The Editors welcome submissions for possible publication in the Ofer Shpilberg, MD
Letters section. Authors of letters should: Avner Shahar, MD
• Include no more than 300 words of text, three authors, and five University of Pittsburgh Graduate School of Public Health
references Pittsburgh, PA 15261
• Type with double-spacing
• Send three copies of the letter, a transfer-of-copyright form (see References
Table of Contents for location) signed by all authors, and a 1. Polycythemia vera: the natural history of 1213 patients followed for 20
covering letter describing any conflicts of interest related to the years. Gruppo Italiano Studio Policitemia. Ann Intern Med. 1995; 123:
contents of the letter 656-64.
Letters commenting on an Annals article will be considered if they 2. Cohen-Goldenberg N, Shpilberg O, Modan B, Chetrit A, Ramot B.
are received within 6 weeks of the time the article was published. Long term follow up in polycythemia vera. Harefuah. 1995;128:673-6.
Only some of the letters received can be published. Published letters
are edited and may be shortened; tables and figures are included In response: Shpilberg and Shahar correctly emphasize that the
only selectively. Authors will be notified that the letter has been main limitation of our study is inherent in its retrospective de-
received. If the letter is selected for publication, the author will be sign. Thus, any finding should be interpreted cautiously, and any
notified about 3 weeks before the publication date. Unpublished conclusion considered as a working hypothesis should be verified
letters cannot be returned. prospectively.
Annals welcomes electronically submitted letters. The Compu- Several hypotheses and open questions exist about the diagno-
Serve address is mhs:annals@acp, and the Internet address is sis of polycythemia vera, given the sparse or outdated data avail-
annals @acp. mhs. CompuServe, com. able (1). Prospective studies of polycythemia vera whose results
have been published are limited by small sample size and by
discrepancies in inclusion criteria. Apart from the studies of the
Polycythemia Vera Study Group (2) and the trial organized by
the European Organization for Research on Treatment of Can-
Natural History of Polycythemia Vera cer (3), no other relatively large randomized, controlled trials can
be found in the literature. Our retrospective study, although by
far the largest in the literature, could not compensate for this
To the Editor: The highly informative article from the Gruppo lack of knowledge.
Italiano Studio Policitemia (1) has increased our knowledge of As clearly stated in our paper, our primary goal was to provide
the course of polycythemia vera. Although we endorse most of information for designing clinical trials. Thus, the incidence of
the author's conclusions, we believe the study has several limita- thrombotic events in polycythemic patients estimated in the ret-
tions inherent in its retrospective design. rospective analysis was used to calculate the sample size for a
Although the diagnostic criteria established by the group are double-blind, placebo-controlled clinical trial testing low-dose as-
useful in the research setting, they tend to be restrictive and may pirin in patients with polycythemia vera. Our other main objec-
exclude as many as one third of patients who have polycythemia tive was to explore the feasibility of organizing an active Italian
vera but incomplete presentations (2). In interpreting the Italian collaborative group for the study of polycythemia vera. Following
group's findings, it seems plausible that some or most of the this experience, a European Collaboration on Low-dose Aspirin
patients who previously had thromboses were actually presenting in Polycythemia Vera was established and funded by the Euro-
at that time with an incomplete form of polycythemia vera. Given pean Union BIOMED 2 program. This study is a randomized
the insidious course of the disease, such patients may differ in trial designed to assess the risk-benefit profile of low-dose aspi-
terms of neoplastic complications from patients who meet the rin in patients with polycythemia vera. Several European coun-
above criteria at the first manifestation of the disease. Thus, they tries are involved in the study, which will enroll approximately
should be analyzed separately. Moreover, the approach to treat- 3000 patients to be followed for 3 to 4 years. Patient recruitment
ing these milder cases may have been less aggressive in terms of began in June 1996 and is expected to be completed by Decem-
myelosuppressive agents. Therefore, the outcome in patients with ber 1998. Patients not randomly assigned to treatment will also
mild disease may have differed considerably with respect to neo- be followed. We hope that this prospective study will be helpful
plastic complications. in assessing the current natural history of the disease.
The authors conclude that survival duration in patients with
polycythemia vera is longer than previously believed, a result Roberto Marchioli, MD
similar to those from several other retrospective studies. Retro- Istituto Mario Negri—Consorzio Mario Negri Sud
spective assessment of survival poses several difficulties. For ex- 66030 Santa Maria Imbaro, Italy
ample, sudden death (such as that from massive myocardial
infarction) occurring before diagnosis is established would ex-
Tiziano Barbui, MD
clude a patient from the cohort. Therefore, such studies tend to
Ospedali Riuniti Bergamo
overestimate survival. Another variable affecting survival is that
patients who receive a diagnosis earlier in the course of their
disease tend to have a longer arithmetic survival. For example, in
another study (2) in which 16% of patients with polycythemia Raffaele Landolfi, MD
vera were identified incidentally while asymptomatic, the overall Universita Cattolica del Sacro Cuor
median survival was 17.4 years. 00168 Rome, Italy
Our final criticism relates to the conclusion that myelosuppres- References
sive agents have an overall unfavorable effect. In their Discussion
1. Marchioli R, Landolfi R, Barbui T, Tognoni G. Feasibility of random-
section, the authors contend that patients receiving myelosup- ised clinical trials in rare diseases: the case of polycythemia vera.
pression therapy may have more severe disease, as supported by Leukemia and Lymphoma. [In press].
the higher frequency of thrombosis. Such potential selection bias 2. Berk PD, Goldberg JD, Donovan PB, Fruchtman SM, Berlin NI, Was-
precludes a separate assessment of the adverse effects of myelo- serman LR. Therapeutic recommendations in polycythemia vera based
suppressive agents. on Polycythemia Vera Study Group protocols. Semin Hematol. 1986;
514 15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6
23:132-43. Table 1. Usefulness of Individual Historical Tests in
3. Treatment of polycythemia vera by radiophosphorus or busulphan: a Predicting the Presence ior Absence of
randomized trial. Leukemia and Hematosarcoma Cooperative Group, Clostridium difficile Colitis*
European Organization for Research on Treatment of Cancer
(E.O.R.T.C). Br J Cancer. 1981;44:75-80.
Index Test True- False- Positive Negative
Positive Positive Likelihood Likelihood
Rate Rate Ratio Ratio
Diagnosis of Clostridium difficile Colitis
Cephalospori n use 0.72 0.53 1.36 0.60
Length of hospital stay >5 days 1.00 0.92 1.09 0.00
To the Editor: In the article by Manabe and colleagues (1) on Duration of antibiotic use >5 days 0.70 0.57 1.23 0.70
Clostridium difficile colitis, only a length of stay of 5 days or Watery stool 0.47 0.60 <1.00 >1.00
longer seems to be of value in managing patients (Table 1). The Semi-formed stool 0.67 0.50 1.34 0.66
"ideal" high true-positive rate and low negative likelihood ratio
make this an excellent screening test for ruling out C. difficile * Adapted from Manabe YC, et al. Ann Intern MecI. 1995;123:835-40.
colitis. I disagree that previous use of cephalosporin is an im-
portant predictor, because the positive likelihood ratio is too low
and the negative likelihood ratio is too high. I recommend that To the Editor: We thank Manabe and colleagues (1) for their
the authors evaluate combinations of historical findings to define careful examination of the diagnostic aspects of C. difficile diar-
a positive index test result (2, 3). rhea. Although the case definition required a positive toxin assay
Rarely does a provider rely on one clinical finding to rule in or result, the sensitivity of a single cytotoxin assay was only 79%.
rule out a diagnosis. A positive index test result could be defined Using different case criteria, other investigators have found the
as a combination of use of antibiotics, antibiotic use for at least sensitivity of cytotoxin assay to range from 67% to 74% (2).
