918 THE NEW ENGLAND JOURNAL OF MEDICINE April 4, 1996

CORRESPONDENCE In this case, the actual diagnosis of pulmonary tuberculosis

CLINICAL PROBLEM-SOLVING: IF AT FIRST YOU
DON’T SUCCEED
To the Editor: In the case of the 22-year-old Laotian immi-
grant described in the Clinical Problem-Solving article enti-
tled “If at First You Don’t Succeed” (Dec. 7 issue),1 does the
one-year interval between the first medical consultation and
the complete relief of symptoms by curative antituberculosis
chemotherapy indicate that she received high-quality health
care? The record says that her physical examination was
“normal.” Can one assume that signs of thyrotoxicosis were
sought in the eyes and not found? That post-tussive rales were
absent? That a goiter was absent? What has happened to the
practice of comprehensive history taking and physical exam-
ination? Is thoroughness no longer considered the foundation
of high-quality medical care? Physicians have been told that
chest x-ray films should not be taken routinely because they
are not “cost effective.” Have physicians come to accept cost
effectiveness as the preeminent justification for a medical pro-
cedure?
Before practice guidelines and issues of cost effectiveness
came to dominate medical care, a symptom-driven initial pres-
entation of a patient to a physician would mandate a prompt
follow-up appointment for a comprehensive examination, which
would include the recording of both positive and pertinent
negative points of the history and physical examination, com-
plete blood count, erythrocyte sedimentation rate, urinalysis,
chest x-ray films, and skin testing with intermediate-strength
purified-protein-derivative tuberculin.
could and should have been made within the first week of
presentation. This would have reflected medical care of high
quality and true cost effectiveness. As it was presented, the
record indicates ineffective care of unacceptable quality.
ALAN F. CARPENTER, M.D.
Los Altos, CA 94024 1890 Granger Ave.
1. Ross JM, Sox HC. If at first you don’t succeed. N Engl J Med 1995;333:
1557-60.
To the Editor: In view of the fact that the patient in the Clin-
ical Problem-Solving case was a Laotian immigrant, a chest
film should have been obtained during the initial diagnostic
workup in the clinic. The prevalence of tuberculosis in the im-
migrant population was cogently described by McKenna and
colleagues.1 Also, the workup assessing the eosinophilia should
have included duodenal aspiration for strongyloides larvae or
a Beale string test.
ROBERT B. PRICE, M.D.
Waynesville, NC 28786 720 Camp Branch Rd.
1. McKenna MT, McCray E, Onorato I. The epidemiology of tuberculosis
among foreign-born persons in the United States, 1986 to 1993. N Engl J
Med 1995;332:1071-6.
To the Editor: In their Clinical Problem-Solving article, Drs.
Ross and Sox failed to discuss the misdiagnosis of iron defi-
ciency and the danger of prescribing iron in the absence of ob-
jective laboratory evidence of iron deficiency. The patient was
treated with iron because of a mild microcytic anemia without
any tests of iron metabolism ever being performed. Moreover,
the expert physician considered several possible reasons for
the failure of iron therapy but never once suggested that tests
for iron status — transferrin saturation and measurement of
transferrin and ferritin — are indicated before anyone is given
iron. The laboratory is more sensitive and specific than the
clinician in making a diagnosis of iron deficiency or iron over-

Instructions for Letters to the Editor

Letters to the Editor are considered for publication (subject to editing and abridgment) provided they do not contain
material that has been submitted or published elsewhere. Please note the following:
• Your letter must be typewritten and triple-spaced.
• Its text, not including references, must not exceed 400 words (please include a word count).
• It must have no more than five references and one figure or table.
• It should not be signed by more than three authors.
• Letters referring to a recent Journal article must be received within four weeks of its publication.
• Please include your full address, telephone number, and fax number (if you have one).
You may send us your letter by post, fax, or electronic mail.
Our address: Letters to the Editor
New England Journal of Medicine
10 Shattuck St.
Boston, MA 02115
Our fax numbers: 617-739-9864 and 617-734-4457
Our Internet address: letters@edit.nejm.org
We cannot acknowledge receipt of your letter, but we will notify you when we have made a decision about publication.
We are unable to provide prepublication proofs. Please enclose a stamped, self-addressed envelope if you want unpublished
material returned to you.
Financial associations or other possible conflicts of interest must be disclosed. Submission of a letter constitutes per-
mission for the Massachusetts Medical Society, its licensees, and its assignees to use it in the Journal ’s various editions
(print, data base, and optical disk) and in anthologies, revisions, and any other form or medium.

