery of substantially all, 99%, of the ammonia. The original paper specifies the time of distillation - 25 minutes following 10 minutes to reach boiling - but not the volume of distillate or the distillation rate. The only publication the reviewer has come across relating to distillation of individual methylamines in the context of TVB determination by any procedure is in Hjorth-Hansen and Bakken's (1947) detailed review of analytical procedures for measuring amines in fish and fish meal. They measured the time course of the distillation of ammonia and methylamines from aqueous solutions using MgO as the alkali. Figure 3 of the publication shows a plot of 7

the proportion of base distilled over against the distillation time, starting from when the distilling liquid started to boil, as smooth curves without experimental points, but the text does not report the volumes of the distillates, nor the distillation rate from which the volumes can be calculated. The authors cite the Egnr and Johansson (1938) paper and Figure 4 of Hjorth-Hansen and Bakken (1947) shows a plot of the fraction of ammonia distilled over against the volume fraction of distillate in the range 99.5% to 99.9% of ammonia distilled over computed from the Egnr and Johansson equation using the value of 14.4 for Q from that paper, but it is not possible to infer from this curve the distillation rate used in determining the experimental curve for ammonia. (The Egnr and Johansson equation presented in Hjorth-Hansen and Bakken (1947) has a typographic error in that the (division) symbol in the quoted formula should be a (minus) symbol). In the case of ammonia and using equation (2) with a value of 13.4 for Q the expected value of Pv for Pm = 0.95 is 0.20. The original starting volume used in Hjorth-Hansen and Bakken (1947) was 250ml so the volume of distillate at this point would be 50ml. The ammonia distillation curve in Figure 3 shows a Pm of 0.95 being obtained at a distillation time of 20.5 minutes giving a mean distillation rate to that point of around 2.5 ml minute-1. The distillation curves in Hjorth-Hansen and Bakken (1947) show TMA distilling over much faster than ammonia as would be expected from relative values of H at 251C. 95% of the TMA is shown as distilling over at 9.1 minutes and using the assumed distillation rate of 2.5 ml min-1 a value of Q of 31.3 for this base is predicted. Though their values of H are similar at 251C the data shows DMA distilling over much slower than ammonia does and 95% of the DMA is recovered in 43.6 minutes equating to a value of 5.23 for Q assuming a distillation rate of 2.5 ml min-1. From these measured and estimated values of Q it can be calculated that the fraction of the original distilling solution to be distilled over, Pv, for 99% recovery of TMA, ammonia and DMA are 0.14, 0.29 and 0.59 respectively. If the sample is suspended in 300ml of water, the conditions of the Lcke and Geidel (1935) procedure, the corresponding volumes to be distilled over are 41ml, 87ml and 176ml. These values should be considered maximum estimates and in practice the actual volumes required to be distilled will be less than these. Egnr and Johansson (1938) pointed to some aspects of experimental procedures that would reduce the amount of distillate required to recover a given proportion of the compound being distilled, the most important of which for the case of the L&G procedure would be refluxing in the distillation apparatus. The effect of refluxing is to increase the concentration of base in the vapours reaching the receiving flask compared with the situation without any refluxing and Egnr and Johansson (1938) derived an equation relating the degree of enrichment to the proportion of water returning to the distillation flask due to refluxing. This enrichment results in an effective increase in the value of Q and Egnr and Johansson (1938) demonstrate the effect with experimental studies using different distillation assemblies. This reflux effect is apparent in the Hjorth-Hansen and Bakken (1947) data when the curve for ammonia in their Figure 3 is compared with the curve calculated from equation (1) using a distillation rate of 2.5 ml min-1. At the start of the distillation the proportion of ammonia distilled over is greater than predicted from the equation and the deviation decreases as distillation proceeds. Hjorth-Hansen and Bakken (1947) recorded their distillation times from when the distilling solution started to boil, a requirement of the L&G procedure, and refluxing would be greater at the start than later when the splash head and distillation head have warmed up. There is insufficient information in the Hjorth-Hansen and Bakken (1947) paper to calculate the effect of refluxing, but judging from the examples in Egnr and Johansson (1938) the consequence is probably to reduce the volume required for 99% recovery of the bases to by at least 10% of the volumes quoted above. It follows from theoretical considerations of distillation discussed above that protocols for 8

determination of of TVB by simple direct distillation should specify the volume of distillate rather than time of distillation. The procedure specified in Lcke and Geidel (1935) requires that the distilling liquid be brought to boiling in 10 minutes and if using their apparatus this would specify the distillation rate and hence the volume of distillate recovered in the 25 minutes distillation time. However, deviations from these standard conditions could lead to errors in the TVB measurements. It would appear from the discussion above that recovering 29% of the original in the distillate will recover more than 99% of the TMA and ammonia, but only about 85% of the DMA. However DMA is a minor constituent in TVB and less than complete recovery of this base would not appreciably affect the accuracy of TVB measurement. It follows from the theoretical treatment of distillation that a smaller volume of original distilling liquid will require a smaller volume of distillate to recover 29% in the distillate. Using an original volume of 100ml, for example, rather than the specified 300ml will require only 9 minutes distillation time at most rather than 25. Apart from any saving in time the reduced reaction time will result in a smaller amount of ammonia being formed by decomposition, (see discussion below), thus leading to a measured TVB value closer to the true value. An alternative procedure to direct distillation is steam distillation. The semi-micro distillation glassware commonly using in the Kjeldahl procedure for nitrogen determination is not suitable because it does not cope well with solid material, and usually more basic glassware is needed. Antonacopoulos (1960) has described an apparatus for steam distillation of foodstuffs which has been extensively used for determination TVB in fish and is often referred just as the 'Antona apparatus'. The glassware is not typically a stock item in catalogues of laboratory suppliers and usually has to be made to order. The sample of minced fish is transferred to a distilling tube with a small amount of water, MgO added, the tube inserted into a flask of water acting as a steam generator, and steam distilled (Antonacopoulos, 1960, 1968; Antonacopoulos, 1973; Antonacopoulos and Vyncke, 1989). The procedure specifies distillation at 10 ml min-1. for 12 minutes giving a distillation volume of 120ml. The traditional glassware used in the distillation step of the Kjeldahl determination of nitrogen is giving way to semi-automatic, rapid, distillation units and this equipment can be used for distillation of muscle samples with MgO (Malle and Tao, 1987;Vyncke, 1996; Surti et al., 2001; European Commission, 2005) and is now a convenient alternative to the Antona apparatus. During steam distillation the distilling mix does not decrease in volume as it does in simple distillation and Egnr and Johansson (1938) developed another expression for this case: (1-Pm) = e-QPv (3)

Symbols the same as equation (2). Inserting the value of 13.4 for Q in the equation predicts that the volume of the distillate for 99% recovery of ammonia and TMA needs to be 34% of the volume of the distilling solution. In the case of steam distillation for determination of TVB the distilling volume is typically in the range 25-50ml leading to sufficient volumes of distillate in the range 9 -18ml. All the TMA will be recovered in this proportion, but the volumes needed for recovery of DMA will be about two and half times these, 22.5 - 45ml. The 120ml specified when using the Antona apparatus then seems excessive even to ensure recovery of DMA. 3. Distillation of Fish Muscle Extracts Distillation of solid samples has some drawbacks in practice in that the distillation apparatus has to able to handle solid samples, and the apparatus has to be dismantled at the end of the 9

distillation to remove the solids. These considerations often lead to bulky and unwieldy apparatus and a slow throughput of samples though the Antona apparatus and rapid distillation units overcome these. Especially the rapid distillation units because these can complete the distillation in 5-10 minutes and tubes can be prepared while other samples are being distilled. An alternative is to prepare extracts of the fish muscle and use semi-micro Kjeldahl glassware for the distillation, invariably by steam distillation. Distillation times are in the order of 10 minutes and throughput is rapid especially if glassware that automatically siphons off the distillation mix at the end is used. Use of extracts does require extra processing steps, but there are advantages in that large sample weights can be used so reducing sample variance, and the extracts might be required for another analyses anyway. Another benefit is that much of the nitrogen-compounds with the potential to decompose to form ammonia are removed from the distillation matrix. Some of the earlier studies on TVB investigated use of aqueous extracts (Tillmans and Otto, 1924; Hess, 1932; Boury and Schvinte, 1935; Lcke and Geidel, 1935; Hillig et al., 1958) or press juice, the liquid squeezed from muscle by pressure, (Beatty, 1938; Watson, 1939), but more often protein-free extracts of various sorts have been used. The use of acidic colloidal ferric hydroxide has been described by Tillmans and Otto (1924), Hjorth-Hansen (1952) and Hjorth-Hansen and Bakken (1947), magnesium sulphate by Tomiyama et al. (1956), formaldehyde by Dyer and Mounsey (1945), and ethanol by Stansby et al. (1944), but use of these precipitants have rarely been reported outside of the original papers. By far the most frequently used protein precipitants used in determination of TVB and volatile amines in fish are trichloroacetic acid (TCA) and perchloric acid (PCA). TCA seems to have first been used in analysis of TMA fish muscle in Canadian laboratories (Dyer and Mounsey, 1945) and rapidly taken up in other laboratories. The concentrations used are usually 5 or 7.5% at a ratio of usually 1:2 or 1:3 fish muscle to extractant. PCA was first used in studies on the post mortem biochemistry of fish muscle (Saito and Arai, 1958; Jones, 1960), (the perchlorate ion can be removed as its insoluble potassium salt, whereas TCA needs to be removed by an organic solvent), though the extracts can be used for determination of TVB and amines. The concentration used is almost invariably 6%, (0.6M), and ratios 1:2 to 1:9 fish:extractant have been used in making the extracts. An extract prepared by grinding muscle with solid TCA can also be used (Shewan et al., 1971). Sodium hydroxide is the common alkalising agent, but MgO (Vyncke et al., 1987) and calcium hydroxide (Hillig et al., 1958) have been used. An aspect of the use of extracts for determination of TVB on which there is lack of uniformity is that of calculating the TVB concentration in the muscle tissue. (And when calculating individual bases in extracts for that matter). Muscle tissue is blended with the extractant, the TVB determined in an aliquot of the extract, and the concentration in the aliquot scaled up to give the concentration in the tissue. In some descriptions of procedures the scaling factor is the ratio of the total mass of sample plus volume of extractant to the volume of aliquot, but, on the basis that the bases are dissolved in the aqueous phase only, the factor should be the ratio of the sum of the water in the sample of fish tissue and the volume of extractant to the volume of aliquot. Using total mass rather than volume of aqueous phase in the scaling factor results in an overestimate of the TVB concentration, (or the concentration of other bases if their analyses are based on extracts), by an amount which depends on the ratio of muscle sample to extractant volume and the water content of the sample. For example a common ratio for extraction with TCA is 1 part fish:2 parts of TCA solution; if total mass is used the scaling factor it is 300/100, but is 280/100 if water masses are used and the water content is 80%, an overestimation by a factor of 1.07 when the total masses are used. Boury and Schvinte (1935) pointed out that the scaling factor should be based on the volume of water phase rather than total mass, but the advice has not always been 10