5 days, and semiformed or watery stool. If all three findings were Manabe and coworkers report that diagnostic sensitivity may
positive, the false-positive rate would be very low and the posi- be increased by obtaining two to three serial stool specimens for
tive likelihood ratio would be very high. No confirmatory labo- cytotoxin assay. However, Aronsson and colleagues (3) reported
ratory test, including a C. difficile assay, would be necessary to that additional specimens increase the rate of detection by only
make a definitive diagnosis. Similarly, the authors could define a 10%. We suggest the combination of cytotoxin assay and toxi-
positive index test result as a combination of fecal leukocytes, genic culture of the initial specimen as a useful alternative solu-
lactoferrin, and positive Gram stain. The use of combinations of tion to this problem (2) that prevents the need for submitting
clinical characteristics rather than individual tests is valuable in multiple specimens. If the result of the direct cytotoxin assay is
deciding whether to use expensive laboratory tests and is more negative and C. difficile is identified on selective medium, the
characteristic of the problem-solving behavior of clinicians. isolate should be tested for elaboration of cytotoxin in vitro.
Confirmation of toxigenicity is important because nontoxigenic C.
David A. Nardone, MD difficile strains are not pathogenic. The addition of culture im-
Veterans Health Administration proves the sensitivity of laboratory diagnosis to as high as 96%
Portland, OR 97207 (2) and permits strain typing of individual isolates. The latter
facilitates recognition of outbreaks of nosocomial infection (4).
References We have found this combined diagnostic approach to be con-
1. Manabe YC, Vinetz JM, Moore RD, Merz C, Charache P, Bartlett JG. venient, efficient, cost-effective, and reasonably rapid (cytotoxin
Clostridium difficile colitis: an efficient clinical approach to diagnosis. assays may be completed in 24 hours; toxigenic cultures, in 4 to
Ann Intern Med. 1995;123:835-40. 5 days). At the University of Colorado, 41 of 120 culture-con-
2. Deyo RA, Rainville J, Kent DL. What can the history and physical firmed cases of C. difficile diarrhea were cytotoxin negative dur-
examination tell us about low back pain? JAMA. 1992;268:760-5.
3. Nardone DA, Roth KM, Mazur DJ, McAfee JH. Usefulness of physical
ing a 12-month period. Similarly, at Northwestern University's
examination in detecting the presence or absence of anemia. Arch Memorial Hospital, 61 of 160 cases detected during a 6-month
Intern Med. 1990;150:201-4. period were cytotoxin negative. Because we and others (5) have
seen patients who have this disorder but no positive result on
To the Editor: Manabe and colleagues (1) have provided valu- direct-stool cytotoxin assay progress to pancolitis or even death,
able guidelines for diagnosing C. difficile colitis in hospitalized clinicians must recognize the limited sensitivity of cytotoxin as-
patients. The study group was defined as patients who had C. says and the potential utility of culture for the diagnosis of C.
difficile toxin as shown by enzyme-linked immunoassay or cyto- difficile diarrhea.
toxin tissue culture assay. Although this definition of C. difficile
colitis is adequate for clinical purposes, we should remember that Ferric C. Fang, MD
the presence of C. difficile toxin is not always associated with C. University of Colorado Health Sciences Center
difficile colitis. On the basis of a positive cytotoxin tissue culture Denver, CO 80262
assay, C. difficile was present in 34 of 268 patients (12.7%) in
Manabe and colleagues' study. This prevalence is similar to that Dale N. Gerding, MD
of C. difficile positivity seen in 5% to 15% of adults who were Lance R. Peterson, MD
treated with antimicrobial agents but did not have related diar- Chicago Lakeside Veterans Affairs Medical Center
rhea (2). Thus, it would be interesting to know the prevalence of Chicago, IL 60611
C. difficile positivity in the patients in Manabe and coworkers'
study who received antibiotics but did not have diarrhea. References
1. Manabe YC, Vinetz JM, Moore RD, Merz C, Charache P, Bartlett JG.
Anil Minocha, MD Clostridium difficile colitis: an efficient clinical approach to diagnosis.
Robert J. Richards, MD Ann Intern Med. 1995;123:835-40.
University of Oklahoma Health Sciences Center 2. Peterson LR, Kelly PJ. The role of the clinical microbiology laboratory
Oklahoma City, OK 73126 in the management of Clostridium difficile-associated diarrhea. Infect
Dis Clin North Am. 1993;7:277-93.
3. Aronsson B, Mdllby R, Nord CE. Diagnosis and epidemiology of Clos-
References tridium difficile enterocolitis in Sweden. J Antimicrob Chemother. 1984;
1. Manabe YC, Vinetz JM, Moore RD, Merz C, Charache P, Bartlett JG. 14(Suppl D):85-95.
Clostridium difficile colitis: an efficient clinical approach to diagnosis. 4. Gerding DN, Johnson S, Peterson LR, Mulligan ME, Silva J. Clostrid-
Ann Intern Med. 1995;123:835-40. ium difficile-associated diarrhea and colitis: SHEA position paper. Infect
2. Bartlett JG. Pseudomembranous enterocolitis and antibiotic-associated Control Hosp Epidemiol. 1995;16:459-88.
colitis. In: Sleisenger MH, Fordtran JS, eds. Gastrointestinal Disease: 5. Lashner BA, Todorczuk J, Sahm DF, Hanauer SB. Clostridium difficile
Pathophysiology, Diagnosis, Management. Philadelphia: WB Saunders; culture-positive toxin-negative diarrhea. Am J Gastroenterol. 1986;81:
15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6 515
To the Editor: Some points made in Manabe and colleagues' tivity. Rather, we prospectively evaluated the combination of
article on C. difficile colitis (1) deserve to be amplified. The clinical and laboratory data to establish a better clinical sense of
authors evaluated patients in whom the diagnosis of C. difficile how much testing needs to be done. Our previous studies show
colitis was entertained rather than all patients, or even just that the rate of false-negative test results is not greatly influenced
patients with diarrhea. Would the conclusions remain the same by the selection of reagents that are commercially available for
for patients in whom it was not clear whether the diagnosis of C. enzyme immunoassays (3). Most important is the stool dilution
difficile colitis should be considered? used in the assay (4). We believe that the toxigenic culture test
Perhaps the patients in whom the duration of diarrhea was Fang and colleagues propose is unrealistic for routine laboratory
longer might be a subgroup worth analyzing. The results of use in the United States. The attractive feature of this test is its
Manabe and colleagues' study were summarized as significant sensitivity. Its disadvantages are that 1) most clinical laboratories
odds ratios for the following four variables: leukocyte positive by do not offer C. difficile cultures, 2) the anticipated cost would be
lactoferrin assay, positive for fecal leukocytes by Gram stain, high (culture plus toxin assay), 3) results would be delayed 3 to
recent cephalosporin use, and presence of semiformed stool. It is 5 days, and 4) most laboratories do not do strain typing. An
not immediately clear how sensitive and specific the presence of additional concern is specificity. MacFarland and colleagues (5)
positive findings for these variables is, alone or in combination, showed that about 30% of hospitalized patients have C. difficile
for the diagnosis of G difficile colitis. in the absence of diarrhea; most of these strains are toxigenic.