The New England Journal of Medicine

Downloaded from nejm.org on November 8, 2018. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
This document was created with FrameMaker 4.0.4P2
Vol. 334 No. 14
load. Iron therapy can injure patients with thalassemia, who
may already be overloaded with iron.
The authors also state that the alpha-thalassemia trait was
diagnosed by hemoglobin electrophoresis. Such a diagnosis is
not usually possible except among newborns in whom Bart’s
hemoglobin can be demonstrated. A defect in alpha-chain
production has an equal effect on the synthesis of hemoglobin
A and A 2 and fetal hemoglobin. Therefore, the alpha-thalas-
semia trait is not associated with electrophoretic abnormali-
ties. Theoretically, you might predict the presence of a faint
hemoglobin H band in persons with the alpha-thalassemia
trait, but this is not encountered in actuality. Therefore, the
questions to be answered are how the alpha-thalassemia trait
was diagnosed in this patient, and whether the diagnosis was
correct.
RAYMOND GAMBINO, M.D.
Teterboro, NJ 07680 Corning Clinical Laboratories
HENRY SOLOWAY, M.D.
Las Vegas, NV 89119 Associated Pathologists Laboratories
To the Editor: The Clinical Problem-Solving article “If at
First You Don’t Succeed” raised a number of important is-
sues. We would like to comment on one: the need for ade-
quate interpretation. In the case described, a non–English-
speaking Laotian woman underwent an extensive workup for
a series of elusive symptoms. From the outset, because they
did not share a common language, both the patient and the
doctor were at a great disadvantage. The patient could not
directly express her worries or describe her symptoms, and
the doctor could not exploit the opportunities afforded by an
examination of the patient’s history to generate prior proba-
bilities.
Interpreters bridge language barriers. Since finding a trained
interpreter may be difficult, patients and providers often rely
on strangers with unknown language skills or on family mem-
bers, including minor children, to interpret. The use of such
interpreters raises serious practical, ethical, and even legal
questions.1 In the case presented, the patient’s husband served
as interpreter. The use of untrained interpreters, particularly
family members, raises two important questions. First, was his
English good enough to provide accurate translation, partic-
ularly with regard to medical terminology? Perhaps ineffec-
tive communication was the real reason for her noncompli-
ance with H2-blocker therapy. Second, would the patient be
willing to share all pertinent information with her husband?
It is possible that the patient might not discuss issues such as
physical or emotional abuse or an extramarital relationship
(e.g., in the context of assessment of risk for the human im-
munodeficiency virus) in his presence. (Perhaps, more impor-
tant, she may have been unable to give informed consent for
iodine-131 treatment.)
Where could the clinician have found a trained interpreter?
Possibilities include local language banks (sometimes hospi-
tal- or university-based) and community organizations. Al-
though it can be awkward to use, the AT&T language line
may be a useful option for uncommonly spoken languages or
during off hours when interpreters may not be available.
Many clinicians lack experience in working with an interpret-
er. Fortunately, both written and videotaped instructions are
available to help clinicians work effectively with interpret-
ers.2,3
Effective communication is important to every aspect of
care, and it is fundamental to achieving the kind of long-term
The New England Journal of Medicine
Downloaded from nejm.org on November 8, 2018. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
CORRESPONDENCE 919
doctor–patient relationship that is rightly lauded at the con-
clusion of the case. Trained interpreters make such commu-
nication possible between people who speak different lan-
guages. As the number of persons with limited proficiency in
English grows in the United States, improving access to qual-
ified interpreters will become increasingly important to the
practice of good medicine.
STEVEN WOLOSHIN, M.D.
LISA M. SCHWARTZ, M.D.
Veterans Affairs
White River Junction, VT 05009 Medical Center
1. Woloshin S, Bickell NA, Schwartz LM, Gany F, Welch HG. Language barri-
ers in medicine in the United States. JAMA 1995;273:724-8.
2. Putsch RW III. Cross-cultural communication: the special case of interpret-
ers in health care. JAMA 1985;254:3344-8.
3. Boston Area Center for Health Education. Bilingual medical interview. Bos-
ton: Department of Health and Hospitals, 1987.
To the Editor: Ross and Sox state that “the sum of all diag-
nostic probabilities must be one” and that “all but the most
accurate tests are most useful when the prior probability is
between 30 and 70 percent.” In the case presented, at least
two of the three main diagnoses (infection with Opisthorchis
sinensis, Graves’ disease, and pulmonary tuberculosis) were
present simultaneously. This is an example that argues strong-
ly against the truth of the first axiom. A change in the prob-
ability of one disease should not necessarily cause a reciprocal
change in the probability of other diagnoses. (Does the fact
that the battery of your car proves to be dead diminish the
probability that you have also run out of gas?) Too often this
axiom, which is a didactic simplification for students who are