followed by other workers. In fact many of the descriptions of procedures used for determination of TVB do not include an explicit statement about calculation of concentration in the fish tissue, and even official or semi-official recommended procedures differ. Some papers on TVB in fish use what the authors refer to as the "Codex" method and cite a document presented to the Codex Committee on Fish and Fishery Products at its 3rd session in 1968 and referenced as Codex Fish 1/7. This document prescribes a scaling factor based on total mass. It is difficult to obtain a copy of this document because the report of the meeting is not readily available, but the experimental procedure is described in detail in Vyncke et al. (1987) B steam distillation of a TCA extract with sodium hydroxide. The description though does not give the formula for calculating the TVB concentration, but presumably the recommendation in the Codex document applies. The TVB procedure in the recommended methods for the analysis of fish products of the Analytical Methods Committee (1979) uses a scaling factor based on the volume of the water phase and this procedure is included, in summary, in Egan et al. (1981). The Official Methods of Analysis (Association of Official Analytical Chemists, 1995) does not have a method for TVB, but the recommended method for TMA determination, method 971.14, utilises a TCA extract as is used in TVB determinations, and the calculation for TMA concentration in the original sample uses a scaling factor based on total mass. FAO (1998) uses a scaling factor based on water volumes in the procedure for TVB in and cites Egan et al. (1981), but in the procedure for determination of TMA uses the mass ratio for calculating TMA citing the AOAC procedure (Association of Official Analytical Chemists, 1995). The procedure for TVB determination recommended in EU legislation (European Commission, 2005) uses distillation from a PCA extract with the scaling factor being based on total mass, though in this case the overestimation is negligible as the extract is made at a 1:9 fish:extractant ratio. 4. Distillation under reduced pressure Decomposition of nitrogen-containing substances during direct or steam distillation can be avoided, or at least reduced substantially, by carrying out the distillation at low temperature B compared with 1001C B under reduced pressure (Tillmans and Otto, 1924; Boury and Schvinte, 1935; Tomiyama and Harada, 1952; Tomiyama et al., 1956; da Costa et al., 1960; Hjorth-Hansen and Bakken; 1947; Pearson and Muslemuddin, 1968). The distilling bath is immersed in a water bath at perhaps 50 or 701C, but because of evaporative cooling the distilling solution will be below that of the water bath. Both muscle extracts and suspensions are amenable to distillation under reduced pressure and recovery of TVB is quite rapid; for example Pearson and Musselmuddin (1968) specify distillation of muscle suspension with MgO for 25 minutes at a pressure of 10-15mm of mercury and a bath temperature of 501 and Tomiyama et al. (1956) using a water bath at 701C reported that most of the TVB distilled over in 5 minutes. The disadvantages of distillation under reduced pressure rather than at atmospheric pressure are the more complex apparatus required and the longer time taken to dismantle and reassemble the apparatus between analyses. These disadvantages seem to deter potential users because distillation at reduced pressure does not appear to have been used outside of the laboratories describing the procedures. 5. The Microdiffusion (Conway) Method What is almost invariably referred to as the 'Conway' method is a popular procedure for measuring TVB judging from published reports; a survey, not comprehensive, by the reviewer more than 100 papers on TVB in fishery products revealed that almost half had used the method in the studies. Distillation takes place at room temperature or a little above in a circular microdiffusion cell B the Conway cell B which has a central well bounded by a wall 11

a little lower than the outer rim of the cells and forming an annular well. The sample, an extract of fish, is mixed with alkali in the annular well and acid is placed in the central well to absorb the evolved bases. The Conway cell seems to have been first used in studies on spoiling fish muscle by Beatty and Gibbons (1937) to measure TMA, but the procedure was soon adopted to measure TVB. The first explicit account of the use of the microdiffusion technique for TVB seems to be Stansby et al. (1944) who did not employ the Conway cell as such, but a similar cell assembled from other glassware. The Conway method probably enjoys its popularity because of the simplicity of the experimental procedure and the low cost of the equipment. The dwell time for complete absorption is two hours at room temperature when saturated potassium carbonate is used as the alkaliser or one hour at 371C, but these times are acceptable in practice. Montgomery (1960) described a rapid procedure based on the time taken to neutralise a fixed amount of acid. The drawbacks of the procedure reside in the care necessary in its execution for accurate and precise results. Because only small amounts of base are involved the analyst must be experienced in titrating from a microburette, for example in applying a thin layer of stopcock grease or similar to the tip of the burette to reduce the drop size and near the end point wiping less than a drop from the tip with the stirring rod. The glassware must be scrupulously clean even to the extent of cleaning the Conway cells in chromic acid between analyses, and the sealing grease must be carefully applied, or flamed, to exclude air bubbles. The standard sodium hydroxide for back titrating the absorbing acid B if standard mineral acid rather that boric acid is used B must be carbonate-free. General texts on the procedure, for example, Conway (1950) describe the principles and theory of absorption in the Conway cell and the practical requirements for accurate and precise analysis. Spinelli (1964) has discussed some of the precautions required for accurate results in the context of using the Conway method for measurement of TMA, but they apply equally to the determination of TVB. 6. Decomposition During Distillation It seems to have been known from early days of measurement of TVB in fish that ammonia is formed during distillation and is included in the value of TVB obtained in an analysis, and it was soon established that the amount of decomposition depended on both temperature of distillation and on the pH of the distilling mix (Boury and Schvinte, 1935; Hjorth-Hansen and Bakken; 1947). The general conclusions of those two studies for minimisation of decomposition ammonia were that intact muscle should be distilled under vacuum with MgO or weaker alkali and that protein-free extracts could be distilled at atmospheric pressure with MgO, but not with strong alkalis. Pearson and Musselmuddin (1968, 1969a, b) made a series of detailed studies on the rates of formation of decomposition ammonia during the determination of TVB in fish by distillation of muscle suspensions of salmon, plaice and haddock at 501C under reduced pressure and at atmospheric pressures with MgO as the alkaliser, (the latter being the classic L&G procedure). They measured the amount of bases distilled from the samples up to 60 minutes distillation with sampling of the distillate at intervals. The figures in the papers of the time course of distillation of bases show an initial curvilinear phase corresponding to the distillation of the intrinsic bases followed by linear phase after about 20 minutes of distillation corresponding to the continuing formation of ammonia by decomposition. The authors report that the rates of decomposition are very much less during distillation under reduced pressure than at atmospheric pressure, though not zero, and do not differ appreciably among the species. The reviewer has made a further analysis of the data by assuming a mathematical model for the course of distillation: 12

TVBt = A(1-expkt) + bt

(3)

where TVBt is the measured TVB concentration at distillation time t, A is the amount of intrinsic TVB and k and b are rate coefficients. The first term represents the distillation of the intrinsic TVB to an asymptote, A, and the second to the decomposition ammonia increasing linearly with time. (The decomposition should also be represented by an asymptotic expression, but over the time course of the experiment a linear expression is adequate). Data were read off the figures in Pearson and Musselmuddin (1968, 1969a, b) and the model fitted to each set, species by storage time by procedure, to estimate values of the coefficients that minimised the sums of squares of deviations of calculated values from the observed values. There were insufficient data values to accurately determine the errors of the estimates, and hence the significance of any differences, but within a procedure there was no suggestion that species or storage time influenced the rate coefficients, though, as would be expected, the estimates of A, the asymptote, increased monotonically with storage time. The mean values of k, the rate coefficient for distillation of intrinsic TVB, over all species and storage times were 0.25 mins-1 and 0.45 mins-1 for distillation at atmospheric and at reduced pressures respectively. Distillation by the procedure at reduced pressure is much faster than that at atmospheric though the procedure as described in Pearson and Musselmuddin (1968) specifies 25 minutes distillation, the same as that at atmospheric pressure. Using the rate constants quoted above it can be calculated that 99% of the intrinsic TVB is distilled over in 19 minutes at atmospheric pressure and 10 minutes at reduced pressure. The rate coefficients for the degradation reaction were 0.34 and 0.054 mg TVB (100g)-1 min-1 for distillation at atmospheric and reduced pressures respectively. These values are within the ranges quoted in the Pearson and Musselmuddin (1968, 1969a, b) papers for the rates at atmospheric pressures, and about the same as the upper value of the range, 0.02-0.05, for distillation at reduced pressure. Vyncke (1970) reports a similar rate, 0.4 mg TVBN (100g)-1 min-1, for distillation of muscle suspension with MgO for between 20 and 60 minutes in the Antona apparatus at atmospheric pressure, and unpublished data from Torry research Station, Aberdeen, UK, (Howgate personal experience), gave a value of 0.36 mg TVBN (100g)-1 min-1. Using the rates estimated from the Pearson and Musselmuddin (1968, 1969a, b) papers the amount of decomposition ammonia generated by a 25 minutes distillation time are 8.6 and 1.4 mg nitrogen at atmospheric and reduced pressures respectively. Distillation time at reduced pressure under the conditions described in Pearson and Musselmuddin (1968) could be reduced to 15 minutes with over 99% recovery of the intrinsic TVB with an associated reduction in the amount of decomposition ammonia nitrogen to 0.8mg. Pearson and Musselmuddin (1968) reported that raising the temperature of the water bath from 501 to 701C increased the rate of decomposition, but did not provide any data on the rate at the higher temperature. It would appear that distillation at a bath temperature lower than 501C would reduce the rate of formation decomposition ammonia even further, but with a concomitant requirement to increase in the distillation time. On balance there might not be any practical advantage to do so.