Finally, the statistical analysis included stepwise multiple re- We appreciate the remarks of Drs. Minochi and Richards and
gression. I am surprised that leukocytes positive by lactoferrin agree that the C. difficile toxin assay is associated with false-
and by Gram stain were both shown to be statistically significant; positive results. The literature they cite for this conclusion is our
it would seem that the two should not be independent variables. own, but it is limited to studies of patients receiving a few
Which independent variables might the two tests be measuring? antibiotics. We have even greater concern about specificity with
culture because of high carriage rates of G difficile, as noted
Stephen Holland, MD above. However, to keep the observations in perspective, it is
University of Illinois at Peoria important to emphasize that almost all enteric bacterial patho-
Peoria, IL 61602 gens are associated with relatively high rates of false-positive
results on standard tests. This applies to salmonella, shigella,
Reference Vibrio cholerae, Campylobacter jejuni, and other organisms. The
1. Manabe YC, Vinetz JM, Moore RD, Merz C, Charache P, Bartlett JG. requirement for the clinician is clinical correlation.
Clostridium difficile colitis: an efficient clinical approach to diagnosis. Dr. Holland correctly points out that both lactoferrin assay and
Ann Intern Med. 1995;123:835-40. Gram stain provide different ways of measuring the presence of
fecal leukocytes. We wish to correct the impression that they
In response: We agree with Dr. Nardone that clinicians use were additively predictive of C difficile disease; they were not.
patterns of clinical and laboratory variables to solve problems. We separately analyzed lactoferrin and Gram stain for fecal
We have analyzed the manner in which the sequential addition leukocytes only to validate the utility of lactoferrin as an adjunct
of our identified predictors affected the diagnosis of C. difficile in diagnosing G difficile disease.
disease (Table 2). Although this information may be useful, the
results are not statistically significant. For practical purposes, Joseph M. Vinetz, MD
physicians will suspect C. difficile-associated enteric complications Yukari Manabe, MD
in any patient with otherwise unexplained diarrhea that occurs John G. Bartlett, MD
with exposure to antibiotics. Decisions about use of laboratory Johns Hopkins University School of Medicine
testing are optimally dictated by the effect of the test results on Baltimore, MD 21205
management and by the probability of a positive result. Cepha-
losporins are the major agents implicated in C. difficile-associated References
diarrhea in recent years (1, 2). Usage rates indicate that clinda- 1. Hirschhorn LR, Trnka Y, Onderdonk, Lee ML, Piatt R. Epidemiology
mycin is probably the most commonly used agent, but not of community-acquired Clostridium difficile-associated diarrhea. J Infect
enough patients in our series were receiving clindamycin to allow Dis. 1994;169:127-33.
2. Anand A, Bashey B, Mir T, Glatt AE. Epidemiology, clinical manifes-
us to verify this well-known association. tations, and outcome of Clostridium difficile-associated diarrhea. Am J
Fang and colleagues summarize the findings of some of their Gastroenterol. 1994;89:519-23.
previously published reports on the sensitivity of tests for C. 3. Merz CS, Kramer C, Forman M, Gluck L, Mills K, Senft K, et al.
difficile toxin. The major point of our paper was not that two to Comparison of four commercially available rapid enzyme immunoassays
three stool specimens are needed for adequate diagnostic sensi- with cytotoxin assay for detection of Clostridium difficile toxin(s) from
stool specimens. J Clin Microbiol. 1994;32:1142-7.
4. Laughon BE, Viscidi RP, Gdovin SL, Yolken RH, Bartlett JG. Enzyme
Table 2. Sensitivities of Predictors of Clostridium immunoassays for detection of Clostridium toxins A and B in fecal
specimens. J Infect Dis. 1984;149:781-8.
5. McFarland LV, Mulligan ME, Kwok MS, Stamm WE. Nosocomial
acquisition of Clostridium difficile-associated infection. NEngl J Med.
Clinical Variablet Sensitivity 1989;320:204-10.
First Stool Second Stool
Surgical Treatment of Asymptomatic Carotid
Overall—ElA-positive only 72(31/43) 84 (36/43) Stenosis
Overall—ElA-positive or tissue
culture-positive 81 (35/43) 91 (39/43) To the Editor: The article by Barnett and colleagues on the
Liquid stool 77(10/13) 92(12/13)
Semi-formed stool 83 (25/30) 97 (29/30)
surgical treatment of asymptomatic carotid artery disease (1)
Lactoferrin-positive 90(19/21) 96(20/21) deserves some clarification. First, the authors appear to misin-
Positive Gram stain 100(8/8) 100(8/8) terpret the failure to find statistically significant effects as imply-
Cephalosporin use 81 (22/27) 93 (25/27) ing the absence of an effect. The Asymptomatic Carotid Athero-
Fecal leukocytes* 90(19/21) 95(20/21) sclerosis Study (ACAS) was not designed to have good power to
Presence of any two predictors§ 94(16/17) 100(17/17) find differences in "disabling strokes" alone, to compare the
Presence of any three predictors§ 100(11/11) 100(11/11) treatments in women alone, or to evaluate differences in treat-
ment effect between men and women.
* Values in parentheses are the number of patients. EIA = enzyme immunoassay. Second, we agree with the authors that Doppler flow velocities
t Variable present on first stool specimen.
* By either lactoferrin or Gram stain. reflect the cross-sectional area change of a vessel. It is precisely
§ Predictors from original article. this measure, and not diameter, that is important for determining
516 15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6
arterial pressure and flow. Nonetheless, Barnett and colleagues bers needed to treat to spare stroke at even the highest degrees
are wrong to imply that a "60% ACAS Doppler" stenosis is a of stenosis are prohibitive.
measurement of area. All Doppler laboratories in the ACAS It is disappointing that the ACAS could not identify deciles of
were validated to diameter measurements provided by arterio- stenosis that carry the greatest risk. An annual increase in stroke
grams before being added to the study group. Thus, in the risk as small as 1% would reduce the number needed to treat
ACAS, 60%, 70%, and 80% Doppler stenoses are equivalent to from 67 to 28 and perhaps tip the scales in favor of surgery. It
84%, 91%, and 96% cross-sectional area narrowing, respectively. is a reasonable hypothesis that in asymptomatic patients with the
Third, the authors make much of the stroke risk determined by highest risk (>80% stenosis), favorable benefit will be found. No
arteriographic linear decile measurements. We await their data randomized trial has yet confirmed this hypothesis, but such a
showing good interobserver and intraobserver reliability in read- finding may emerge from the European Asymptomatic Carotid
ing contrast angiograms to the nearest decile. Because the linear Surgery Trial (5).
diameter of a normal carotid artery on digital minimized film is
an average of 3 mm, the measurement of 60%, 70%, and 80% Henry JM. Barnett, MD
diameter stenosis would require exact measurements of 1.2-, 0.9-, Michael Eliasziw, PhD
and 0.6-mm minimal residual lumens, respectively. Given the Heather Meldrum, BA
vagaries of patient movement artifact, contrast-dye load and edge The John P. Robarts Research Institute
definition, projection angle, and intraluminal plaque irregularity, London N6A 5K8, Ontario, Canada
we believe that this task is daunting (perhaps impossible), even if
a jeweler's loupe is used. References
Finally, it is erroneous and misleading to imply that the results 1. Endarterectomy for asymptomatic carotid artery stenosis. Executive
of ACAS are inconclusive. Peer reviewers (who critiqued the Committee for the Asymptomatic Carotid Atherosclerosis Study.
study before it began and periodically while it was in progress), JAMA. 1995;273:1421-8.
the Data and Safety Monitoring Committee, and the unblinded 2. Barnett HJ, Meldrum HE, Eliasziw M. The dilemma of surgical treat-
Statistical Coordinating Center certified that the contrary was ment for patients with asymptomatic carotid disease. Ann Intern Med.
true. The Data Safety and Monitoring Committee stopped the 1995;123:723-5.