working on an exercise on one given disease, is misleading for
clinicians (not for the discussant, who did not take it into ac-
count).
A test is useful when it brings probability up or down to-
ward the test threshold or the test-treatment threshold.1 Tests
applied on the basis of a prior probability between 30 and 70
percent will in fact have a high numerical yield. Clinicians of-
ten order powerful tests in order to increase the probability of
disease from 99 percent to 99.99 percent (biopsy) or to lower
it from 0.1 percent to 0.001 percent (test for the human im-
munodeficiency virus) while ruling in or out life-threatening
diseases. Is the usefulness of the tests in these cases less than
when the probability is increased from 50 percent to 90 per-
cent? Was the yield of the finding of acid-fast bacilli in the
case under discussion not as important as the result of skin
testing with purified protein derivative? Its absolute yield was
certainly much less, since the prior probability of tuberculosis
was already high, given the chronic cough, the weight loss,
the fatigue, the infiltrate on the chest radiogram, and the pos-
itive skin test itself. For life-threatening diseases and for toxic
or hazardous treatments, the last steps in the evolution to-
ward certainty are considered by clinicians to be as important
as the steps in the middle of the probability range.
We are convinced that the use of a linear scale to represent
the probability of a disease is not appropriate for clinicians;
rather, clinicians think in terms of a logarithmic scale, on
which small numerical differences close to 0 percent and 100
percent can be as important as large numerical differences in
the medium range of probability.2
JEF VAN DEN ENDE, M.D.
ALFONS VAN GOMPEL, M.D.
B-2000 Antwerp, Belgium Institute of Tropical Medicine
920 THE NEW ENGLAND JOURNAL OF MEDICINE
1. Pauker SG, Kassirer JP. The threshold approach to clinical decision making.
N Engl J Med 1980;302:1109-17.
2. van den Ende J, van Gompel A, van den Enden E, van Damme W, Janssen
P. Bridging the gap between clinicians and clinical epidemiologists: Bayes
theorem on an ordinal scale. Theor Surg 1994;9:195. abstract.
To the Editor: I am the public health nurse coordinator who
monitored the patient in the Clinical Problem-Solving article
during her treatment for tuberculosis. The organism isolated
from the patient was not resistant to any antituberculosis med-
ication. Initially, a telephone report from the laboratory indi-
cated resistance to isoniazid and ethambutol. At that time,
these medications were dropped from the four-drug regimen
and streptomycin and ciprofloxacin were added. When the fi-
nal printed report came from the laboratory, it indicated sen-
sitivity of the organism to all medications. When the labora-
tory was asked to clarify these results, it noted that the initial
telephone report incorrectly indicated resistance but that the
finding was believed to be a result of antibiotic degradation.
All subsequent testing indicated 100 percent sensitivity to an-
tibiotics. Rather than changing the drug regimen again, we
continued to treat the patient with rifampin, pyrazinamide,
streptomycin, and ciprofloxacin.
VALERIE F. BENOIT, R.N., M.A.
Claremont, NH 03743-2280 Public Health District Office
The authors reply:
To the Editor: Drs. Van den Ende and Van Gompel are cor-
rect in chiding us for failing to say that the probabilities of
mutually exclusive events must total one. However, our case
does not illustrate their point. To understand our argument,
one must first know the definition of probability: a person’s
estimate of the likelihood of an event. In retrospect our pa-
tient had several diseases at once, but since the discussant did
not know this until the end of the case, the discussant’s prob-
abilities should sum to one. Perhaps, as Drs. Van den Ende
and Van Gompel claim, physicians do think in terms of a log-
arithmic scale of uncertainty. In the absence of any proof of
this statement, we stand by our claim that testing will have
the largest effect on uncertainty when the pretest probability
is intermediate.
The comments of the other letter writers relate to the de-
tails of clinical care. Drs. Woloshin and Schwartz eloquently
underscore the obligation of physicians caring for those who
speak a different language to obtain a qualified interpreter.
Dr. Price correctly raises the point that the possibility of in-
fection with strongyloides should have been aggressively pur-
sued, since the sensitivity of a single or even multiple stool
specimens is not high. Concentration of the stool increases
the yield significantly, and negative serologic results make in-
fection unlikely. Duodenal aspiration in pursuit of this diag-
nosis is not commonly performed at our institution but is an
effective tool in the right clinical setting.
Dr. Carpenter hypothesizes that a comprehensive initial
examination would have uncovered clues leading to an earlier
diagnosis and implies that the physicians involved in the pa-
tient’s care might have been distracted by guidelines or cost
considerations. A careful review of the patient’s record does
not support the first contention — the chart documents the
performance of careful comprehensive examinations on more
than one occasion. Few physicians believe that determination
of the erythrocyte sedimentation rate or chest radiography
would have been appropriate at the patient’s first visit. The