There seems no doubt that amides, glutamine and asparagine in particular, are the major source of the decomposition ammonia during distillation of whole muscle (Rehbein and Oehlenschlger, 1988). The mean amide content of the muscle protein fraction of four species of teleost fish reported by Connell and Howgate (1959) was 6.24mg amine N (100g)-1 protein N. Assuming the total nitrogen content of teleost fish muscle is around 2.8g (100g)-1 of which about 85% is protein nitrogen, the amide nitrogen content is then around 150mg (100g)-1 muscle tissue, very much more than is needed to account for the observed 13

decomposition ammonia in the TVB procedure. Muscle tissue of elasmobranch fish contains large amounts of urea which is readily decomposed to form ammonia by even mild alkalis and generally TVB measured by distillation at atmospheric pressure is not recommended as a measure of spoilage for this type of fish. However, Pearson and Musselmuddin (1969b) showed that distillation under reduced pressure using MgO as the alkaliser does not result in decomposition of urea and provides a true measure of intrinsic TVB in elasmobranch muscle. The formation of decomposition ammonia from amides in protein can be avoided by using protein-free extracts though other nitrogen-containing compounds in such extracts can decompose under appropriate conditions. The glutamine content of cod protein-free extracts was reported by Mackie and Ritchie (1974) to be 1.4mg (100g)-1 of muscle tissue, equivalent to 0.12mg amide nitrogen (100g)-1, a negligible amount in the context of TVB content. Pearson and Musselmuddin (1969) has a plot of the time course of TVB content of haddock TCA extract by the L&G procedure which shows no increase in TVB after the initial 20 minutes. Oehlenschlger (1988) showed that the higher the pH of the alkalised PCA extract, measured at room temperature, the greater the rate of deamination during distillation. He found the rate of decomposition when using NaOH as the alkali was 1.3 mg N (100g)-1 (50ml)-1 distillate in the case of redfish (Sebastes marinus) and 3.1mgN (100g)-1 (50ml)-1 distillate in the case of cod (Gadus morhua). The distillation rate in the Antona apparatus is typically 10 ml min-1 so the decomposition rates are 0.26 and 0.62 mgN (100g)-1 min-1 for redfish and cod respectively. Rehbein and Oehlenschlger (1982, 1988) reported using the L&G method, (presumably using MgO as the alkaliser though the paper is not explicit on this point), with the Antona unit to measure TVB in PCA extracts and also measured the ammonia and methylamine contents of the extracts. They found there was rapid formation of decomposition ammonia during the time the intrinsic TVB distilled over and some continuous formation of decomposition ammonia on further distillation which was almost negligible in fresh fish but greater in spoiled. Some unpublished results from Torry Research Station (Howgate, 1993) confirmed these findings of rapid initial formation of decomposition ammonia followed by slow continuous formation at a rate which was greater in stale than in fresh fish. 7. Other methods Measurement of TVB by distillation at room temperature has the advantage that there is no, or only negligible, formation of ammonia by decomposition of nitrogen-containing compounds. The distillation is commonly carried out in the Conway cell, but that has some disadvantages experimentally, as described above, mostly associated with its being a micro method. Alternatively the distillation can be effected on a macro scale by aeration of the alkalised mixture as Clark and Almy (1917) suggested in their early consideration of methods that had potential for assessing spoilage of fish. Egnr and Johansson (1938) has a discussion of theoretical aspects of aeration and Hjorth-Hansen and Bakken (1947) made further studies to compare theory and practice. Stansby et al., (1944) included aeration in their comparison of methods for determination of TVB, but otherwise the method has not been used in studies of TVB in fish. TVB can be measured by Flow Injection Analysis (FIA) (Wekell, et al., 1987; Hollingworth et al., 1990; Ruiz-Capillas and Horner, 1999; Baixas-Nogueras et al., 2001) in which the bases are partitioned from an alkalised extract through a PTFE membrane into a carrier stream containing bromothymol blue indicator. The change in colour of the indicator is monitored and provides a measure of the total bases. An advantage of the procedure is that it is carried out at room temperature and should measure the intrinsic TVB content. Pivarnik et al. (1998) tested the ammonia electrode in its standard configuration as a TVB electrode and 14

obtained a good correspondence between measurements of TVB in fish by electrode and by distillation. A collaborative trial of the method (Ellis et al., 2000) revealed that some laboratories experienced practical difficulties with the procedure, and there were moderately large discrepancies among results from the participating laboratories. There have been proposals for simple devices for measuring TVB in head spaces by colour changes of membranes impregnated with a suitable pH-sensitive dye (Hungerford et al., 1997; Loughran and Diamond, 2000), but they do not appear to have been adopted in practice. 8. Relationship Between Methods The measured TVB content consists of the intrinsic TVB in the sample plus the decomposition ammonia formed during distillation. The results of determination of TVB by two methods on the same samples should then be related by the expression TVB1 = TVB2 + c (4)

where c is the difference in the amount of decomposition by the two methods. If a matrix of values of c were determined for pairs of procedures then values determined by one procedure could be corrected to those expected of another. Such an approach would suppose that c does not also depend on species of fish or type of fish product as would happen if the components that decompose to form ammonia in some procedures varied by species or product. Vyncke (1970) compared 200 paired values for TVB content, (species of fish not stated), by the L&G and the Codex methods and calculated the regression equation Y = 0.93X + 5.30, where Y is TVB content by the Codex method and X the corresponding value by the L&G method. The regression coefficient is appreciably different from the value of 1.0 to be expected from the expression quoted earlier, but, as has been pointed out above, the Codex procedure uses a scaling factor based on total mass which overestimates the true concentration. A 1:2 fish:extract ratio is used in the Codex method which leads to a 7% overestimate of the TVB concentration assuming an average water content in the samples of 80%, and when the Y values, Codex method, are corrected by this factor the regression equation becomes Y = 0.99X + 5.67. This regression coefficient is practically not different from 1.0 and it seems from this data set that the Codex method, and by extension any procedure based on steam distillation of a protein-free extract, is biased by approximately +6mgN (100g)-1 compared with the L&G procedure. Pearson and Muslemuddin (1971) compared results of TVB determination by distillation under reduced pressure (Pearson and Muslemuddin, 1968) and by the L&G procedure for 15 samples of fish covering three species, cod, haddock and plaice, and expressed the relationship as the exponential equation Y = 0.223X1.31 where Y is the value by distillation under reduced pressure and X those the L&G procedure. The authors used this expression to convert quality class limits based on the L&G procedure to corresponding values using vacuum distillation. The authors do not report the original data on which the expression was derived, but earlier papers (Pearson and Muslemuddin, 1968; 1969a) between them tabulate nine sets of data obtained by the L&G, vacuum distillation, and Conway procedures on the same samples of three species of fish, salmon, haddock, and plaice. (The determinations by the Conway method were carried out on a TCA extract and the authors report they corrected the results for water content of the samples so that they corresponded to those by the direct distillation procedures). Calculating the regression coefficients by the usual least squares procedure when both variable variables are subject to observational error, as they are here, results in a biased, smaller, estimate of the regression coefficient, and instead the reviewer calculated the structural relationship between the three pairs of relationships (Kendall and 15

Stuart, 1967). In all cases the slopes of the structural relationships were not significantly different from 1.0 and in the case of Conway and distillation under reduced pressure the constant, c in the expression above, was not significantly different from 0.0. The constants for the relationships of L&G against either the Conway or the distillation under educed pressure procedures were significantly different from 0.0 and the mean bias of L&G compared with the other two procedures was +11.9mgN (100g)-1 of sample. The reviewer also fitted the data of the L&G and the distillation under reduced pressure methods to Pearson and Musselmuddin's (1971) exponential model, again as a structural relationship. The exponential model resulted in a slightly lower residual variance compared with the linear model, but when tested as an F ratio the difference was not significantly different (p=1.9) and the simpler linear model can adequately represent the relationship between the methods. Stansby et al. (1944) used 60% ethanol to prepare protein-free extracts and measured TVB in these extracts by the Conway method and by direct distillation using sodium borate as the alkaliser. They compared TVB contents of nine samples of silver salmon stored in ice or at room temperature by these two procedures and also on press juice from the same samples using the Conway procedure. The data are shown in Table 1 of the paper and were used by the reviewer to calculate the structural relationship between pairs of procedures. Results for press juice by the Conway method were expressed on a 100ml of press juice basis in the paper and were corrected to a 100g basis assuming the fish contained 80% water. For all three comparisons the structural relationships were statistically significantly different from 1.0 suggesting systematic differences in what was being measured by the three procedures and not just biases arising from decomposition. Davidovich and Giannini (1984) measured TVB in Patagonian hake (Merluccius hubbsi) stored for up to nine days in ice by the L&G and Conway methods and obtained the relationship Y = 2.37X + 5.27 where Y is TVB content by the L&G method and X the corresponding value by the Conway method. The regression coefficient is much greater than is to be expected from the simple linear relationship above, and judging from the scatter diagram of values by the two methods included in the paper the error bounds can not include a value of 1.0 even if the structural relationship were calculated. It is possible there is a systematic error operating here and it is worth noting that the higher values of TVB by the L&G method shown in a figure in the paper, around 33mgN (100g)-1, are about twice those reported by Lupin et al. (1980) for the same species stored in ice for nine days. Malle et al. (1989) compared TVB values determined by rapid distillation of a TCA extract with NaOH and by the Conway method for 261 samples and obtained a regression of Y = 1.017X + 0.85 where Y is TVB content by the rapid distillation procedure and X the corresponding value by the Conway method. Both procedures utilised TCA extracts and the authors used scaling factors based on the mass ratio rather than volume ratio. This would not have affected the value of the regression coefficient if the same extraction ratio had been the same in both, but they used a ratio of 1:2 for the rapid distillation method and 1:1 for the Conway method. Assuming on average the samples contained 80% water, (the authors do not give information about the species used other than they were marine fish), and correcting by the scaling factors based on volume ratios for the two procedures gives a regression coefficient of 0.982. This is very close to the expected value of 1.0, especially bearing in mind that the least squares regression procedure provides a biased, and lower, value of the coefficient compared to that of the structural relationship. Botta et al. (1984) compared six methods for determining TVB by measuring the TVB contents of the same samples of ice-stored cod during storage. The procedures were direct distillation of muscle with MgO the essentially the L&G procedure, steam distillation of muscle with MgO in a rapid distillation unit, the Conway procedure on a TCA extract, distillation of a TCA extract with NaOH in a semi-micro distillation unit, distillation of an 16

acidic magnesium sulphate extract with NaOH under reduced pressure (Tomiyama et al., 1956), and distillation of an ethanolic extract with sodium borate as described by Stansby et al. (1944). The test material was cod stored in ice for up to 18 days and the TVB contents were measured on the same sample by each of the six procedures. The results are shown in Figure 2 of the paper and shows the TVB contents increasing exponentially with storage time following a dwell without any increase. (Changes in TVB during spoilage will be discussed in more detail in Part 2 of this review). This behaviour can be modelled by an expression of the form: Y = a + ek(t-d) (5)