3. Beneficial effect of carotid endarterectomy in symptomatic patients with
study earlier than planned because the observed advantage of
high-grade carotid stenosis. North American Symptomatic Carotid End-
surgery was greater than we assumed when designing the study. arterectomy Trial Collaborators. NEngl J Med. 1991;325:445-53.
4. Eliasziw M, Smith RF, Singh N, Holdsworth DW, Fox AJ, Barnett HJ,
John E. Castaldo, MD et al. Further comments on the measurement of carotid stenosis from
James F. Toole, MD angiograms. Stroke. 1994;25:2445-9.
Virginia J. Howard, MSPH 5. Halliday AW, Thomas D, Mansfield A, for the Steering Committee and
ACAS Executive Committee for the collaborators. The Asymptomatic Carotid Surgery Trial (ACST):
Winston-Salem, NC 27157-1078 rationale and design. Eur J Vase Surg. 1994;8:703-10.
1. Barnett HJ, Meldrum HE, Eliasziw M. The dilemma of surgical treat- Abuse History and Gastrointestinal Illness
ment for patients with asymptomatic carotid disease. Ann Intern Med.
To the Editor: I congratulate Dr. Drossman for his timely
review of the management of refractory functional gastrointesti-
In response: We agree completely that data from trial sub- nal disorders (1). He suggests that the predominant symptom
groups should be interpreted cautiously. Nevertheless, it is dis- should determine the best medication (for example, loperamide
appointing that the survival curves for the secondary analysis of for diarrhea and anticholinergic agents for pain). He recom-
disabling stroke remained superimposed for 4.5 years (1). The mends psychoactive medications when pain is unrelated to
lack of benefit for women could be due to small numbers rather changes in gut function, thus downplaying the role of antidepres-
than an absence of benefit. The perioperative complication rate sant agents in functional gastrointestinal disorders in the pres-
of stroke in women (3.6%) was double that in men (1.7%). ence of gut dysfunction. Greenbaum and colleagues (2) have
We did not imply that the ACAS was reporting 60% as an shown a beneficial effect of desipramine on diarrhea and on the
area stenosis. Our comment was that "60% is only equivalent to abdominal pain in diarrhea-predominant irritable bowel syn-
a linear measurement on an arteriogram if the cut-point formula drome. Similarly, Cannon and associates (3) have shown that
of ACAS is applied" (2). We stated that "in clinical practice, the imipramine relieves symptoms in patients who have chest pain
description of a 60% stenosis from a Doppler report may be despite normal coronary angiograms and regardless of abnormal
misinterpreted." Practitioners reading a reported "60% stenosis" results of esophageal motility and provocative testing. The mech-
are less familiar with the nuances of area and linear measure- anism of action of these drugs remains to be established.
ments than are aficionados. We cautioned the readers of the
potential of misrepresenting a ultrasound report that indicates
Anil Minocha, MD
stenosis of 60% by presuming that the report is equivalent to
University of Oklahoma Health Sciences Center
Oklahoma City, OK 73126
Arteriographic linear measurements remain the standard
against which endarterectomy has been evaluated in symptomatic
patients. Good interobserver and intraobserver agreements indi-
cate that with care and patience, arteriographic measurements 1. Drossman DA. Diagnosing and treating patients with refractory func-
tional gastrointestinal disorders. Ann Intern Med. 1995;123:688-97.
are reproducible (3, 4). The stakes are high, and additional 2. Greenbaum DS, Mayle JE, Vanegeren LE, Jerome JA, Mayor JW,
minutes needed for readings are well spent. Greenbaum RB, et al. Effects of desipramine on irritable bowel syn-
The term "inconclusive" denotes that the ACAS results do not drome compared with atropine and placebo. Dig Dis Sci. 1987;32:257-
translate readily into clinical usefulness. Symptomatic and asymp- 66.
tomatic patients with severe stenosis face major differences in 3. Cannon RO, Quyyumi AA, Mincemoyer R, Stine AM, Gracely RH,
risk. The stroke rate in the medical therapy group of the North Smith WB, et al. Imipramine in patients with chest pain despite normal
American Symptomatic Carotid Endarterectomy Trial (NASCET) angiograms. NEngl J Med. 1994;330:1411-7.
was 13.5% per year; in the ACAS, the rate was 2.2%. If the
number of patients needed to treat is used as the measure of In response: I appreciate Dr. Minocha's comments, but they
greatest clinical importance, 67 asymptomatic patients and 6 require some clarification. In my article, I state that "if pain is
symptomatic patients require carotid endarterectomy to prevent the primary problem, it helps to determine whether the pain
one stroke in 2 years. Only skilled surgeons can achieve this occurs in relation to changes in gut function, which would indi-
result. With the morbidity and mortality rate of 4.5% reported in cate the need for a medication directed at the gut (such as an
the other randomized trials on asymptomatic patients, the num- antimotility drug). If the pain is continuous, severe, and unre-
15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6 517
lated to changes in gut function, psychoactive medications for reliable IgD determinations. For example, on the basis of a case
central analgesia [antidepressant agents] are indicated." history and examination of a serum sample, only 6 of 85 Dutch
I agree that antidepressant agents can be used in the presence laboratories could diagnose HIDS correctly. Few specialized lab-
of gut dysfunction; conversely, motility-acting agents are not oratories in the United States routinely measure IgD levels. In
suited for severe, continuous pain with or without gut dysfunc- one study (3), laboratories were asked about measurement of
tion. In addition to the study by Greenbaum and colleagues (1) IgD levels. All laboratories used one radial immunodiffusion
(which showed reduction of diarrheal symptoms and the number method to assess IgD and measured only a few (<20) samples
of slow contractions in the rectosigmoid in patients with diar- weekly (3). Although this method has a limited sensitivity, the
rhea-predominant irritable bowel syndrome), two recent studies range of serum IgD levels in HIDS is such that the diagnosis
have shown positive effects of antidepressant agents on small- should be made. However, most samples in the United States are
bowel motility (2, 3). Imipramine, a tricyclic antidepressant agent shipped, and IgD is susceptible to spontaneous fragmentation
with substantial anticholinergic effects, was shown to slow jejunal during storage due to proteolytic enzymes that are present in
phase III propagation velocity and to prolong orocecal transit serum and may influence the results (4). This problem is avoided
time in normal persons and in patients with diarrhea-predomi- in a recently developed enzyme-linked immunosorbent assay
nant irritable bowel syndrome (2). In another study by the same (ELISA) used to determine IgD levels in human serum. We use
research group (3), administration of paroxetine—a selective se- a commercial monoclonal antibody that binds to Fc regions of
rotonin reuptake inhibitor—reduced orocecal transit time in con- IgD (Dako, Copenhagen, Denmark) and human serum with
trols and patients with diarrhea-predominant irritable bowel syn- known IgD concentration (Behring, Marburg, Germany) as a
drome. No data were available on possible effects on symptoms. reference. The lower limit of detection with our ELISA is 3
To expand on my previous recommendations, I emphasize U/mL.