The New England Journal of Medicine
Downloaded from nejm.org on November 8, 2018. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
April 4, 1996
larger concern, that physicians might become less careful giv-
en the pressures of cost containment and managed-care in-
centives, is certainly appropriate, although not germane in
our largely fee-for-service setting. A chest x-ray film and tu-
berculin skin test should have been done far earlier in the pa-
tient’s clinical course.
We agree with Drs. Gambino and Soloway that the labora-
tory diagnosis of iron deficiency would have greatly assisted the
clinicians caring for the patient. They further question whether
hemoglobin electrophoresis could have diagnosed both the he-
moglobin E and alpha-thalassemia traits. In this case, electro-
phoretic quantification of hemoglobin E revealed a value of 27
percent, which is lower than is usually seen with the hemoglo-
bin E trait and more suggestive of a mixed hemoglobinopathy
with alpha-thalassemia trait.1 Finally, Ms. Benoit has reported
that the patient’s strain of Mycobacterium tuberculosis was not
drug-resistant. This information was known to her medical
team but absent from the records we reviewed.
JONATHAN M. ROSS, M.D.
HAROLD C. SOX, M.D.
Dartmouth–Hitchcock
Lebanon, NH 03756 Medical Center
1. Tuchinda S, Rucknagel DL, Minnich V, Boonyaprakob U, Balankura K, Su-
ratee V. The coexistence of the genes for hemoglobin E and a thalassemia in
Thais, with resultant suppression of hemoglobin E synthesis. Am J Hum Ge-
net 1964;16:311-35.
ANGIOGENESIS AND TUMOR GROWTH
To the Editor: Folkman’s article on angiogenesis (Dec. 28
issue)1 focused on clinical applications, with particular em-
phasis on the treatment of cancer. There is evidence that an-
giogenesis is indeed a factor in the progression of human car-
cinoma, but the extent of its role is in question.2 Although an
intratumoral lymphatic system is usually not well developed
in solid tumors, an elaborate system of lymphatic and prelym-
phatic channels in close proximity to the primary cancer and
invasive areas has been reported in breast cancer.3 It is there-
fore important to consider the possible impact of lymphagen-
esis (lymphatic-vessel formation) on the growth and metasta-
sis of solid tumors, particularly breast tumors.4
Lymphatic invasion has greater prognostic strength than
blood-vessel invasion for overall survival.5 Dissemination
through the lymphatic vessels to regional lymph nodes is one
of the most powerful prognostic indicators of metastatic dis-
ease. The lymphatic network is important in the progression
of breast carcinoma.
Variation in the dependence of tumor on the angiogenic
and lymphatic systems is likely. The use of antiangiogenic fac-
tors to inhibit the proliferation and migration of endothelial
cells in an attempt to control cancer has begun, but the de-
gree to which these agents affect lymphagenesis remains to
be determined. Specific antilymphagenic agents, if they exist,
may generate new protocols for the treatment of cancer.
Unless one is referring solely to the vascular system, we
suggest “vasogenesis” rather than “angiogenesis” for the proc-
ess of vessel formation in tumors. This distinction is also cru-
cial in distinguishing between antilymphagenic and antian-
giogenic agents.
JOHANNES P. VAN NETTEN, PH.D.
STEPHEN A. CANN, B.SC.
NICHOLAS G. VAN DER WESTHUIZEN, M.B.
Victoria, BC V8R 1J8, Canada Royal Jubilee Hospital
Vol. 334 No. 14 CORRESPONDENCE
1. Folkman J. Clinical applications of research on angiogenesis. N Engl J Med
1995;333:1757-63.
2. Costello P, McCann A, Carney DN, Dervan PA. Prognostic significance of
microvessel density in lymph node negative breast carcinoma. Hum Pathol
1995;26:1181-4.
3. Hartveit F. Attenuated cells in breast stroma: the missing lymphatic system
of the breast. Histopathology 1990;16:533-43.
4. Cann SA, van Netten JP, Ashby TL, Ashwood-Smith MJ, van der Westhuizen
NG. Role of lymphagenesis in neovascularisation. Lancet 1995;346:903.
5. Lauria R, Perrone F, Carlomagno C, et al. The prognostic value of lymphatic
and blood vessel invasion in operable breast cancer. Cancer 1995;76:1772-8.
To the Editor: In his excellent overview of angiogenesis,
Folkman described the role of fibroblast growth factor in neo-
plastic and non-neoplastic diseases. There have been interest-
ing reports on the role of somatostatin in the inhibition of
angiogenesis.1,2 This effect is probably mediated by the soma-
tostatin receptor subtype 2, which has been shown to inhibit
the effect of the INT2 gene, which encodes fibroblast growth
factor, in several tissues. Somatostatin and its analogues have
a potential role in treating some malignant conditions, either
through a direct antitumor effect or through potentiation of
the effect of other chemotherapies.3 This effect could be ex-
plained by the inhibition of angiogenesis in tumors expressing
the somatostatin receptor subtype 2. Tumors in different stag-
es of differentiation vary in expression of the receptor, which
could explain the conflicting results of studies to date.
New techniques for the selective detection of receptor sub-
types call for new studies of the role of somatostatin and
receptor-specific analogues in oncology. The effect of these
agents in non-neoplastic conditions is also under investigation,
and there are promising results in the treatment of ocular
neovascularization.4 We are currently evaluating the possible