where Y is the TVB concentration, a is the initial value of TVB at start of storage, k is the rate constant, t is the storage time, and d is the dwell before the exponential growth phase of TVB. The TVB contents were measured by the different procedures on the same sample of fish so if the value obtained consists of the intrinsic TVB content of the sample plus the ammonia formed by decomposition then values of the parameters k and d should be the same, within experimental error, but values of a will differ among procedures. Figure 2 of Botta et al. (1984) shows a scatter diagram of TVB contents against storage time, in days and the values were picked off the diagram and the model fitted to the data by minimising the sum of squares of deviations of the observed from the fitted values. Inspection of the curves in the original figure and consideration of the values of the parameters obtained from fitting the model suggested the results for the Stansby et al. (1944) procedure, distillation of an ethanolic extract, were anomalous compared with the others. The fitted value of a, the initial value, was 2.0mgN (100g)-1 compared with 7.4mgN (100g)-1 by the Conway procedure, the method that is considered not to produce any ammonia by decomposition, and the rate constant was appreciably lower than those of the other methods. Botta et al. (1984) record that recovery of a mixture of TMA and ammonia added to the fish sample was only 72% in the case of the Stansby et al. (1944) procedure. The results for this set of were therefor not included in further processing of the data. The estimated values of d and k for the other methods were very similar and a reduced model using the same values of these two parameters was fitted to all the data giving an estimate of d of 5.0 days and of k of 0.28 days-1. The combined residual mean square of this model was larger than that of the full model, but the difference, tested as an Analysis of Variance, was not significant. However a reduced model in which all three parameters were the same gave a significantly larger residual mean square compared with either of the previous models. These results supported the model described above that the differences between the TVB results obtained with the different methods, other than the Stansby et al. (1944) procedure, were in the initial values and were due to different amounts of decomposition ammonia. The lowest values of a were 7.4 and 8.7mgN (100g)-1 for the Conway and the distillation at reduced pressure methods. Botta et al. (1984) do not describe how results were calculated, but presumably they were on a mass ratio basis and these results, based on extracts would be slight overestimates compared to calculation on a volume basis. The two highest values were obtained by distillation of fish muscle with MgO; 18.1 mgN (100g)-1 by the L&G procedure, 15.3 by rapid steam distillation. The distillation time in the former method is 20 minutes compared with 10 minutes or less in the latter and the difference can be accounted for by the additional decomposition with the longer distillation time. The initial value by distillation of a TCA extract with NaOH was 12.8 mgN (100g)-1. Regulations in the European Union (EU) have set limits to TVB concentrations for some species of fish and the recommended method of analysis, steam distillation in a rapid distillation unit of a PCA extract at a ratio of 1:9, muscle:extractant, made alkaline with 17

NaOH, is set out in an EU Regulation (European Commission, 2005). The Regulation though allows for other procedures to be used and Vyncke (1996) compared the results obtained by the recommended procedure with those alternatives. He found a very high correlation between the procedures and presents figures for results of four pairs of them along with the regression equations. In all cases the regression coefficients are close to 1.0 and the intercepts close to 0.0, but the author does not list the errors of the parameters. The reviewer has further analysed the results by reading data from the figures and calculating the structural relationships and confidence bounds of the parameters. The results included some very high TVB concentrations and in view of the findings discussed above that the amount of decomposition ammonia might be greater in very stale fish compared with fresh the statistical analyses were restricted to comparisons in which TVB contents were less than 50 mgN (100g)-1. Vyncke (1996) used a Kjeltec Tecator unit for distillation of an extract according to the recommended method and compared the results with distillation in an Antona unit (Antonacopoulos, 1960). One set of data, Figure 1 in the paper, compared distillation of PCA extracts, the EU recommended procedure. The structural relationship of TVB contents using the Antona apparatus on TVB contents using the Tecator has a slope of 1.085 with 95% Confidence Limits (CL) of 0.96 and 1.22, that is, including 1.0. When the structural relationship is forced to have a slope of 1.0, the intercept is 2.93, that is, TVB determined using the Altona apparatus gives a value almost 3mgN (100g)-1 larger than when using the recommended EU procedure with the Tecator unit. Vyncke (1996) records that the distillation time using the Tecator unit was six minutes compared with 12 minutes using the Antona apparatus. As was discussed above decomposition ammonia is produced during distillation of protein-free extracts with NaOH and the cause of the bias of the Altona over the Tecator procedures can be attributed to the extra distillation time in the Altona procedure. The EU Regulation permits direct distillation of muscle alkalised with MgO in the Altona apparatus and TVB results by this procedure were compared with those using the Tecator unit. The structural relationship of Altona results on Tecator had a slope of 0.967 with 95% CL of 0.85 and 1.10, that is, again including 1.0. Forcing a slope of 1.0 on the structural relationship gives an intercept of 2.97. Again, bearing in mind the earlier discussion of the rate of formation of decomposition ammonia by distillation of muscle with MgO, this bias of the Altona over the Tecator can be accounted for by the longer distillation time when using the Altona apparatus. The direct distillation in the Altona apparatus was also compared with distillation of the corresponding PCA extract alkalised with NaOH and distilled in the Antona apparatus. The paper, Vyncke (1996), does not describe the details of the extraction procedure and the calculation of concentration, but presumably the procedure described in the EU Regulation was used. This, as discussed above, does not allow for the water content of the sample and would overestimate the TVB content derived from the PCA extraction by a factor of 1.02. When this correction is applied to the original data the structural relationship of direct distillation of muscle on distillation of PCA extract has a slope of 1.037 with 95% CL of 0.951 and 1.132. Forcing the structural relationship to have a slope of 1.0 gives an intercept of -1.38, that is direct distillation with MgO gives a TVB content 1.38mgN (100g)-1 higher than the distillation of a PCA extract does. The distillation conditions of the two methods are the same so this bias most likely can be attributed to a difference in the amount of decomposition ammonia produced. The fourth comparison in Vyncke (1996) is PCA extract as specified in the EU recommended method against a 1:3 fish:extractant in TCA, (Codex method), which is specified in the Regulation as an alternative. Again, as discussed above, the TVB content in the sample would be calculated on the mass ratio and would overestimate the TVB content on a flesh basis. After correcting values as given in the paper, Figure 4, for both extractants the slope of the structural relationship of TCA on PCA results was 1.097 with 95% CL of .975 and 1.235. When the structural relationship is forced to a slope of 1.0 the intercept is 0.38 with 95% CL of -0.12 to 0.89, that is includes a value of 0.0. 18

It appears from this Vyncke (1996) study that equation (4) above is a good model for the results with c depending on the material being distilled B intact muscle tissue or extract B and the distillation time. B. TMA 1. The Conway Method In the early decades of the last century TMA was determined by tedious analytical procedures based on separate measurement of ammonia and selective decomposition of the amines with nitrous acid in a TVB distillate (Weber and Wilson,1918; Okoloff, 1932; Boury and Schvinte, 1935; Hjorth-Hansen and Bakken, 1947). Many of the analytical procedures introduced since then have not been specific for TMA, but have used formaldehyde to suppress interference from ammonia and methylamines other than TMA. Beatty and Gibbons (1937) introduced a simple method of analysis using the Conway microdiffusion cell (Conway 1950) in which 1ml of press juice, the liquor obtained from minced fish muscle by mechanical pressure, is mixed with 0.5ml of formalin solution in the annular ring of the cell, the mixture made alkaline with saturated potassium carbonate and the liberated bases absorbed in standard acid in the central well. The excess acid is back titrated as usual to give the amount of bases distilled over. Under the experimental conditions specified the ammonia is complexed and does not distil over, but DMA is probably not completely, or at all, complexed and the method is probably not specific for TMA as Beatty and Gibbons (1937) themselves point out. The Beatty and Gibbons method for determination of 'TMA' is operationally the same as that of the Conway method for measuring TVB and has the same advantages and disadvantages as already discussed for the latter determination and if TVB is being measured by the microdiffusion procedure it is convenient to measure the former as well. Though many other procedures for determination of TMA have been described since Beatty and Gibbons (1937) the microdiffusion method is still in use; the reviewer is aware of at least 10 publications in research journals in or since 2000 which report its use. 2. The Picrate Method Dyer (1945) described a colorimetric method based on a simplified version of one proposed for determination of amines in blood (Richter et al., 1941). The principle of the Dyer (1945) procedure is that an aliquot of an extract of fish muscle is taken, formaldehyde added to prevent interference by ammonia, and the mixture made alkaline. The free bases are extracted into solvent, usually toluene, and reacted with picric acid to give a yellow-coloured picrate salt which can be measured in a photometer. The materials, equipment and expertise required for the picrate method are available in any food analytical laboratory and the procedure is suitable for use commercial and regulatory laboratories as well as in research. The most expensive item is a photometer, but a reasonably accurate estimate of optical density can be achieved by comparing the test sample against a suitable range of standards. Though the procedure for determination of TMA by the picrate method is quite straightforward some precautions must be taken to ensure accurate and precise results. The original procedure (Dyer, 1945) used saturated potassium carbonate as the alkaliser, but the combination of formaldehyde with potassium carbonate does not completely prevent partition of DMA into the solvent phase. Further studies showed that potassium hydroxide (KOH) was more efficient in suppressing interference from DMA and in giving higher recoveries of TMA (Hashimoto and Okaichi, 1957; Shewan et al., 1971; Tozawa et al., 1971; Keay and Hardy, 1972; Bullard and Collins,1980), but the there are other factors as well. The amines are not completely partitioned into the solvent phase and the fraction of TMA extracted depends on 19

the composition of the alkaliser. For example, Bullard and Collins (1980) found that 97% of the TMA was extracted into the toluene phase using 45% KOH, 80% using 25% KOH and 72% using 50% potassium carbonate. The effects of the different alkalisers is almost certainly a salting out effect, not an effect of alkalinity. They also found that none of the alkali/formaldehyde combinations completely suppressed the interference of DMA. 25% KOH gave the least interference of DMA with a colour yield for DMA relative to TMA of about 6% compared with about 13% for 45%KOH. Wong and Gill (1987) reported that the picrate procedure using 25% KOH overestimated TMA content compared with an High Performance Liquid Chromatography (HPLC) procedure and Prez-Villarreal and Howgate (1986) reported that the picrate procedure using 45%KOH overestimated TMA by 20% compared with a GLC procedure. It appears from the various studies of the effect of the nature and concentration of alkali in the picrate procedure that 45% KOH should be used in situations where concentrations of DMA are expected to be low compared with TMA, for example in chill-stored fish, because this gives a more complete recovery of the TMA and the procedure is more sensitive, but where DMA is expected to be high, for example in frozenstored fish, 50% potassium carbonate should be used to reduce interference from DMA at the expense of some loss of sensitivity to TMA. Judging from publications in research journals when the picrate procedure is used in studies of TMA content in fishery products the practice is about evenly divided between using potassium carbonate or KOH as the alkaliser. It is perhaps worth noting that the method for 'Trimethylamine nitrogen in seafood' recommended in the AOAC Official Methods of Analysis specifies saturated potassium carbonate as the alkaliser, whereas the procedure specified by the Analytical Methods Committee uses 45% KOH It is usual to standardise the picrate procedure for each batch of analyses by preparing a calibration curve at the time using standard solutions of TMA. Bearing in mind that extraction of TMA into the solvent phase is not complete, especially when using 25% KOH or potassium carbonate, it is important that the working standards be made up in a solution that replicates the solute contents of test extracts. For example if a 1:3, sample:extractant, extract is prepared in 10% TCA and the sample is 80% water then the working standards should be made up in 8% TCA. Formaldehyde in near neutral solutions reacts with salts of ammonia and MMA to release equimolar amounts of acid and this reaction can be utilised to estimate TMA by an extension of the TVB procedure in which formaldehyde is added to the neutralised distillate from a TVB determination and titrating further the acid released (Boury and Schvinte, 1935; Analytical Methods Committee, 1979). It is often assumed in accounts of the reactions of formaldehyde with DMA that a similar reaction occurs with this amine as well, but according to Boury and Schvinte (1935) formaldehyde reacts with DMA salts, but does not release acid. The difference between the titrations of a TVB distillate before and after addition of formaldehyde therefore estimates the amounts of DMA and TMA. However, DMA is usually a minor component of the volatile bases in spoiling fish and the procedure would usually provide an adequate estimate of TMA content for many purpose in quality control of fishery products. The titrations need to carried out with care to give accurate results. The volumes of solution before addition of formaldehyde should be kept as small as possible and carbonatefree alkali should be used in the titrations. A pH indicator changing colour at pH 7.0 should be used and it might be useful to have a buffer solution at pH 7.0 with the indicator added as a reference to match the end points. 3. Distillation with Formaldehyde 20