three points: First, antidepressant agents may improve symptoms We found that storing IgD for as long as 5 months or repeated
of the irritable bowel syndrome by affecting motility, independent freeze-thawing cycles had no influence on the detection of IgD.
of mood-altering or analgesic effects, but confirmatory studies This suggests that fragmentation does not influence the results of
correlating symptoms with changes in motility and adjusting for our assay. The use of an accurate ELISA for measuring IgD
depressive mood are needed. A practical approach would be to levels could enhance recognition of cases of HIDS in the United
choose a tricyclic agent with anticholinergic properties for diar- States.
rhea-predominant irritable bowel syndrome and a selective sero-
tonin reuptake inhibitor for constipation-predominant irritable Joost PH. Drenth, MD, PhD
bowel syndrome. Ina S. Klasen, PhD
Second, motility-altering agents are better suited for treating Jos W.M. van der Meer, MD, PhD
such symptoms as postprandial pain and diarrhea on an as- University Hospital St. Radboud
needed basis. Conversely, because antidepressant agents require 6500 HB Nijmegen, the Netherlands
several weeks to become effective and have a long duration of
action, they are best prescribed when symptoms are frequent or References
1. Drenth JP, Haagsma CJ, van der Meer JW, and the International
Finally, the role for motility-altering agents in other functional Hyper-IgD Study Group. Hyperimmunoglobulinemia D and periodic
gastrointestinal disorders (for example, functional dyspepsia, fever syndrome. The clinical spectrum in a series of 50 patients. Med-
esophageal motility disorders, and anorectal disorders) is not well icine (Baltimore). 1994;73:133-44.
established, requires further study, and must be determined on 2. Grose C, Schnetzer JR, Ferrante A, Vladutiu AO. Children in hyper-
an individual basis. Because of their central analgesic effects, immunoglobulinemia and periodic fever syndrome. Pediatr Infect Dis.
however, antidepressant agents may help when symptoms are 1996;15:72-7.
severe and refractory. 3. Mancini G, Carbonara AO, Heremans JF. Immunochemical quantita-
tion of antigens by single radial immunodiffusion. Immunochemistry.
Douglas A. Drossman, MD 4. Skvaril F, Radl J. The fragmentation of IgD during storage. Clin Chim
University of North Carolina Acta. 1967;15:544-6.
Chapel Hill, NC 27599
1. Greenbaum DS, Mayle JE, Vanegeren LE, Jerome JA, Mayor JW, Meta-Analysis and Bouillabaisse
Greenbaum RB, et al. Effects of desipramine on IBS compared with
atropine and placebo. Dig Dis Sci. 1987;32:257-66.
2. Gorard DA, Libby GW, Farthing MJ. Effect of a tricyclic antidepressant To the Editor: In his thoughtful editorial on the recent contro-
on small intestinal motility in health and diarrhea-predominant irritable versy over calcium channel blockers (1), Dr. Messerli correctly
bowel syndrome. Dig Dis Sci. 1995;40:86-95. points out that the research methods used in two recent studies
3. Gorard DA, Libby GW, Farthing JG. Influence of antidepressants on on this topic (2, 3) are prone to various biases. Although we
whole gut orocaecal transit times in health and irritable bowel syn- agree with the general comments made in the editorial, we make
drome. Aliment Pharmacol Ther. 1994;8:159-66.
three points about the case-control study by Psaty and col-
Dr. Messerli states that this study may have been subject to
selection bias and that hypertensive patients with coronary artery
Recognition of IgD and Periodic Fever disease may have been more likely to be treated with calcium
antagonists than with diuretics. Psaty and colleagues, however,
To the Editor: The hyperimmunoglobulinemia D and periodic present results that address this concern: Among persons with
fever syndrome (HIDS) is characterized by recurrent febrile at- cardiovascular disease, 52.5% were receiving diuretics and only
tacks with abdominal symptoms, joint involvement, skin lesions, 37.4% were receiving calcium channel blockers. Furthermore, all
and lymphadenopathy. The syndrome has been diagnosed in 66 persons with known cardiovascular disease were excluded from
patients, most from Europe (1). The clinical picture and the the study's principal analysis.
elevated serum IgD levels (MOO U/mL) complete the diagnosis. A second form of bias that can arise in case-control studies is
Although no treatment is available, a correct diagnosis removes recall bias, in which differential recollection of previous expo-
uncertainty and allows the patient to be informed on the benign sures can lead to spurious associations between disease and
prognosis of the syndrome. exposure variables. Although recall bias may have affected the
To our knowledge, HIDS has been diagnosed in only two recording of potential confounding variables, the exposure vari-
patients in the United States (2). Although population variation able of medication use was determined from computerized phar-
may explain the lack of cases, we believe that under-reporting of macy data, not from patient recall.
cases plays a substantial role. Possible reasons for such under- Finally, Psaty and colleagues did their study on a well-defined
reporting include lack of awareness and limited availability of patient sample—enrollees in the Group Health Cooperative of
518 15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6
Puget Sound. Through careful screening of computerized data, References
researchers could identity 95% of incident myocardial infarctions. 1. Messerli FH. Case-control study, meta-analysis, and bouillabaisse: put-
Sampling bias was thus less likely because the study cases were ting the calcium antagonist scare into context. Ann Intern Med. 1995;
representative of incident disease in the underlying patient pop- 11:888-9.
ulation. 2. Psaty BM, Heckbert SR, Koepsell TD, Siscovick DS, Raghunathan TE,
Dr. Messerli is correct in stating that case-control studies are Weiss NS, et al. The risk of myocardial infarction associated with
antihypertensive drug therapies. JAMA. 1995;274:620-5.
like bouillabaisse: "One tainted fish will spoil the pot." However, 3. Joint National Committee on Detection, Evaluation, and Treatment of
clinicians should recognize and appreciate the substantial High Blood Pressure. The Fifth Report of the Joint National Commit-
strengths of the study by Psaty and coworkers and should not tee on Detection, Evaluation, and Treatment of High Blood Pressure
dismiss its findings solely on the basis of its case-control design. (JNC V). Arch Intern Med. 1993;153:154-83.
William A. Ghali, MD, MPH In response: As Drs. Ghali and Hershman point out, Psaty and
Warren Y. Hershman, MD, MPH colleagues attempted to avoid an indication bias. No matter how
Boston University Medical Center carefully a retrospective case control-study is done, however, an
Boston, MA 02118 indication bias can never be ruled out: In retrospect it is impos-
sible to second guess the reason a specific cardiovascular drug
References was prescribed for a given patient. In another report from the
1. Messerli FH. Case-control study, meta-analysis, and bouillabaisse: put- same patient cohort (1), the authors freely admit that "as ex-
ting the calcium antagonist scare into context [Editorial]. Ann Intern pected those with cardiovascular disease were more likely to be
Med. 1995;123:888-9. taking calcium channel blockers and beta blockers " Because
2. Psaty BM, Heckbert SR, Koepsell TD, Siscovick DS, Raghunathan TE, all patients had hypertension (which, by definition, is a cardio-
Weiss NS, et al. The risk of myocardial infarction associated with vascular disease), it was impossible to exclude persons with
antihypertensive drug therapies. JAMA. 1995;274:620-5. known cardiovascular disease.