therapeutic effect of the somatostatin analogue octreotide in
rheumatologic diseases.
HAIM PARAN, M.D.
Kfar-Sava, Israel Meir Hospital
DAPHNA PARAN, M.D.
Tel Aviv, Israel Tel Aviv Medical Center
1. Woltering EA, Barrie R, O’Dorisio TM, et al. Somatostatin analogues inhibit
angiogenesis in the chick chorioallantoic membrane. J Surg Res 1991;50:
245-51.
2. Patel PC, Barrie R, Hill N, Landeck S, Kurozawa D, Woltering EA. Postre-
ceptor signal transduction mechanisms involved in octreotide-induced inhi-
bition of angiogenesis. Surgery 1994;116:1148-52.
3. Weckbecker G, Tolcsvai L, Stolz B, Pollak M, Bruns C. Somatostatin ana-
logue octreotide enhances the antineoplastic effects of tamoxifen and ovar-
iectomy on 7,12-dimethylbenz(a)anthracene-induced rat mammary carcino-
mas. Cancer Res 1994;54:6334-7.
4. Grant MB, Caballero S, Millard WJ. Inhibition of IGF-I and b-FGF stimulat-
ed growth of human retinal endothelial cells by the somatostatin analogue,
octreotide: a potential treatment for ocular nonvascularization. Regul Pept
1993;48:267-78.
To the Editor: Folkman addressed the “switch” of tumor cells
to the angiogenic phenotype, but it is important to point out
that the p53 gene has an important role in controlling tumor
angiogenesis. Dameron et al. have reported that the loss of
the wild-type p53 allele in cultured fibroblasts from patients
with the Li–Fraumeni syndrome results in a decrease in levels
of the endogenous negative regulator of neovascularization,
thrombospondin-1.1 Other authors showed that a mutant p53
gene determines neovascularization in some tumor cell lines
by increasing the production of protein kinase C and enhanc-
ing the expression of the vascular endothelial growth factor.2
A clinical implication of these findings is that tumor-microves-
The New England Journal of Medicine
Downloaded from nejm.org on November 8, 2018. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
921
sel density and evaluation of p53 can have synergistic prog-
nostic value in node-negative breast carcinoma.3
FABIO PUGLISI, M.D.
SIMONA SCALONE, M.D.
VINCENZO DILAURO, M.D.
33100 Udine, Italy University of Udine
1. Dameron KM, Volpert OV, Tainsky MA, Bouck N. Control of angiogenesis
in fibroblasts by p53 regulation of thrombospondin-1. Science 1994;265:
1582-4.
2. Kieser A, Weich HA, Brandner G, Marme D, Kolch W. Mutant p53 potenti-
ates protein kinase C induction of vascular endothelial growth factor expres-
sion. Oncogene 1994;9:963-9.
3. Gasparini G, Weidner N, Bevilacqua P, et al. Tumor microvessel density, p53
expression, tumor size, and peritumoral lymphatic vessel invasion are rele-
vant prognostic markers in node-negative breast carcinoma. J Clin Oncol
1994;12:454-66.
Dr. Folkman replies:
To the Editor: Van Netten et al. propose the term “vasogen-
esis” for both angiogenesis and “lymphagenesis.” The pre-
ponderance of the experimental evidence is that tumors in-
duce new capillary blood vessels, but there is little evidence
that tumors induce new lymphatic vessels. Certain findings
argue against tumor-induced lymphangiogenesis. First, FLT4,
a marker of lymphatic endothelium, was detected in melano-
mas in the granulation tissue and dermis of the skin around
the tumor, but not in the tumor.1 Second, in a study of a vari-
ety of tumors, no functional lymphatic vessels were found.2
Moreover, no lymphangiogenic factors have been isolated
from tumors; microlymphangiographic and histologic studies
of mice bearing a sarcoma in the tail reveal functional lym-
phatic vessels only outside the tumor, not within it; and tu-
mors implanted in the mouse or rabbit cornea induce blood-
vessel growth, but not lymphatic-vessel growth. Unless it can
be demonstrated that tumors stimulate the growth of new
lymphatic vessels, “vasogenesis” does not seem to be an im-
provement over the currently accepted terms.
In the study cited by van Netten et al., Costello et al.3 do
not question whether tumor growth is dependent on angio-
genesis, but whether tumor-microvessel density is a valid prog-
nostic marker. To date, the vast majority of reports show a
strong correlation between tumor-microvessel density and the
risk of metastasis.4
Puglisi et al. emphasize the role of the p53 suppressor gene
in the switch to the angiogenic phenotype. I agree. My review
cited the work of Dameron et al.5
Paran and Paran offer a possible mechanism for the anti-
angiogenic activity of somatostatin and its analogues. Their
suggestion that these molecules may be useful in the therapy
of ocular neovascularization and rheumatologic disease is a
good one and merits further study.
JUDAH FOLKMAN, M.D.
Boston, MA 02115 Children’s Hospital
1. Kaipainen A, Korhonen J, Mustonen T, et al. Expression of fms-like tyrosine
kinase 4 gene becomes restricted to lymphatic endothelium during develop-
ment. Proc Natl Acad Sci U S A 1995;92:3566-70.
2. Dumont RE. Factor VIII-related antigen. J Natl Cancer Inst 1993;85:674-6.
3. Costello P, McCann A, Carney DN, Dervan PA. Prognostic significance of
microvessel density in lymph node negative breast carcinoma. Hum Pathol
1995;26:1181-4.
4. Weidner N. Intratumoral microvessel density as a prognostic factor in cancer.
Am J Pathol 1995;147:9-19.
5. Dameron KM, Volpert OV, Tainsky MA, Bouck N. Control of angiogenesis
in fibroblasts by p53 regulation of thrombospondin-1. Science 1994;265:
1582-4.
922 THE NEW ENGLAND JOURNAL OF MEDICINE April 4, 1996