If it is assumed that formaldehyde fixes ammonia and methylamines other than TMA then it would seem reasonable that adding formaldehyde to the distilling mix in the determination of TVB should allow only TMA to distill over. Beatty and Gibbons (1937) refer to trying distillation in the presence of formaldehyde as a procedure for measuring TMA in fish, but do not give any results of their trials. The principle was investigated by Benoit and Norris (1942) who, in the introduction to their paper, cited publications going back to the end of the 19th century which had used distillation in the presence of formaldehyde as a means of separating mixtures of methylamines. Benoit and Norris (1942) distilled solutions of ammonia and methylamines with formaldehyde under reduced pressure at 301C using sodium carbonate as the alkaliser. They showed that TMA was quantitatively recovered at all concentrations of formaldehyde used and that ammonia did not distill over. The behaviours of DMA and MMA were anomalous in that at low concentrations of formaldehyde distillation of these bases was partially suppressed, but as its concentration was increased a higher proportion of these bases distilled over. The procedure then is not specific for TMA B though they used it in a study of TMAO content of various species of aquatic animals (Norris and Benoit, 1945) B and the reviewer is not aware of the procedure being otherwise used for determination of TMA in fishery products. Malle and Tao (1987) later revived the principle of the Benoit and Norris (1942) procedure, (though without citing that or other papers), by distilling a TCA extract with formaldehyde at atmospheric pressure using sodium hydroxide as the alkaliser. They do not report recoveries from model solutions of methylamines, but compared TMA values of fish samples obtained by their procedure with those obtained by the microdiffusion or the picrate procedures. There is a high correlation between the two sets of data, but fitting a structural relationship to the results reproduced in the paper shows that the Malle and Tao procedure overestimates the TMA content compared with the microdiffusion procedure by about 22% over all the samples. The reviewer's own experience (unpublished) of the Malle and Tao method with solutions of ammonium chloride shows that the ammonia is not completely fixed and the proportion distilling over is very dependent on the alkalinity of the distilling mix. Even at low alkalinities, (but sufficient to allow for rapid distillation of amines), about 15% of the ammonia comes over and at the alkalinity expected from the amounts of extract and alkali used by Malle and Tao (1987) 25% distilled over. Though the amine/formaldehyde reactions summarised in the paper occur in approximately neutral solutions at room temperature the reactions occurring under the distillation conditions of the Malle and Tao (1987) procedure will be more complex. The formaldehyde would decompose under the hot alkaline conditions by the Cannizarro reaction at least and lose its effectiveness, and decomposition products would undergo the Mannich reaction with the dimethylamine with the possible release of other amines. 4. Gas Liquid Chromatography Amines were amongst the first groups of compounds to be separated by gas liquid chromatography (GLC) when the technique was introduced (James et al., 1952) and was soon applied to the determination of amines in fishery products (Groninger, 1958; Hughes, 1958; Hughes,1959). Since then a variety of analytical procedures and column packings for effecting the separation have been described in the literature of determining amines in fishery products. Amines tend to 'tail' and variations in column packings tend to be directed towards reducing this effect. A more fundamental difference in procedures is in the treatment of the sample before injection on the column. Earlier procedures (Groninger, 1958; Gruger, 1962; Hughes, 1958; Hughes,1959) injected the distillate from a TVB-type distillation directly on to the column. This approach has the disadvantage that injection of water onto the columns 21

usually has an adverse effect on their performance. This can be avoided by extracting the amines from an aqueous extract into an organic solvent and injecting an aliquot of that (Nonaka, et al., 1967; Keay and Hardy, 1972; Tokunaga, et al., 1977; Lundstrom and Racicot, 1983; Perez-Martin, et al., 1987; Krzymien and Elias, 1990; Oetjen and Karl, 1999). The effects of incomplete partition of the amines into the solvent already discussed above with regard to the picrate picrate method apply to this procedure as well and recoveries of amines using different solvents are measured and discussed in Lundstrom and Racicot, (1983). The analysis of headspace vapours above an alkalinised suspension or extract of a sample avoids the need for a distillation or extraction step (Miller, et al., 1972; Kruse and Stockemer, 1989; Krzymien and Elias, 1990; Fiddler, et al., 1991). The headspace of course will contain volatiles other than amines and Miller, et al., (1972) used a nitrogen detector to reduce interference from non-nitrogen-containing compounds, but Kruse and Stockemer, (1989) and Fiddler, et al., (1991) in comparing the nitrogen detector with standard FID found the latter was satisfactory. An extension of the simple head space sampling is to use solidphase microextraction (Bn et al., 2001; Chan et al., 2006). Chan et al. (2006) used their procedure to measure TMA in two species of freshwater fish and found high, >30 mgN (100g)-1, in chill-stored samples, which is very surprising bearing in mind that generally freshwater fish does not contain TMAO and produce TMA during storage. They also found similarly high concentrations of TMA in samples stored at -201C, which is also very surprising bearing in mind that TMA is formed in fish muscle by bacterial action B assuming TMAO is present B and bacterial growth would not occur at this temperature. It is difficult to avoid the conclusion that the results point to an artefact in the Chan et al. (2006) procedure. 5. Other Methods FIA already mentioned earlier with regard to measurement of TVB can be adapted to measure TMA by adding formaldehyde to the flow steam (Zhi et al., 1995; Sadok et al., 1996; Ruiz-Capillas and Horner, 1999; Baixas-Nogueras et al., 2001). Ammonia is adequately complexed by the procedure and Sadok et al. (1996) reported that interference from DMA in this system is negligible, (colour yield about 1/60th that of an equimolar amount of TMA), and FIA procedures incorporating formaldehyde appear to be specific for TMA in fish extracts. Chang et al., (1976) report using the ammonia electrode to measure TMA by modifying the filling solution to contain TMA rather than ammonia. Under the conditions specified in the paper ammonia gives a negligible response, though DMA gives a response which is less 10% that of TMA depending on the concentration of formaldehyde used. Mitsubayashi et al., (2004) have described a FIA procedure for TMA in fish which uses an immobilised enzyme biosensor as the detector. Other procedures for measuring TMA in fishery products have been proposed based on a variety of principles, though applications of them do not seem to have been reported other than in the original papers describing the procedures; they include: HPLC (Charest and Dunn, 1984; Gill and Thompson, 1984; Monser and Greenway, 1996; zogul et al. 2002); enzymatic (Wong and Gill, 1987; Wong et al., 1988); indicator films (Loughran and Diamond, 2000; Byrne et al., 2002); solid state and biosensors (Storey et al., 1984; Gamati et al., 1991; Loechel, et al., 2003; Mitsubayashi et al., 2005; Zhang et al., 2005); and capillary electrophoresis (Timm and Jrgensen, 2002). C. DMA Reay (1937) alluded to a method for determination of DMA as the coloured copper dithiocarbamate and Beatty and Collins (1940) provides a description of the procedure. The Beatty and Collins (1940) method was later modified by Dyer and Mounsey (1945) and their 22

procedure has been in use since then for the determination of DMA. In the Dyer and Mounsey (1945) procedure an extract of fish muscle is reacted with ammoniacal copper sulphate and carbon disulphide to give the dithiocarbamate which is extracted into benzene and the colour measured to give the DMA content. Benzene is toxic and can be replaced by other solvents (Mackie and Thomson, 1974). DMA interferes with the picrate method for determining TMA because it partitions to some extent into the solvent phase, the extent depending on the alkali used in the procedure, (see the discussion above) and this behaviour was exploited by Castell, et al., (1974) to determine DMA in fish extracts. Colour yields, absorbance/unit concentration, are determined for both TMA and DMA using 50% potassium carbonate and using 25% KOH as the alkali. These factors are used in a pair of simultaneous equations to calculate DMA and TMA concentrations in test samples analysed using the two alkalis. This method depends upon four factors being accurately and precisely measured and bearing in mind the usual experimental errors associated with the picrate method and the generally low concentrations of DMA found in spoiling fish it has the potential to be prone to high relative errors; judging from the research literature the method has been very little used in studies on fish quality compared with the dithiocarbamate method. Any of the GLC or HPLC methods already cited for measurement of TMA will also measure DMA and Mackie and Thomson (1974) showed that results of DMA by both the copper dithiocarbamate and the a GLC procedure gave comparable results. D. Ammonia Ammonia can be determined in the distillate from TVB determination using Nessler's reagent or by formol titration, but the result will not measure the true ammonia content of the sample unless the distillation has been carried out at low-temperature using a mild alkali. Ammonia can be determined in aqueous or protein-free extracts of tissue using the enzymatic method attributed first to da Fonseca-Wollheim (1973). The experimental procedure is described in Cheuk and Finne (1984), though most publications just cite the relevant pages of the current edition of the compilation of enzymatic methods (Bergmeyer, 1983-6). Several suppliers of laboratory chemicals offer kits for determination of ammonia in biological materials by the enzymatic method and the method is by far that most frequently used for the specific determination of ammonia in fishery products. The procedure requires use of a spectrophotometer reading in the UV, but Leblanc and Gill (1984) modified it to produce a colour readable in the visible region. Ammonia can be separated from methylamines by GLC, but the problem with this approach lies in detection because the FID commonly used with GLC systems does not respond to ammonia and the procedures described in papers cited above relating to determination of TMA by GLC do not measure ammonia. Sze et al., (1963) describe the separation of ammonia and methylamines by GLC though they do not specify the detector; given the date of the publication it was probably a conductivity detector. A similar situation applies to direct determination of ammonia by HPLC B it does not absorb in the UV range of detectors typically used in HPLC systems unless derivatized. For example, Lin and Lai (1980) described the measurement of ammonia and methylamines in fishery products as their dabsyl derivatives. IV. DISCUSSION Though measurement of TVB has been used in studies of spoilage of fishery products for almost a century by now there is not an universally accepted analytical procedure for the measurement. The main defect in the analytical procedures is that ammonia is produced from nitrogen-containing substances in the fish tissue or in extracts of tissue unless the distillation is carried out below about 501C. Any decomposition ammonia is additional to the intrinsic 23