3. Furberg CD, Psaty BM, Meyer JV. Nifedipine. Dose-related increase in It is by no means unexpected that patients receiving /3-blockers
mortality in patients with coronary heart disease. Circulation. 1995;92:
had better outcomes than did those receiving calcium antago-
nists. Given that most of these short-acting calcium antagonists
were not approved for treating hypertension and that they were
prescribed at an inappropriate dosing schedule (that is, once a
To the Editor: I read with interest the recent editorial by Dr.
day or twice a day as Dr. Psaty recently conceded ), the
Messerli (1). As both a health services researcher-clinical epidem-
outcome is not surprising. Nifedipine capsules given once or
iologist and a practicing general internist, I have found the
controversy about calcium channel blockers to be enlightening twice a day will, at best, decrease arterial pressure for a few
and frustrating. I strongly agree with Dr. Messerli that controlled hours only and, at worst, lead to hypotension, sympathetic stim-
studies documenting the safety and efficacy of calcium channel ulation, reflexive tachycardia, and possibly myocardial ischemia
blockers are desperately needed. (It is interesting that such trials (3). Such a regimen deviates from accepted standards of medical
have not been done, and yet billions of dollars are spent each practice, and it should be no surprise when the outcome also
year prescribing and marketing these drugs.) deviates from these standards.
Finally, I am not certain about the take-home message of Psaty
I do, however, disagree with two of his other statements.
and colleagues' study to strictly follow the Joint National Com-
Messerli correctly states that case-control studies are subject to
possible selection bias. This means that if physicians preferen- mittee guidelines. Recent case-control studies (4, 5) have shown
tially indicate that patients are at risk for death (presumably an increased risk for sudden death with j3-blockers and diuretics
because of ischemic heart disease) with calcium channel blockers compared with other therapies for hypertension. Interestingly,
rather than with other antihypertensive agents, the increased rate news media coverage of these reports was minimal compared
of myocardial infarction may be caused by the indication, not the with that elicited by the calcium antagonists. These data notwith-
drug. He fails to mention, however, that Psaty and colleagues (2) standing, diuretics are the only drug class for which a reduction
have shown a clear dose-related decrease in the rate of myocar- in mortality has been unequivocally shown in patients with hy-
dial infarctions among hypertensive patients receiving j3-blockers pertension. Such data are lacking for all others (such as j3-block-
(despite a dose-related increase in the rate of myocardial infarc- ers, a-blockers, angiotensin-converting enzyme inhibitors, and an-
tion in hypertensive patients receiving calcium channel blockers). giotensin receptor blockers). Why single out calcium antagonists?
j3-blockers are subject to the same selection bias as are calcium Perhaps we should be more concerned about the ink that is
channel blockers; however, patients at Group Health Coopera- currently wasted on misleading case-control studies, meta-anal-
tive of Puget Sound who take j3-blockers fare much better than yses, and other "bouillabaisse" and about the underlying reasons
do those who take calcium channel blockers. for the excessive media coverage than about the safety of the
long-acting calcium antagonists.
Second, Messerli shows results for four placebo-controlled tri-
als indicating that patients who receive long-acting calcium chan-
nel blockers survive longer than do patients who receive placebo. Franz H Messerli, MD
This finding is misleading in two ways. First, only one of the Ochsner Clinic and Alton Ochsner Medical Foundation
trials examined patients with hypertension. Psaty and colleagues New Orleans, LA 70121
did not contend that calcium channel blockers have no potential
uses. They clearly state that these agents are associated with References
increased risk for myocardial infarction in patients with hyper- 1. Psaty BM, Lee M, Savage PJ, Rutan GH, German PS, Lyles M.
tension. Second, the comparison group in each of these studies Assessing the use of medications in the elderly: methods and initial
received a placebo, but none of the Group Health patients in experience in the Cardiovascular Health Study. The Cardiovascular
Psaty and colleagues' study did. Patients receiving calcium chan- Health Study Collaborative Research Group. J Clin Epidemiol. 1992;
nel blockers had an increased risk for myocardial infarction 2. Psaty BM. U.S. Food and Drug Administration hearing. 25 January
compared with patients receiving other drugs. The take-home 1996.
message from Psaty and colleagues' study was that the Joint 3. Wilson DC, Schwarts GL, Textor SC, Zachariah PK, Sheps SG. Pre-
National Committee on Detection, Evaluation, and Treatment of cipitous fall in blood pressure in the treatment of chronic hypertension.
High Blood Pressure guidelines (3) should be followed: Hyper- Proceedings of the 5th International Symposium on Calcium Antago-
tension should be initially treated with diuretics or /3-blockers, or nists: Pharmacology and Clinical Research, Houston, Texas, September
both, and calcium channel blockers should be second- or third- 1991.
line agents. 4. Hoes AW, Grobbee DE, Lubsen J, Man in 't Veld AJ, van der Does E,
Hofman A. Diuretics, j3-blockers, and the risk for sudden cardiac death
in hypertensive patients. Ann Intern Med. 1995;123:481-7.
William M. Tierney, MD 5. Siscovick DS, Raghunathan TE, Psaty BM, Koepsell TD, Wicklund KG,
Indiana University School of Medicine Lin X, et al. Diuretic therapy for hypertension and the risk of primary
Indianapolis, IN 46202 cardiac arrest. N Engl J Med. 1994;330:1852-7.
15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6 519
Table 1 . Plasma TGF-/51 Levels in Healthy Donors and Patients w i t h the Sepsis Syndrome*
TGF-/31 Level /> Value
Healthy Donors Patients with the Sepsis
(n = 21) Syndrome (n = 26)
TGF-/31 measured in normal plasma 13.0 ± 1.4 (4.8-26.6) 27.0 ± 4 . 5 (2.9-102.4) 0.01t
TGF-/31 measured in platelet-poor plasma 9.9 ± 1.1 (3.2-20.6) 17.4 ± 2 . 8 (3.4-68.1) 0.05
* Levels were measured using an enzyme-linked immunosorbent assay. Values are expressed as the mean ± SD (range). TGF-/31 = transforming growth factor-/31.
t P value determined by using the Mann-Whitney U test.
Rating Long-Term Care Facilities on Pressure Elevated Levels of Circulating Transforming
Ulcers Growth Factor-/31 in Patients with the Sepsis
To the Editor: Berlowitz and colleagues (1) used a large Vet-
erans Administration database to identify clinical and functional To the Editor: The sepsis syndrome is associated with the
status variables that predict pressure ulcer development. Al- presence of circulating proinflammatory cytokines. Recent inves-
though they refer to poor nutrition as a known risk factor, their tigations have indicated the presence of naturally occurring anti-
multivariable analysis identifies no variable that is closely related inflammatory mediators (such as interleukin-lra, soluble tumor
to nutritional status. At least three possible explanations exist. necrosis factor receptor [sTNF R], and interleukin-10). Among
One explanation is simply that the database did not contain the anti-inflammatory cytokines, transforming growth factor-/31
information on nutrient intake or "nutritional status." The sec- (TGF-/31) has been shown to repress the production of inflam-
ond is that some nutritional variables were included, but no matory cytokines by activated macrophages (1) and to induce the
association was found. The associations among nutrient intake, release of sTNF R and interleukin-lra (2). Circulating TGF-/31
nutritional status, and the presence or development of pressure levels are present in healthy persons, and the role of TGF-/31 in
ulcers are in fact surprisingly loose (2). Malnutrition is generally homeostasis has been clearly established in knock-out mice,
automatically acknowledged as a risk factor, although many stud- which die within 3 weeks of developing multifocal inflammatory
ies (perhaps this one) found no association between nutrition disease (3).
and pressure ulcers. We investigated whether the levels of circulating TGF-/31 were
The third possibility is complex. The term "malnourished" can increased in patients with the sepsis syndrome. We measured the
refer to an otherwise healthy, starving child, a case in which levels of plasma TGF-j31 in 26 patients at the time of diagnosis.