TOXIC EPIDERMAL NECROLYSIS

To the Editor: Roujeau et al. (Dec. 14 issue)1 present data
showing a morbidity of 1 in 230,000 associated with toxic epi-
dermic necrolysis and the Stevens–Johnson syndrome due to
one week of trimethoprim–sulfamethoxazole therapy. If the
death rate is assumed to be 15 percent,2 the mortality rate
would be 1 in 1,550,000. It would be interesting to know the
actual mortality among their index patients.
To put this excess mortality in context, it is similar to
the transfusion-associated frequency of human immunodefi-
ciency virus transmission in Germany (1 in 800,000 to 1 in
2,000,000)3,4 and the United States (1 in 450,000 to 1 in
660,000).5 In many countries, including the United States and
Germany, as a result of recent court rulings, patients are to be
informed about this adverse effect and the therapeutic alter- A
natives, even if blood transfusion is only a remote possibility.
A consistent approach would imply that before prescribing
trimethoprim–sulfamethoxazole one should counsel patients
about the associated risk of death and obtain informed con-
sent. However, it is difficult for patients to understand the
small risk figures. Being told about a singular, exactly quan-
tified adverse effect may impede patients’ perception of other,
often more important adverse effects. Compliance will proba-
bly be reduced. We question whether explicit counseling about
a very low mortality risk benefits patients and would like to
know the authors’ views on this.
The authors conclude with the statement that “prescribing
physicians should still consider that alternative therapies have
substantially lower excess risks.” What are the rational op-
tions for the practitioner? What is the impact of the excess
mortality due to toxic epidermic necrolysis on the overall mor- B
tality associated with trimethoprim–sulfamethoxazole? Can
we conclude with sufficient confidence that alternative thera- Figure 1. Peeled Skin from a Patient with Toxic Epidermal Ne-
pies, such as amoxicillin or fluoroquinolones for urinary tract crolysis (Panel A) and a Patient with Staphylococcal Scalded
infections, are associated with lower overall mortality rates? Skin Syndrome (Panel B).
Since trimethoprim–sulfamethoxazole is inexpensive, how The level of the skin split in toxic epidermal necrolysis is below
much in additional costs for alternative therapies should be the epidermis, revealing a full-thickness necrotic epidermis (he-
matoxylin and eosin, 200). The level of split in staphylococcal
deemed acceptable? scalded skin syndrome is the epidermal granular-cell layer (he-
FRANZ F. WAGNER, M.D. matoxylin and eosin,  400).
WILLY A. FLEGEL, M.D.
D-89081 Ulm, Germany University of Ulm
skin syndrome is the epidermal granular-cell layer, whereas in
1. Roujeau J-C, Kelly JP, Naldi L, et al. Medication use and the risk of Stevens– toxic epidermal necrolysis the level is subepidermal, with a
Johnson syndrome or toxic epidermal necrolysis. N Engl J Med 1995;333:
1600-7.
full-thickness necrotic epidermis (Fig. 1). It is interesting to
2. Roujeau JC, Stern RS. Severe adverse cutaneous reactions to drugs. N Engl note that in his original 1956 description of toxic epidermal
J Med 1994;331:1272-85. necrolysis, Lyell2 included a case of adult staphylococcal
3. Glück D, Elbert G, Dengler T, et al. HIV lookback of the Red Cross blood scalded skin syndrome. This led to two decades of confusion
banks in Germany. Infusionsther Transfusionmed 1994;21:368-75.
4. Zeiler T, Kretschmer V, Sibrowski W, et al. A retrospective investigation con-
because the terminology for staphylococcal scalded skin syn-
cerning the practice of HIV lookback methods, the incidence of HIV-1/2- drome and toxic epidermal necrolysis overlapped, with terms
positive blood donors and the risk of transfusion-associated HIV infection such as Lyell’s syndrome, staphylococcal-induced toxic epi-
from blood donations of governmental and communal blood transfusion dermal necrolysis, and drug-induced scalded skin syndrome.
services in Germany. Infusionsther Transfusionmed 1994;21:362-7.
5. Lackritz EM, Satten GA, Aberle-Grasse J, et al. Estimated risk of transmis-
These last three terms should no longer be used.
sion of the human immunodeficiency virus by screened blood in the United HENRY M. FEDER, JR., M.D.
States. N Engl J Med 1995;333:1721-5.
DIANE M. HOSS, M.D.
University of Connecticut
To the Editor: Because in children toxic epidermal necrolysis Farmington, CT 06030 Health Center
can be easily confused with staphylococcal scalded skin syn- ROBERT L. DIMOND, M.D.
drome, Roujeau and colleagues included children in their Dartmouth–Hitchcock
study “only if they had mucous-membrane erosions or target- Lebanon, NH 03756 Medical Center
like lesions, or had had a skin biopsy.” For the rapid differen-
1. Amon RB, Dimond RL. Toxic epidermal necrolysis: rapid differentiation be-
tiation of staphylococcal scalded skin syndrome from toxic tween staphylococcal- and drug-induced disease. JAMA 1975;111:1433-7.
epidermal necrolysis, a frozen section of peeled skin1 can be 2. Lyell A. Toxic epidermal necrolysis: an eruption resembling scalding of the
analyzed. The level of the skin split in staphylococcal scalded skin. Br J Dermatol 1956;68:355-61.