ammonia present in the sample which leads to an overestimation of the true TVB concentration in the sample. Different procedures result in different amounts of decomposition ammonia being produced and though results from methods might be linearly related there are biases between them. Procedures in which distillation is performed at room at normal atmospheric pressure, such as the Conway method, or under vacuum give the true TVB concentration. The various methods that have been proposed and used are mainly derived from empirical trials and study of the physical chemical principles involved in the distillation of the bases comprising TVB suggests the procedures are not optimum for time to complete the distillation while minimising decomposition. The completion of distillation should be based on the volume distilled over rather than distillation time and the volume of the mix being distilled should be kept as small as possible to minimise the volume of distillate required for complete recovery of the bases. The amount of decomposition ammonia is linearly related to the distillation time and distillation rate should be as fast as possible in the apparatus used to minimise the time. A survey of the published literature on TMA in spoiling fish shows that the most frequently used analytical methods are the Conway procedure and the picrate procedure, though many others have been proposed. The picrate method is not completely specific for TMA and a proportion of any DMA in the sample is included with the TMA, the proportion depending on the combination of alkali and formaldehyde concentrations used. Extraction of amine into the organic phase is not complete under the conditions specified in protocols for the analysis, though this fact is not referred to in the protons, and care must be taken that the extraction step is standardised and consistent, and applied in the same way to the standards used for calibration. Of the other methods that have been used in studies of TMA in fishery products, GLC procedures have the advantage of being specific for TMA, (and other methylamines), though the high capital and running costs of the equipment and the overheads of time for preparing, conditioning and maintaining columns militate against their use other than in research environments. FIA methods appear promising as they can measure both TVB and TMA by a small change in procedure, and the TVB measured is the true intrinsic concentration. The capital and running costs of FIA methods are markedly less than those of GLC, but unfortunately there are only few reports of the use of this procedure in studies of spoiling fish. It is uncommon to measure ammonia in studies of spoiling fish B teleost fish anyway B which is unfortunate because in some circumstances of fishing ground, source B wild or farmed - and season ammonia can increase more than TMA during spoilage. Ammonia can be estimated by subtraction of TMA and DMA concentrations from TVB, (volatile amines other than TMA and DMA are absent or of negligible amount in spoiling fish), but this approach is not accurate unless the analytical procedure used does not produce decomposition ammonia. Ammonia can be measured along with other amines by GLC, but an appropriate detector must be used because the FID typically used in GLC is not sensitive to ammonia. It can be determined by a specific enzymatic procedure. Only small amounts of DMA are formed in fish muscle during spoilage, and not in all species, and its measurement is not considered useful in monitoring spoilage. I can be measured by GLC, but otherwise by a wet chemical method as copper dithiocarbamate. Methylamines, and in some methods, TVB, are measured on protein-free extracts and protocols for analysis differ as to how concentrations measured in an aliquot are scaled up to concentrations in the sample. Some scale up by the total mass of the macerate, for example 100g of sample plus 200ml of extractant, and some by the volume of the water phase in the 24

macerate, for example 80ml assuming 80% water content in the sample plus 200ml of extractant. The former would overestimate the concentration in the sample by a factor of 1.07 compared with the latter. It is disappointing that after nearly a century of measuring volatile bases in spoiling fish that analytical procedures for measuring TVB and TMA, the two measures that have most been studies and been used as indices of spoilage, have not been standardised. Even the procedure for calculating sample concentration from measurements on extracts has not been standardised. The errors and biases in results of analyses by the various procedures inevitably raise doubts about the precision and accuracy of published data, and diminish the reliability of measurements when used in commercial quality control or in official inspection.

25

REFERENCES Analytical Methods Committee, (1979), Recommended general methods for the examination of fish and fish products. Analyst, 104, 434-450. Anderson, A.G. (1908). On the decomposition of fish. 26th Annual Report of the Fishery Board for Scotland, 1907, Part III, Scientific Reports. Pp13-39. HMSO, London. Antonacopoulos, N. (1960), Improved apparatus for quantitative distillation of steam volatile substances. (Verbesserte Apparatur zur quantitativen Destillation wasserdampfflchter Stoffe). Z. Lebensm. -Unters. -Forsch., 113, 113-116. Antonacopoulos, N. (1968): In: Handbuch der Lebensmittelchemie, Bd. 111/2. Pp 1493-1494. I. Acker, ed. Springer Verlag, Berlin. Antonacopoulos, N. (1973). In: Fische und Fischerzeugnisse Pp 224-225. W. Ludorff and V. Meyer, eds. Paul Parey, Berlin Antonacopoulos, N. and Vyncke, W. (1989). Determination of volatile basic nitrogen: a third collaborative study by The West European Fish Technologists ' Association (WEFTA). Z. Lebensm. -Unters. -Forsch., 189, 309-316. Association of Official Analytical Chemists, (1995). Official Methods of Analysis of AOAC International, 16th ed.. AOAC International, Arlington, Va. Baixas-Nogueras, S., Bover-Cid, S., Vidal-Carou, M.C., Veciana-Nogus, M.T., and Marin-Font, A.. (2001). Trimethylamine and volatile basic nitrogen determination by flow injection/gas diffusion in Mediterranean hake. J. Agric. Food Chem., 49, 1681-1686. Beatty, S.A. (1937). The measurement of spoilage of fresh fish. Prog. Rep. Atlantic Sta. Fish. Res. Board Can. no.53, 3-5. Beatty, S.A. (1938 ). Studies in fish spoilage. II. The origin of trimethylamine during the spoilage of cod muscle press juice. J. Fish. Res. Board Can., 4, 63-68. Beatty, S.A. (1939 ). Studies in fish spoilage. III. The trimethylamine oxide content of the muscles of Nova Scotia fish. J. Fish. Res. Board Can., 4, 229-232. Beatty, S.A. and Gibbons, N.E. (1937). The measurement of spoilage in fish. J. Biol. Board Can., 3, 77-91. Beatty, S.A. and Collins, V.K. (1940). Studies of fish spoilage. VII. Dimethylamine production in the spoilage of cod. J. Biol. Board Can., 5, 32-35. Bn, A., Hayman, A., Reynard, E., Luisier, J.L.and Villettaz, J.C. (2001). A new method for the rapid determination of volatile substances: the SPME-direct method. Part II. Determination of the freshness of fish, Sens. Actuators B, 72, 204-207. Benoit, G.J. and Norris, E.R. (1942). Effect of formaldehyde on the volatilizations of ammonia, mono-, di-, and trimethylamine. Ind. Eng. Chem., Anal. Edn, 14, 823-826.

26

Bergmeyer, H. U. Ed. (1983-6). Methods of enzymatic analysis, 3rd edn. Verlag Chemie, Weinheim. Botta, J.R. (1995). Evaluation of seafood freshness quality. VCH Publishers Inc., New York Botta, J.R., Lauder, J.T. and Jewer, M.A. (1984). Effect of methodology on total volatile basic nitrogen (TVB-N) determination as an index of quality of fresh Atlantic cod (Gadus morhua). J. Food Sci., 49, 734-736, 750. Boury, M. and Schvinte, J. (1935). Spoilage of fish. (L'altration du poisson). Rev. Trav. l'Office Pech. Maritim., 8, 282-333. (In French). Brocklesby, H.N. and Riddell, W.A. (1937 ). Chemical and biochemical studies of halibut. 2. Iced fish. Prog. Rep. Pacific Biol. Sta. Fish. Res. Board Can., No.33, 17-19. Bullard, F.A. and Collins, J. (1980 ). An improved method to analyse trimethylamine in fish and the interference of ammonia and dimethylamine. Fish. Bull., 78, 465-473. Byrne, L., Lau, K.T. and Diamond, D. (2002). Monitoring of headspace total volatile basic nitrogen from selected fish species using reflectance spectroscopic measurements of pH sensitive films. Analyst, 127,1338-1341. Castell, C.H., Smith, B. and Dyer, W.J. (1974). Simultaneous measurements of trimethylamine and dimethylamine in fish and their use for estimating quality of frozen stored gadoid fillets. J. Fish. Res. Board Can., 31, 383-89. Chan, S.T., Yao, M.W.Y., Wong, Y.C., Wong, T., Mok, C.S. and Sin, D.W.M. (2006). Evaluation of chemical indicators for monitoring freshness of food and determination of volatile amines in fish by headspace solid-phase microextraction and gas chromatography-mass spectrometry. Eur. Food Res. Technol., 224, 67-74 Chang, G.W., Chang, W.L. and Lew, K.B.K. (1976). Trimethylamine-specific electrode for fish quality control. J. Food Sci., 41, 723-724. Charest, R. and Dunn, A. (1984). Chromatographic separation of choline, trimethylamine, trimethylamine oxide, and betaine from tissues of marine fish. Anal. Biochem., 136, 421-424. Cheuk, W.L., Finne, G. (1984). Enzymatic determination of urea and ammonia in refrigerated seafood products. J. Agric. Food Chem., 32, 14-18. Clark, E.D. and Almy, L.H. (1917). Preliminary studies on chemical methods of detecting deterioration in fish flesh. J. Assoc. Off. Agric. Chem., 2, 231-236. Connell, J.J. and Howgate, P.F. (1959). The amino-acid composition of some British food fishes. J. Sci. Food Agric., 10, 241-244. Conway, E.J. (1950). Microdiffusion Analysis and Volumetric Error. Third edn. Crosby Lockwood and Son Ltd, London. Cook, A.S. (1931 ). A study of the occurrence of trimethylamine in marine animals. Can. Chem. Metall., 15, 22. da Costa, A.A.; Tomiyama, T. and Stern, J.A. (1960). The applicability of the vacuum 27