providing food would clearly help. The term can also describe a As shown in Table 1, mean plasma levels of TGF-/31 were
cachectic patient with metastatic lung cancer: All measures of significantly higher in septic patients than in healthy donors. In 8
nutritional status are abnormal, but dietary intervention to aid patients, the circulating TGF-/31 level was greater than 26.6
survival is useless. Some nursing home residents may more ng/mL, which was the highest level observed in the healthy
closely resemble the latter than the former. Nutrient intake and donors. When levels in platelet-poor plasma were measured, the
nutritional variables might both be affected (although differently) difference between the groups was borderline. Levels of TGF-/31
by chronic, wasting illness or by extreme, fragile debility, but were not correlated with outcome or other cytokine levels. These
aggressive provision of nutrients would be fruitless. results indicate that high levels of circulating TGF-/31 can be
If analysis of this large database shows no relation between associated with the sepsis syndrome. The contribution of platelets
nutrition and development of pressure ulcers, I believe the au- as a source of this cytokine in plasma may be increased during
thors should emphasize this fact. Currently, many persons are sepsis. The immunosuppression seen in experimental sepsis and
attached to feeding tubes or receive dietary supplements on the in patients with the septis syndrome might reflect the presence of
basis of a belief that such a relation exists. increased levels of TGF/31 (4).
In conclusion, it appears that what Louis Pasteur called "Na-
Thomas E. Finucane, MD tura medicatrix" when he studied puerperal sepsis (5) occurs in
Johns Hopkins University School of Medicine many patients with the sepsis syndrome. In other words, the
Baltimore, MD 21224 systemic inflammatory response syndrome is associated not only
with the exacerbation of the production of proinflammatory cy-
References tokines but also with the increased release of many anti-inflam-
1. Berlowitz DR, Ash AS, Brandeis GH, Brand HK, Halpern JL, Mos- matory actors, including specific interleukin-1 and tumor necrosis
kowitz MA. Rating long-term care facilities on pressure ulcer develop- factor inhibitors and such cytokines as interleukin-10 and TGF-/31.
ment: importance of case-mix adjustment. Ann Intern Med. 1996; 124:
557-63. Christelle Marie, MSc
2. Finucane TE. Malnutrition, tube feeding and pressure sores: data are
incomplete. J Am Geriatr Soc. 1995;43:447-51.
Jean-Marc Cavaillon, PhD
In response: We agree with Dr. Finucane that the association
among nutrient intake, measures of nutritional status, and the Marie-Reine Losser, MD
development of pressure ulcers is complex. Unfortunately, our Hopital Lariboisiere
database did not contain information on these variables. Such Paris, France
information would have been useful, both to better define this
clinical condition and to understand differences in patient mix
that may affect a facility's rate of pressure ulcer development. References
Further studies are needed in this area. 1. Bogdan C, Nathan C. Modulations of macrophage function by trans-
forming growth factor-/3, interleukin-4 and interleukin-10. Ann N Y
Acad Sci. 1993;685:713-39.
Dan R Berlowitz, MD, MPH 2. Turner M, Chantry D, Katsikiq P, Berger A, Brennan FM, Feldman M.
Gary H. Brandeis, MD Induction of the interleukin-1 receptor antagonist protein by transform-
Mark A. Moskomtz, MD ing growth factor-j3. Eur J Immunol. 1991;21:1635-9.
Center for Health Quality, Outcomes, and Economic Research 3. Shull MM, Ormsby I, Kier AB, Pawlowski S, Diebold RJ, Yin M, et al.
Bedford, MA 01730 Targeted disruption of the mouse transforming growth factor-/31 gene
520 15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6
results in multifocal inflammatory disease. Nature. 1992;359:693-9. Enoxaparin-Induced Skin Necrosis
4. Miller-Graziano CL, Szabo G, Griffey K, Mehta B, Kodys K, Catalano
D. Role of elevated monocyte transforming growth factor-j3 production
in post-trauma immunosuppression. J Clin Immunol. 1991;11:95-102. To the Editor: Enoxaparin, a low-molecular-weight heparin
5. Pasteur L. Compte Rendu de la Seance l'Academie des Sciences. 3 used to treat and prevent deep venous thrombosis, has been
May 1880. evaluated in several clinical trials (1, 2). The wide use of the drug
makes it imperative that possible side effects, including skin
reactions, be studied (3, 4). We report the rare occurrence of
skin necrosis that developed after enoxaparin administration. To
our knowledge, previous reports of side effects of enoxaparin
Non-Insulin-Dependent Diabetes Mellitus have described local skin inflammation and pruritis, but not
Developing during Interferon-a Therapy for necrosis. One report (5) noted that a different low-molecular-
Chronic Hepatitis C weight heparin caused skin necrosis. We surveyed approximately
8000 subcutaneous injections of enoxaparin (40 mg once a day or
20 mg twice a day) administered in the Meir General Hospital in
To the Editor: Development of non-insulin-dependent (auto- Kfar-Saba, Israel, in 1993-1994. The only remarkable complica-
immune) diabetes mellitus has been suggested in patients treated tions we identified were two cases of skin necrosis.
with interferon-a (1-3). We describe a patient who developed
this disease with no evidence of an immune process while receiv- Patient 1, a 59-year-old woman, had surgery for myoma of the
ing interferon-a for hepatitis C. uterus. Because of her history of atrial fibrillation, treatment with
enoxaparin, 40 mg in a single daily subcutaneous injection, was
A 50-year-old man was referred to us because of chronic active
hepatitis C. His cousin had insulin-dependent diabetes, but no initiated. The drug was given alternately in the right and left
family member had a history of non-insulin-dependent diabetes. arms. Six days later, the injection sites on both arms became red,
Recombinant interferon-a was started at a dose of 32 million U tender, and indurated. A small black area, 25 mm in diameter,
three times a week. Before therapy, the patient had glucose appeared on the left arm. Examination of a biopsy specimen of
intolerance. After 1 week of treatment, his fasting plasma blood the lesion indicated severe vasculitis in the dermis and in the
glucose level was 5.5 mmol/L. After 7 weeks of therapy, the dermal-epidermal junction. Results of tests of blood clotting and
glucose level suddenly increased to 25.6 mmol/L. The patient's blood count were within normal ranges. Enoxaparin treatment
hemoglobin A l c value was 9.5%. The thyroid-stimulating hor- was discontinued, and the necrotic wound was treated with silver
mone value was in the normal range, and tests for thyroid sulphadiazine cream. After local debridement, the wound was
peroxidase antibodies were negative. Interferon-a therapy was allowed to close spontaneously.
discontinued, and diabetes was controlled after 0.55 U of insulin Patient 2 was a 34-year-old women in her 42nd week of
per kg of body weight per day was given for 1 week. The pregnancy. She had a cesarean delivery because of rupture of the
patient's insulin requirement then decreased quickly, and he uterus and fetal distress. When deep venous thrombosis devel-
stopped taking insulin 3 months after discontinuing interferon-a oped in her right foot after surgery, the patient was treated with
therapy. His glucose levels were well controlled by diet alone. 40 mg of subcutaneous enoxaparin injections once a day. The
Results of tests for islet-cell antibodies by immunofluorescence drug was given periumbilically or in either arm. A purple-red,
and for insulin autoantibodies by immunoprecipitation were neg- tender area that appeared in the periumbilical region 10 days
ative before interferon-a therapy and 10 days and 1.5 months later became progressively necrotic (Figure 1). Examination of a
after interferon-a therapy was discontinued. Results of tests for wedge biopsy specimen showed epidermal necrosis and hyalin-
glutamic acid decarboxylase antibodies by immunoprecipitation ized vessels with mononuclear perivascular infiltration in the
were negative 10 days after cessation of interferon-a therapy. dermis. After local treatment with silver sulphadiazine cream, the
The patient's basal and poststimulation C peptide values were necrotic wound was debrided and left to close spontaneously.