The New England Journal of Medicine

Downloaded from nejm.org on November 8, 2018. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
Vol. 334 No. 14
The authors reply:
To the Editor: Drs. Wagner and Flegel comment on the ex-
cess mortality due to Stevens–Johnson syndrome or toxic epi-
dermal necrolysis in relation to the administration of sulfona-
mides and question whether other antiinfective drugs are safer.
Our study was designed to quantify the risks of occurrence
of these severe skin reactions in relation to medication use.
The results demonstrated that the risk of Stevens–Johnson
syndrome or toxic epidermal necrolysis was significantly high-
er with antibacterial sulfonamides than with aminopenicillins
or quinolones. Among our 245 patients, the actual death rate
was 16 percent (38 deaths among 241 patients with informa-
tion on vital status). Fortunately, most patients survived the
reactions, and our study therefore lacked the statistical pow-
er to detect a significant difference in the number of deaths
among users of various drugs. A higher risk of developing
these reactions clearly indicates not only a higher morbidity
but also an increased risk of mortality. We agree that it is dif-
ficult to appreciate the clinical relevance of very rare adverse
events and to counsel patients on such risks. It is of course
necessary to consider an overall estimation of the risk–benefit
ratio. In that respect our study provided useful data for quan-
tifying one type of side effect.
Dr. Feder and coworkers advocate the use of histologic ex-
amination of a frozen section of peeled skin for the rapid dif-
ferentiation of staphylococcal scalded skin syndrome from
toxic epidermal necrolysis. That technique, used in some of
the patients included in our study, has been used in Créteil
for years and has proved helpful. However, we now rely on ex-
amination of the frozen section of a full-thickness skin biopsy,
because it is as easy to do and is less prone to misinterpreta-
tion of the level of separation.
JEAN-CLAUDE ROUJEAU, M.D.
94010 Créteil, France Hôpital Henri Mondor
DAVID W. KAUFMAN, SC.D.
Boston, MA 02118 Boston University School of Medicine
AUTOIMMUNE HEPATITIS
To the Editor: Meyer zum Büschenfelde and Lohse (Oct. 12
issue)1 suggest that patients with mild autoimmune hepatitis
should be treated. However, few data support the contention
that such patients will benefit. Although the results of large
controlled trials support the use of corticosteroids in the
treatment of autoimmune hepatitis,2-4 these beneficial effects
are fully established only in patients with the most severe his-
tologic manifestations.5 The effects of corticosteroids in re-
ducing mortality appeared to be confined to patients whose
initial live biopsies revealed bridging necrosis, multilobular
necrosis, or cirrhosis. There were no deaths among patients
whose initial liver biopsies revealed only chronic active hepa-
titis. Furthermore, corticosteroid therapy does not unequivo-
cally lower the risk of cirrhosis in patients who have only
chronic active hepatitis.
Koretz et al. have shown that only a minority of patients
with chronic active hepatitis appear to derive proved benefit
from corticosteroid treatment.6 What, then, should be recom-
mended for patients with mild autoimmune hepatitis? Should
corticosteroids be used in an asymptomatic patient in whom
the disease is manifested histologically only as periportal or
piecemeal necrosis? I believe the answer is a qualified no and
that the decision to treat must be made on an individual basis
and be guided by the available data. Pending additional infor-
The New England Journal of Medicine
Downloaded from nejm.org on November 8, 2018. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
CORRESPONDENCE 923
mation, it may be most prudent to withhold corticosteroid or
immunosuppressive therapy in patients with asymptomatic,
mild, autoimmune hepatitis.
JOSEPH C. YARZE, M.D.
Gastroenterology Associates
Glens Falls, NY 12801 of Northern New York
1. Meyer zum Büschenfelde K-H, Lohse AW. Autoimmune hepatitis. N Engl J
Med 1995;333:1004-5.
2. Cook GC, Mulligan R, Sherlock S. Controlled prospective trial of cortico-
steroid therapy in active chronic hepatitis. Q J Med 1971;40:159-85.
3. Soloway RD, Summerskill WH, Baggenstoss AH, et al. Clinical, biochemical,
and histological remission of severe chronic active liver disease: a controlled
study of treatments and early prognosis. Gastroenterology 1972;63:820-33.
4. Murray-Lyon IM, Stern RB, Williams R. Controlled trial of prednisone and
azathioprine in active chronic hepatitis. Lancet 1973;1:735-7.
5. Wright EC, Seeff LB, Berk PD, Jones A, Plotz PH. Treatment of chronic ac-
tive hepatitis: an analysis of three controlled trials. Gastroenterology 1977;
73:1422-30.
6. Koretz RL, Lewin KJ, Higgins J, Fagen ND, Gitnick GL. Chronic active hep-
atitis: who meets treatment criteria? Dig Dis Sci 1980;25:695-9.
The authors reply:
To the Editor: Yarze correctly points out that the role of im-
munosuppressive therapy in mild cases of autoimmune hepati-
tis has not yet been defined by controlled trials. Until such
studies are conducted, clinical reasoning and experience should
guide therapy. We believe that patients with mild autoimmune
hepatitis should receive immunosuppressive therapy, albeit at
lower doses than are recommended in more aggressive disease.
A large proportion of patients with autoimmune hepatitis al-
ready have cirrhosis when they come to medical attention. In
most of them the cirrhosis has developed in the absence of
clinical symptoms. In many, mild elevations of aminotransfer-
ase levels have been documented without a definite diagnosis
being made or therapy commenced. Furthermore, the majori-
ty of patients with cryptogenic cirrhosis have features of auto-
immune hepatitis, suggesting that subclinical hepatitis has led
to their cirrhosis.1 Most important, in many patients untreated
disease has a spontaneously fluctuating course. If liver biopsies
are performed in these patients during a phase of spontaneous
remission, little inflammation will be observed. However, when
the patients are left untreated, reactivation is likely to occur,
usually within one year, and it may go undetected, because the
only symptom, if any are present, may be fatigue.2 The only
justifiable alternative, we believe, would be very close follow-
up. However, regular biopsies would need to be included in
such an approach, because fibrosis may develop rapidly.
The rate at which cirrhosis develops in untreated patients
with autoimmune hepatitis is not well documented.3 The ther-
apeutic trials cited by Yarze have not included sufficiently
long observation periods or sufficient numbers of patients with
milder disease to show a survival benefit in this subgroup. An-
other study whose follow-up period lasted at least 10 years (or
until death, in the case of 44 patients) found a marked survival
benefit that was independent of the patients’ aminotransferase
levels at the start of immunosuppressive therapy.4 That report
concluded, “This study confirms the well-established fact that
steroid therapy is mandatory, and suggests that the required
duration of treatment is about five years.” In addition, since
all these studies were undertaken before the hepatitis C virus
was discovered, patients with viral hepatitis who do not benefit
from immunosuppressive therapy were included.
Autoimmune hepatitis often occurs in young patients. It is
a chronic disease that has been shown to be progressive in
many, if not most, cases. In the more aggressive form of the
924 THE NEW ENGLAND JOURNAL OF MEDICINE April 4, 1996

disease, the survival benefit of therapy is dramatic. Therefore, 2800

it makes good sense to treat milder disease as well.
4200
K.-H. MEYER ZUM BÜSCHENFELDE, M.D., PH.D. 2700

No. of Accidents
ANSGAR W. LOHSE, M.D.
D-55101 Mainz, Germany Johannes Gutenberg University
4000
1. Czaja AJ, Carpenter HA, Santrach PJ, Moore SB, Homburger HA. The na- 2600
ture and prognosis of severe cryptogenic chronic active hepatitis. Gastroen-
terology 1993;104:1755-61.
2. Czaja AJ, Ludwig J, Baggenstoss AH, Wolf A. Corticosteroid-treated chronic 3800
active hepatitis in remission: uncertain prognosis of chronic persistent hepa- 2500
titis. N Engl J Med 1981;304:5-9.
3. Thaler H. The natural history of chronic hepatitis. In: Schaffner F, Sherlock
S, Leevy CM, eds. The liver and its diseases. New York: Stratton Interconti-
nental Medical, 1974:207-15. 2400 3600

re

e
ly

ly
r

r
te

te
4. Kirk AP, Jain S, Pocock S, Thomas HC, Sherlock S. Late results of the Royal

