The minimum information needed to calculate sample size for a randomized controlled trial in which a specific event is being

counted includes the power, the level of significance, the underlying event rate in the population under investigation and the size of the treatment effect sought. The calculated sample size should then be adjusted for other factors, including expected compliance rates (Kirby et al., 2002). Estimation of the sample size in a clinical trial should follow a few very important principles. The first and most important factor is specifying the primary outcome measure (Collins et al., 1984). The second most important factor is estimating the desired treatment effect which is worth registering in order to obtain a minimally clinically significant effect. It is usually estimated through pilot studies and/or literature search. To make predictions about recruitment rates, the following factors should be taken into account: Patient population characteristics (such as age/sex distribution, disease stage while diagnosed, geographical location, and social environment), intervention and outcomes in the study, trial setting, trial complexity and support, and available health care resources should be taken into account (McDonald et al., 2006). In this assignment we need to calculate sample size for a prospective clinical study: Hypothesis: Does continuous magnesium sulphate infusion reduce the need for mechanical ventilation and improve control of muscle spasms and autonomic instability? Subject population: Patients presenting with tetanus. Single study site: Specialized ICU unit at the Hospital for Tropical Diseases in Vietnam. Study design: Randomized, double blind, placebo controlled. Primary outcome: Requirements of assisted ventilation, odds ratio. Statistical analysis: Intention to Treat.

Sample size calculation Only primary outcome of requirement of assisted ventilation was be used for sample size calculation. Calculation was done using the approach described by Whitley E and Ball (2002) for the sample size required to compare proportions in two equally sized groups. Given proportions: in control arm p1=0.75 in treatment arm p2=0.55 the mean of two proportions p=( p1+p2)/2=0.65

SD = (p(1-p))= (0.650.35)=0.477 Standardized difference= (Target difference)/SD=(p1-p2)/SD=0.2/0.477=0.42 cp,power for P value 0.5 and 80% power 7.9 (Table 2, ) Number of patients required in each arm: n=[p1 (1-p1) + p2(1-p2)]/ ( p1-p2)2cp,power n=[0.750.25+0.550.45]/0.227.9=[0.1875+0.2475]/0.047.9=86 We will round up number of patients per group to 90. Due to the protocol design there should be no loss to follow up. This is a placebo controlled double blind study and the treatment may have to be discontinued for patients presenting with severe adverse effects (renal failure, ECG changes, cardiac arrests, hypotention, high serum magnesium, etc.). Primary outcome data could be missed if the treatment is interrupted prior to a tracheostomy patients being put on ventilator due to SAEs. Special instructions should be put in place in the trial protocol to account for such patients in data analysis. All patients should be entered into the study analysis. We probably should boost our enrollment by 10% as a precaution to be able to account for unanticipated circumstances. These assumptions bring the sample size to 200.

Estimation of recruitment rates

According to the baseline characteristics of the patients of interest presented in Thwaites et al., (2004) there are around 300 patients admitted every year (ranging from 249 to 320) if recruitment is not interrupted. During four years out of ten (1993-2002) when recruitment was interrupted, total number of cases was as low as 175 in 2000. Analysis of the age distribution in 2002 presented by Thwaites and colleagues, shows that only 23% of patients are children 18 years old and younger. Thus there are about 230 (3000.77) adult patients treated at the ICU yearly. There are very few female patients of child bearing age (only five patients 43 years old and younger in the adult category) and the possibility of pregnant patients affecting the sample size should be negligible. Some patients will not qualify for the study by the inclusion criteria. Thwaites et al (2006) included in a similar study only patients with a tracheostomy in situ. Thwaites et al. (2004) showed that at the ICU tracheostomy was done only in patients with a severe condition which comprised 40% all admitted patients. This brings down the number of eligible patients per year to around 90. We will assume conservatively that around 25% of assessed patients will be excluded from the study if they are removed from the ventilator within four hours or have contraindications to magnesium sulfate therapy based on their serum creatinine level, urine output, ECG, BP, fluid resuscitation requirements, and

refusal to sign an informed consent. This assumption will lower the number of eligible patients to 65-70 per year. With the assumption of 65-70 patients being eligible for the study, it will take 34 months to recruit 200 subjects and it may take longer if the study starts during the wet season (May-October) when a significantly smaller number of patients (21 versus 26) are admitted to the ICU (Thwaites et al., 2004). This estimate is in agreement with the time-line of actual clinical trial done by Thwaites and colleagues in 2002-2005 and published in 2006. If the requirements for a tracheostomy could be omitted, 170 patients could be eligible for the study yearly and the study could be completed in approximately in 14 months.

In the presented study the recruitment rate estimate is rather straightforward. The ICU trial site setting presumes that patients are assessed for recruitment as they are admitted for treatment. The financial cost of intervention (magnesium sulfate) is negligible. Randomization and blinding are the major areas of concern. Patients are available for treatment as soon as they are randomized. Loss to follow up is non-existent. Unscheduled interruption of the intervention is unavoidable due to the possibility of SAEs and should be planned accordingly.

References

Collins JF, Williford WO, Weiss DG et al. (1984) Planning patient recruitment: Fantasy and reality . Statistics in Medicine. 3(4): 435-443. Kirby A, Gebski V and Keech AC (2002) Determining the sample size in a clinical trial MJA 177 (5): 256257. McDonald AM, Knight RC, Campbell MK et al. (2006) What influences recruitment to randomised controlled trials? A review of trials funded by two UK funding agencies Trial 7:9 [Online] Available from: http://www.trialsjournal.com/content/7/1/9 (Accessed: 5 December 2010) Thwaites CL, Yen LM, Nga NTN et al. (2004) Impact of improved vaccination programme and intensive care facilities on incidence and outcome of tetanus in southern Vietnam, 1993-2002. Transactions of the Royal Society of Tropical Medicine and Hygiene; 98 (11): 671-677. Thwaites CL, Yen LM, Loan HT et al. (2006) Magnesium sulphate for treatment of severe tetanus: a randomised controlled trial. Lancet 368 (9545): 1436-1443. Whitley E and Ball J. Statistics review 4: Sample size calculations.Critical Care 2002, 6:33-341