distillation technique for the determination of volatile acids and volatile bases in fish flesh. In: Chilling of Fish. Fish processing technologists meeting, Rotterdam, The Netherlands, 25-29 June, 1956. Pp244-250. Ministry of Agriculture, Fisheries and Food, The Hague, Netherlands. da Fonseca-Wollheim, F., Direct determination of plasma ammonia without removal of protein. (Direkte Plasmaammoniakbestimmung ohne Enteiweissung). (1973). Z. klin. Chem. klin. Biochem., 11, 426-431 Davidovich, L.A. and Giannini, D.H. (1984). Correlation between total volatile bases contents in fish determined by the distillation and the microdiffusion methods. Lebensmitt. -Wiss. -Technol., 17, 287-289. Dyer, W.J. (1945). Amines in fish muscle: I. Colorimetric determination of trimethylamine as the picrate salt. J. Fish. Res. Board Can., 6, 351-358. Dyer, W.J. and Mounsey, Y.A. (1945). Amines in fish muscle: II Development of trimethylamine and other amines. J. Fish. Res. Board Can., 6, 359-367. Egan, H, Kirk, R.S., and Sawyer, R. (1991). Pearson's composition and analysis of foods. Longman, Harlow, UK. Egnr, H. and Johansson, V.J. (1938). Distillation of ammonia. Ann. Agric. Col. Sweden, 5, 113-130. Ellis, P.C., Pivarnik, L.F. and Thuam, M. (2000). Determination of volatile bases in seafood using the ammonia on selective electrode: collaborative study. J. AOAC Int., 83, 933-943. Emerson, K., Russo, R.C., Lund, R.C. and Thurson, R.V. (1975). Aqueous ammonia equilibrium calculations: effects of pH and temperature. J. Fish. Res. Board Can., 32, 2379-2383 European Commission. (2005). Commission Regulation No 2074/2005 of 5 December 2005 laying down implementing measures for certain products under Regulation (EC) No 853/2004 of the European Parliament and of the Council and for the organisation of official controls under Regulation (EC) No 854/2004 of the European Parliament and of the Council and Regulation (EC) No 882/2004 of the European Parliament and of the Council, derogating from Regulation (EC) No 852/2004 of the European Parliament and of the Council and amending Regulations (EC) No 853/2004 and (EC) No 854/2004. Off. J. Eur. Comm., L338, 27-59. Annexe II, Section II, Chapter III. FAO (1998). Manuals of food quality control. Food analysis: quality, adulteration and tests of identity. FAO Food and Nutrition Papers 14/8. Pp139-142. Food and Agriculture Organisation, Rome. Fiddler, W., Doerr, R.C. and Gates, R.A. (1991). Gas chromatographic method for determination of dimethylamine, trimethylamine, and trimethylamine oxide in fish-meat frankfurters. J. Assoc. Off. Anal. Chem.,74, 400-403. Gamati, S., Luong, J.H. and Mulchandani, A. (1991 ). A microbial biosensor for trimethylamine using Pseudomonas aminovorans cells. Biosens. Bioelectron., 6, 125-31.

28

Gill, T.A. and Thompson, J.W. (1984). Rapid, automated analysis of amines in seafood by ion-moderated partition HPLC. J. Food Sci., 49, 603-606. Groninger, H. (1958). Fish spoilage. I. Determination of bacterial metabolites by gas chromatography. Comm. Fish. Rev., 20(11), 23-26. Gruger, E.H. (1972). Chromatographic analyses of volatile amines in marine fish. J. Agric. Food Chem., 20, 781 Hashimoto, Y. and Okaichi, T. (1957). On the determination of trimethylamine and trimethylamine oxide. A modification of the Dyer method. Bull. Jap. Soc. Sci. Fish., 23, 269-272. (In Japanese) Hebard, C.E., Flick, G.J., and Martin, R.E. (1982). Occurrence and significance of trimethylamine oxide its derivatives in fish and shellfish. In: Chemistry and Biochemistry of Marine Food Products. Pp 149-304. R.E. Martin, G.J. Flick, C.E. Hebard, and D.R. Ward eds. Avi Publishing Co., Inc., Westport, Co. Hess, E. (1932). The influence of low temperatures above freezing upon the rate of autolytic and bacterial decomposition of haddock muscle. Contrib. Can. Biol. Fish., 7, 147-163 Hillig, F., Shelton, L.R., Loughrey, J.H. and Eisner, J. (1958). Chemical indices of decomposition in cod. J. Assoc. Off. Agric. Chem., 41, 763-776. Hjorth-Hansen, S. (1952) Mthode pour le dosage de loxyde de trimthylamine. Anal. Chim. Acta, 6 ,438-441. (In French). Hjorth-Hansen, S. and Bakken, K. (1947). Investigations on analytical methods for estimation of ammonia and methylamines in fish. (Underskelser over analysemetoder for ammoniak og metylaminer i fisk). Reports from the Norwegian Fisheries Research Laboratory, 1, No. 6. (Fiskeridirektoratets Skrifter. Serie Underskelser ved Statens Fiskeriforsksstatsjon, 1, No. 6). (In Norwegian). Hollingworth, T.A., Wekell, M.M., Sullivan, J.K.: Torkelson, J.D. and Throm, H.R.. (1990). Chemical indicators of decomposition for raw surimi and flaked artificial crab. J. Food Sci., 55, 349-352. Howgate, P. (1993). Unpublished paper presented to the meeting of the West European Fish Technologist's Association in Hamburg, 1993. Hughes, R.B. (1958). Volatile amines of herring flesh. Nature, 181, 1281-1282. Hughes, R.B. (1959). Chemical studies on the herring (Clupea harengus). I. Trimethylamine oxide and volatile amines in fresh, spoiling and cooked herring flesh. J. Sci. Food Agric., 10, 431-436. Hungerford, J. Manger, R., Wekell, M. and Hollingworth, T. (1997). Rapid chemical tests: laboratory and field screening methods for seafood analysis. In: Fish Inspection, Quality Control, and HACCP: a Global Focus. Proceedings of the Conference held May 19-24, 1996, Arlington, Virginia, USA. Pp538-550. R.E. Martin, R.L. Collette and J.W. Slavin eds. Technomic Publishing Co., Inc., Lancaster, Pa, USA. 29

James, A.T., Martin, A.J.P., and Howard-Smith, G., (1952). Gas-liquid chromatography: the separation and microestimation of ammonia and methylamines. Biochem. J., 52, 238-242. Jones, N.R. (1960). The separation and determination of free purines, pyrimidines and nucleoside in cod muscle. Analyst, 85, 111-115. Keay, J.N. and Hardy, R. (1972). The separation of aliphatic amines in dilute aqueous solution by gas chromatography and application of this technique to the quantitative analysis of tri- and dimethylamine in fish. J. Sci. Food Agric., 23, 9-19 Kendall, M.G. and Stuart, A. (1967). Chaper 29. Functional and structural relationships. The Advanced Theory of Statistics. Vol. 2. Pp 375-416. Charles Griffin and Co. Ltd., London. Kruse, R. and Stockemer, J. (1989). Application possibilities of headspace-gas chromatography for investigations on fishes and shrimps: Determination of mono-, di- and trimethylamines. Arch. Lebensmittelhyg., 40, 87-89. Krzymien, M.E. and Elias, L. (1990). Feasibility study on the determination of fish freshness by trimethylamine headspace analysis. J. Food Sci., 55, 1228-1232. LeBlanc, R.J. and Gill, T.A. (1984). Ammonia as an objective quality index in squid. Can. Inst. Food Sci. Technol. J., 17, 195-201. Lin, J-K. and Lai, C-C. (1980). High Performance Liquid Chromatographic determination of naturally occurring primary and secondary amines with dabsyl chloride. Anal. Chem., 52, 630-635. Loechel, C., Basran, A., Basran, J., Scrutton, N.S. and Hall, E.A.H. (2003). Using trimethylamine dehydrogenase in an enzyme linked amperometric electrode. Part 1. Wild-type enzyme redox mediation. Analyst, 128, 166-172. Loughran, M. and Diamond, D. (2000). Monitoring of volatile bases in fish sample headspace using an acidochromic dye. Food Chem., 69, 97-103. Lcke, F. and Geidel, W. (1935). Determination of volatile basic nitrogen in fish as a measure of their freshness. (Bestimmung des flchtigen basischen Stickstoffs in Fischen als Mastab fur ihren Frischezustand). Z. Unters. Lebensmitt., 70, 441-458. (In German). Lundstrom, R.C. and Racicot, L.D. (1983 ). Gas chromatographic determination of dimethylamine and trimethylamine in seafoods. J. Assoc. Off. Anal. Chem., 66,1158-1163. Lupin, H.. M.., Giannini, D. H., Soule, C. L., Davidovich, L. A. and Boeri, R. L. (1980). Storage life of chilled Patagonian hake. J. Food Technol., 15, 285-300. Mackie, I.M. and Ritchie, A.H. (1974). Free amino acids of fish flesh. In: Proceedings of the IVth International Congress of Food Science and Technology, Madrid, Spain, 23-27 September, 1974. Pp 29-38. Instituto de Agroqumica y Tecnologa de Alimentos, Valencia. Mackie, I.M. and Thomson, B.W. (1974). Decomposition of trimethylamine oxide during iced and frozen-storage of whole and comminuted tissue of fish. Proceedings of the IVth 30

International Congress of Food Science and Technology, Madrid, Spain, 23-27 September, 1974. Pp. 243-250. Instituto de Agroqumica y Tecnologa de Alimentos, Valencia. Malle, P. and Tao, S.H. (1987). Rapid, quantitative determination of trimethylamine using steam distillation. J. Food Prot., 50, 756-760. Malle, P., Vanelle, A.M. and Petit, A. (1989). Total volatile nitrogen rates in salt water fish muscle. Data for standardisation of the determination, expression and use of T.V.B.N. (Teneur en azote basique volatil total du tissu musculaire des poissons marins. Elements pour une normalisation de la determination, de l'expression et de l'exploitation de l'ABVT). Recl. Med. Vet., 165, 395-402. (In French). Miller, A., Scanlan, R.A., Lee, J.S. and Libbey, L.M. (1972). Quantitative and selective gas chromatographic analysis of dimethyl- and trimethylamine in fish. J. Agric. Food Chem., 20, 709-711. Mitsubayashi, K., Kubotera, Y., Yano, K., Hashimoto,Y., Kon, T., Nakakura, S., Nishi,Y. and Endo, H. (2004). Trimethylamine biosensor with flavin-containing monooxygenase type 3 (FMO3) for fish-freshness analysis. Sens. Actuators B, 103, 463-467. Monser, L.I. and Greenway, G.M. (1996). Liquid chromatographic determination of methylamines. Determination of trimethylamine in fish samples using a porous graphitic carbon stationary phase. Anal. Chim. Acta, 322, 63-68. Montgomery, W.A. (1960). A rapid method for the estimation of volatile bases in fish muscle. In: Chilling of Fish. Fish processing technologists meeting, Rotterdam, The Netherlands, 25-29 June, 1956. Pp 234-243. Netherlands: Ministry of Agriculture, Fisheries and Food, The Hague. Nichols, M.S. and Foote, M.E. (1931). Distillation of free ammonia nitrogen from buffered solutions. Ind. Eng. Chem., 3, 311-313. Nonaka, J., Mitani, H. and Koizumi, C. (1967). Determination of volatile amines in fish muscle by gas liquid chromatography. I. Trimethylamine. Bull. Jap. Soc. Sci. Fish., 33, 753-757. Norris, E.R. and Benoit, G.J. (1945). Studies on trimethylamine oxide I. Occurrence of trimethylamine oxide in marine organisms. J. Biol. Chem., 158, 433- 438. Oehlenschlger, J. (1988). The determination of volatile basic nitrogen (TVB-N). Effect of pH on the deamination of acid extracts of sea fish during steam distillation. Informationen fr die Fischwirtschaft, 35, 31-34. Oehlenschlger, J. (1997). Volatile amines as freshness/spoilage indicators. A literature review. In: Seafood from producer to consumer, integrated approach to quality (edited by J.B. Luten, T. Brrensen, and J. Oehlenschlger). Pp 571-584. Elsevier, Amsterdam. Oehlenschlger, J. and Rehbein, H. (1982 ). The chemical composition and analysis of volatile base nitrogen in fish. (ber die Zusammensetzung des fluechtigen Basenstickstoffs in Fischen). Informationen fr die Fischwirtschaft, 29, 33-34.