Im efor
r te

r te
fo

af

af
te ia

te ia
Be
Free Hospital prospective controlled trial of prednisolone therapy in hepatitis

B
af ed

k
af ed
w

w
B surface antigen negative chronic active hepatitis. Gut 1980;21:78-83.

m
1

DAYLIGHT SAVINGS TIME AND TRAFFIC
ACCIDENTS
To the Editor: It has become increasingly clear that insuffi-
cient sleep and disrupted circadian rhythms are a major pub-
lic health problem. For instance, in 1988 the cost of sleep-
related accidents exceeded $56 billion and included 24,318
deaths and 2,474,430 disabling injuries.1 Major disasters, in-
cluding the nuclear accident at Chernobyl, the Exxon Valdez oil
spill, and the destruction of the space shuttle Challenger, have
been linked to insufficient sleep, disrupted circadian rhythms,
or both on the part of involved supervisors and staff.2,3 It has
been suggested that as a society we are chronically sleep-
deprived4 and that small additional losses of sleep may have
consequences for public and individual safety.2
We can use noninvasive techniques to examine the effects
of minor disruptions of circadian rhythms on normal activi-
ties if we take advantage of annual shifts in time keeping.
More than 25 countries shift to daylight savings time each
spring and return to standard time in the fall. The spring shift
results in the loss of one hour of sleep time (the equivalent in
terms of jet lag of traveling one time zone to the east), where-
as the fall shift permits an additional hour of sleep (the equiv-
alent of traveling one time zone to the west). Although one
hour’s change may seem like a minor disruption in the cycle
of sleep and wakefulness, measurable changes in sleep pattern
persist for up to five days after each time shift.5 This leads to
the prediction that the spring shift, involving a loss of an
hour’s sleep, might lead to an increased number of “micro-
sleeps,” or lapses of attention, during daily activities and thus
might cause an increase in the probability of accidents, espe-
cially in traffic. The additional hour of sleep gained in the fall
might then lead conversely to a reduction in accident rates.
We used data from a tabulation of all traffic accidents in
Canada as they were reported to the Canadian Ministry of
Transport for the years 1991 and 1992 by all 10 provinces. A
total of 1,398,784 accidents were coded according to the date
of occurrence. Data for analysis were restricted to the Mon-
day preceding the week of the change due to daylight savings
time, the Monday immediately after, and the Monday one
week after the change, for both spring and fall time shifts.
Data from the province of Saskatchewan were excluded be-
cause it does not observe daylight savings time. The analysis
of the spring shift included 9593 accidents and that of the fall
shift 12,010. The resulting data are shown in Figure 1.
The loss of one hour’s sleep associated with the spring shift
1
Im
Spring Shift Fall Shift
Figure 1. Numbers of Traffic Accidents on the Mondays before
and after the Shifts to and from Daylight Savings Time for the
Years 1991 and 1992.
There is an increase in accidents after the spring shift (when an
hour of sleep is lost) and a decrease in the fall (when an hour
of sleep is gained).
to daylight savings time increased the risk of accidents. The
Monday immediately after the shift showed a relative risk of
1.086 (95 percent confidence interval, 1.029 to 1.145; x 2 9.01,
1 df; P0.01). As compared with the accident rate a week later,
the relative risk for the Monday immediately after the shift was
1.070 (95 percent confidence interval, 1.015 to 1.129; x 2 6.19,
1 df; P0.05). Conversely, there was a reduction in the risk of
traffic accidents after the fall shift from daylight savings time
when an hour of sleep was gained. In the fall, the relative risk
on the Monday of the change was 0.937 (95 percent confidence
interval, 0.897 to 0.980; x 2 8.07, 1 df; P0.01) when com-
pared with the preceding Monday and 0.896 (95 percent con-
fidence interval, 0.858 to 0.937; x 2 23.69; P0.001) when
compared with the Monday one week later. Thus, the spring
shift to daylight savings time, and the concomitant loss of one
hour of sleep, resulted in an average increase in traffic acci-
dents of approximately 8 percent, whereas the fall shift result-
ed in a decrease in accidents of approximately the same mag-
nitude immediately after the time shift.
These data show that small changes in the amount of sleep
that people get can have major consequences in everyday ac-
tivities. The loss of merely one hour of sleep can increase the
risk of traffic accidents. It is likely that the effects are due to
sleep loss rather than a nonspecific disruption in circadian
rhythm, since gaining an additional hour of sleep at the fall
time shift seems to decrease the risk of accidents.
Vancouver, BC V6T 1Z4, STANLEY COREN, PH.D.
Canada University of British Columbia
1. Leger D. The cost of sleep-related accidents: a report for the National Com-
mission on Sleep Disorders Research. Sleep 1994;17:84-93.
2. Coren S. Sleep thieves. New York: Free Press, 1996.
3. Mitler MM, Carskadon MA, Czeisler CA, Dement WC, Dinges DF, Graeber
RC. Catastrophes, sleep, and public policy: consensus report. Sleep 1988;11:
100-9.
4. Webb WB, Agnew HW Jr. Are we chronically sleep deprived? Bull Psycho-
nom Soc 1975;6:47-8.
5. Monk TH, Folkard S. Adjusting to the changes to and from daylight saving
time. Science 1976;261:688-9.

1996, Massachusetts Medical Society.

The New England Journal of Medicine

Downloaded from nejm.org on November 8, 2018. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.