31

Oetjen, K. and Karl, H. (1999). Improvement of gas chromatographic determination methods of volatile amines in fish and fishery products. Deut. Lebensm.-Rundsch., 95, 403-407. Okoloff, F. (1932). Determination of ammonia, trimethylamine and other amines in foods. (Bestimmung von Ammoniak, Trimethylamin und anderen Aminen in Nahrungsmitteln). Z. Lebens. Unters. Forsch., 63, 129-154. (In German). lafsdttir, G., Luten, J., Dalgaard, P. Careche, M. Verrez-Bagnis, V. Martinsdottir, E. and Heia, K. (1997). In: Methods to determine the freshness of fish in research and industry. Proceedings of the final meeting of the concerted action "Evaluation of fish freshness" AIR3CT94 2283, Nantes, Nov. 12-14 1997. International Institute of Refrigeration, Paris. zogul, F., Taylor, K.D.A., Quantick, P. and zogul, Y. (2002). Biogenic amines formation in Atlantic herring stored under MAP using a rapid HPLC method with improved derivatization procedures. Int. J. Food Sci. and Technol., 37, 515-522. Pearson, D. and Muslemuddin, M. (1968). The accurate determination of total volatile nitrogen in meat and fish, Part I. techniques and application to beef and salmon. J. Assoc. Public Anal., 6, 117-123. Pearson, D. and Muslemuddin, M. (1969a). The accurate determination of total volatile nitrogen in meat and fish. Part II application to white teleostean fish. Journal of the Association of Public Analysts, 7, 50-54. Pearson, D. and Muslemuddin, M. (1969b). The accurate determination of total volatile nitrogen in meat and fish. Part III. Application to elasmobranch fish. J. Assoc. Public Anal., 7, 73-82. Pearson, D. and Muslemuddin, M. (1971). The accurate determination of total volatile nitrogen in meat and fish. Part IV. Calculated critical values for beef and white fish. J. Assoc. Public Anal., 9, 28-29. Perez-Martin, R.I., Franco, J.M., Molist, P. and Gallardo, J.M. (1987). Gas chromatographic method for the determination of volatile amines in seafoods. Int. J. Food Sci. Technol., 22, 509-514. Prez-Villarreal, B. and Howgate, P. (1987). Spoilage of European hake, (Merluccius merluccius), in ice. J. Sci. Food Agric., 41, 335-350. Pivarnik, L.F., Thuam, M. and Ellis, P.C. (1998). Rapid determination of volatile bases in fish by using an ammonia ion-selective electrode. J. AOAC Int., 81, 1011-1022. Poller, K. and Linneweh, W. (1926 ). The occurrence of trimethylamine oxide in Clupea harengus (ber das Vorkommen von Trimethyamin-oxide in Clupea harengus). Ber. Deut. Chem. Ges., 59, 1362-1365. (In German). Reay, G.A. (1937). The nitrogenous extractives of fish. Report of the Food Investigation Board for the Years, 1935-1937, Section IV - Fish. Pp. 69-71. Department of Scientific and Industrial Research, London. Rehbein, H. and Oehlenschlger, J. (1982 ). Composition of the volatile basic N fraction 32

(TVB-N) in acid extracts and alkaline distillates from sea fish fillets. (Zur Zusammensetzung der TVB-N-Fraktion (fluechtige Basen) in sauren Extrakten und alkalischen Destillaten von Seefischfilet). Arch. Lebensmittelhyg., 33, 44-48. (In German) Rehbein, H. and Oehlenschlger, J. (1988 ). Determination of volatile basic nitrogen (TVB-N). Deamination of pure substances occurring in fish muscle during steam distillation after mild alkalization. (Desaminierung von im Fischmuskel vorkommenden Reinsubstanzen bei Wasserdampfdestillation nach milder Alkalisierung). Informationen fr die Fischwirtschaft, 35, 136-139. (In German). Richter, D., Lee, M.H. and Hill, D. (1941). The rate of removal of amines from the blood. Biochem. J., 35, 1225-1230. Ruiz-Capillas, C. and Horner, W.F.A. (1999). Determination of trimethylamine nitrogen and total volatile basic nitrogen in fresh fish by flow injection analysis. J. Sci. Food Agric., 79, 1982-1986. Sadok, S., Uglow, R.F. and Haswell, S.J. (1996). Determination of trimethylamine in fish by flow injection analysis. Anal. Chim. Acta, 321, 69-74. Sander, R. (2007). Compilation of Henry's law constants for inorganic and organic species of potential importance in environmental chemistry. http://www.mpch-mainz.mpg.de/~sander/res/henry.html. Accessed April, 2008. Saito, T. and Arai, K. (1958). Slow freezing of carp muscle and inosinic acid formation. Arch. Biochem. Biophys., 73, 315-319. Shewan, J.M., Gibson, D.M. and Murray, C.K. (1971). The estimation of trimethylamine in fish muscle. In: Fish Inspection and Quality Control. Pp 183-186. R. Kreuzer, ed. Fishing News (Books) Limited, London. Spinelli, J. (1964). Evaluation of the micro-diffusion method for the determination of tertiary volatile base in marine products. Fish. Ind. Res., 2, 17-19. Stansby, M.A., Harrison, R.W. and Dassow, J. Sater, M. (1944). Determining volatile bases in fish - comparison of precision of certain methods. Ind. Eng. Chem. Anal. Edn., 16, 593-596. Storey, R.M., Davis, H.K., Owen, D. and Moore, L. (1984 ). Rapid approximate estimation of volatile amines in fish. J. Food Technol., 19, 1-10. Surti, T., Taylor, T. and Ma'Ruf, F. (2001). The effect of storage at tropical ambient temperature on the quality and shelf life of grouper (Plectropomus maculatus). Int. J. Food Sci. Technol., 36, 517-522. Sze, Y.L., Borke, M.L. and Ottenstein, D.M. (1963). Separation of lower aliphatic amines by gas chromatography. Anal. Chem., 55, 240-242. Tarr, M.L.A. (1939). The bacterial reduction of trimethylamine oxide to trimethylamine. J. Fish. Res. Board Can., 4, 367-377.

33

Tillmans, J. and Otto, R. (1924). The detection of the onset of spoilage of fish. (ber den Nachweis der beginnenden Fischfulnis). Z. Unters. Nahr. Genussmitt., 47, 25-37. (In German). Tillmans, J. and Mildner, H. (1916). The detection of the onset of spoilage of flesh. (ber den Nachweiss beginnender Fleischfulnis), Z. Unters. Nahr. Genussmitt., 32, 65-75. Timm, M. and Jrgensen, B.M. (2002). Simultaneous determination of ammonia, dimethylamine, trimethylamine, and trimethylamine-N-oxide in fish extracts by capillary electrophoresis with indirect UV-detection. Food Chem., 76, 509-518. Tokunaga, T., Iida, H. and Miwa, K. (1977 ). Gas chromatographic analysis of amines in fish. Bull. Jap. Soc. Sci. Fish., 43, 219-227. Tomiyama, T., de Costa, A.A. and Stern, J.A. (1956). A rapid vacuum distillation procedure for the determination of volatile acids and volatile bases in fresh fish. Food Technol., 10, 614-617. Tomiyama, T. and Harada, Y. (1952). Studies on the method for testing the spoilage of food - IV. A rapid vacuum-distilling method for the determination of volatile base. Bull. Jap. Soc. Sci. Fish., 18, 112-116. Tozawa, H., Enokihara, K. and Amano, K. (1971). Proposed modification of Dyer's method for trimethylamine determination in cod fish. In: Fish Inspection and Quality Control. R. Kreuzer, ed. Pp 187-190. Fishing News (Books) Limited, London. Vyncke, W. (1970). Comparison of two methods for determining Volatile Basic Nitrogen (TVN). FAO Technical Conference on Fish Inspection and Quality Control, Halifax, Canada, 15-25 July 1969. FAO Fisheries Reports, No. 81, Vol 2. FE: FIC/69/0/60. Rome: Food and Agriculture Organization. Vyncke, W. (1996). Comparison of the official EC method for the determination of total volatile bases in fish with routine methods. Archiv fr Lebensmittelhygiene, 47, 110-112. Vyncke, W., Luten, J., Brnner, K. and Moermans, R. (1987). Determination of total volatile bases in fish: a collaborative study by the West European Fish Technologists' Association (WEFTA). Z. Lebens. Unters. Forsch., 184, 110-114 Watson, D.W. (1939 ). Studies on fish spoilage. IV. The bacterial reduction of trimethylamine oxide. J. Fish. Res. Board Can., 4, 252-266. Weber, F.C. and Wilson, J.B. (1919). The formation of ammonia and amines in canned sardines during storage. J. Ind. Eng. Chem., 11, 121-126 Wekell, M.M., Hollingworth, T.A. and Sullivan, J.J. (1987 ). Application of flow injection analysis (FIA) to the determination of seafood quality. In: Seafood Quality Determination. D.E. Kramer and J. Liston, eds. Pp. 17-26. Elsevier Science Publishers BV. Amsterdam. Winkles, G.H. (1855). On the existence of trimethylamine in the brine of salted herring. Quart. J. Chem. Soc. London, 7, 63-68.

34

Wong, K., Bartlett, F. and Gill, T.A. (1988). A diagnostic test strip for the semiquantitative determination of trimethylamine in fish. J. Food Sci., 53, 1653-1655. Wong, K. and Gill, A. (1987). Enzymatic determination of trimethylamine and its relationship to fish quality. J. Food Sci., 52, 1-3,6. Zhang, Z., Xu, K., Xing, Z. and Zhang, X. (2005). A nanosized Y2O3-based catalytic chemiluminescent sensor for trimethylamine. Talanta, 65, 913-917. Zhi, Z-L., Rios, A. and Valcrcel, M. (1995). Direct determination of trimethylamine in fish in the flow-reversal injection mode using a gas extraction sampling device. Anal. Chem., 67, 871-877.

35