not low before, while, or after insulin therapy was received. Skin necrosis occurring after administration of a low-molecu-
Previous reports (1-3) have described patients with islet-cell lar-weight heparin is rare, and the pathogenesis of this compli-
antibodies and insulin autoantibodies during interferon-a therapy cation is not well understood. Our experience may alert medical
or when diabetes was diagnosed. Pancreatic autoimmunity prob- personnel to the possibility of such a side effect.
ably did not occur in our patient because he had no autoimmune
markers. Interferon-a could have led to reduced sensitivity of the
patient's liver and peripheral tissues to insulin and to exacerba- Moshe Fried, MD
tion of preexisting glucose intolerance (4). The high doses of Sraya Kahanovich, MD
interferon-a our patient received could explain the intensity of Rephael Dagan (Reiss), MD
the metabolic disorder. We believe that interferon-a can exacer- Meir General Hospital
bate glucose intolerance independently of an autoimmune pro- 44281 Kfar-Saba, Israel
cess, especially if given in high doses.
Pierre Chedin, MD
Juliette Cahen-Varsaux, MD
Victor Dupony Hospital
95107 Argenteuil, France
Nathalie Boyer, MD
92118 Clichy, France
1. Fabris P, Betterle C, Floreani A, Greggio NA, De Lazzari F, Naccarato
R, et al. Development of type 1 diabetes mellitus during interferon-alfa
therapy for chronic HCV hepatitis. Lancet. 1992;340:548.
2. Murakami M, Iriuchijima T, Masatomo M. Diabetes mellitus and
interferon-a therapy [Letter]. Ann Intern Med. 1995;123:318.
3. Guerci AP, Guerci B, Levy-Marchal C, Ongagna J, Ziegler O, Candilo-
ros H, et al. Onset of insulin-dependent diabetes mellitus after inter-
feron-alfa therapy for hairy cell leukemia. Lancet. 1994;343:1167-8.
4. Koivisto VA, Peklonen R, Cantel K. Effect of interferon on glucose Figure 1. Necrosis in periumbilical area of patient 2 that devel-
tolerance and insulin sensitivity. Diabetes. 1989;38:641-7. oped after treatment with enoxaparin.
15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6 521
References significant, and they place patients at greater risk by doing radial
1. Noble S, Peters DH, Goa KL. Enoxaparin: a reappraisal of its pharma- artery puncture or cannulation. However, Williams and Schenken
cology and clinical applications in the prevention and treatment of (4) reviewed 9020 cases in the literature and their own experi-
thromboembolic disease. Drugs. 1995;49:388-410. ence with more than 25 000 radical artery punctures. They found
2. Spiro TE, Johnson GJ, Christie MJ, Lyons RM, MacFarlane DE, that these procedures were all done with no ischemic complica-
Blasier RB, et al. Efficacy and safety of enoxaparin to prevent deep tions and concluded that the risk for complication is so low that
venous thrombosis after hip replacement surgery. Ann Intern Med.
1994;121:81-9. prescreening with an Allen test is unwarranted. This conclusion
3. Manoharan A. Heparin-induced skin reaction with low-molecular- would not be useful in the patient in the intensive care unit who
weight heparin. Eur J Haematol. 1992;48:234. is likely to be at higher risk for ischemia because of the use of
4. Phillips JK, Majumdar G, Hunt BJ, Savidge GF. Heparin-induced skin vasoactive drugs and hypotension.
reaction due to two different preparations of low molecular weight Cases of false-negative ulnar flow and false-positive ulnar flow
heparin (LMWH). Br J Haematol. 1993;84:349-50. should be identified. Glavin and Jones reported a sensitivity of
5. Ojeda E, del Carmen Perez M, Mataix R, et al. Skin necrosis with low 1.0 for oximetry compared with Doppler analysis of ulnar flow.
molecular weight heparin. Br J Haematol. 1992;82:620. Interestingly, they cited seven patients in their group in whom
flow was detected by oximetry but not by Doppler. Unfortu-
nately, they had no patients in whom oximetry failed to detect
flow. This finding limits their conclusions. My colleagues and I
Use of Pulse Oximetry for Assessing Ulnar have frequently identified patients in whom ulnar flow was com-
Collateral Flow promised enough to warrant use of another site.
I agree with Dr. Kruse that diagnostic groups in which oxim-
To the Editor: O'Mara and Sullivan (1) report the anecdotal etry is not useful must be reported. We want to avoid the
use of pulse oximetry for evaluating collateral patency before chauvinism of the Allen test, which may erroneously exclude too
radial artery cannulation, and they propose that the technique be many cases from radial use, and the liberalism of "no testing,"
called the Sullivan test. Other reports on this use of finger pulse which may cause complications in the critically ill patient even if
oximetry and pulse transduction have been published, and there it does not cause complications in the general hospital popula-
is a potentially important distinction between the two methods. tion.
Fuhrman and colleagues (2) found that finger plethysmography
could detect abnormal circulatory patterns but that significant Karen O'Mara, DO
changes in blood flow in the hand did not result in decreased Resurrection Medical Center
oxyhemoglobin saturation as assessed by pulse oximetry. They Chicago, IL 60631
concluded that "plethysmography can be used to demonstrate
palmar collateral circulation, but pulse oximetry cannot." On the References
basis of results of their own studies, Glavin and Jones (3) rec- 1. Glavin RJ, Jones HM. Assessing collateral circulation in the hand—
ommended that pulse monitors and pulse oximetry not be used four methods compared. Anaesthesia. 1989;44:594-5.
together for this purpose. 2. Allen EV. Thromboangitis obliterans: methods of diagnosis of chronic
Notwithstanding the controversy over the necessity of collateral occlusive arterial lesions distal to the wrist with illustrative cases. Am J
flow assessment in general (4) and the findings of Glavin and Med Sci. 1929;178:237-44.
3. Mangano DT, Hickey RF. Ischemic injury following uncomplicated ra-
Jones (3), it appears that finger pulse detection may have merit dial artey catheterization. Anesth Analg. 1979;58:55-7.
for assessing ulnar collateral flow. However, further study is 4. Williams T, Schenken JR. Radial artery puncture and the Allen test.
required before pulse oximetry can be routinely recommended Ann Intern Med. 1987;106:164-5.
for this application.
James A. Kruse, MD
Wayne State University Microsporidia in Humans
Detroit, MI 48201
To the Editor: Microsporidia are ubiquitous, obligate intracel-
lular parasites that can infect all classes of animals. Five gen-
1. O'Mara K, Sullivan B. A simple bedside test to identity ulnar collateral era—Enterocytozoon, Encephalitozoon, Septata, Pleistophora, and
flow [Letter]. Ann Intern Med. 1995;123:637.
2. Fuhrman TM, Pippin WD, Talmadge LA, Reilley TE. Evaluation of Nosema—and a sixth taxon, which contains unclassified micro-
collateral circulation of the hand. J Clin Monit. 1992;8:28-32. sporidia, infect humans (1, 2). The sources and modes of trans-
3. Glavin RJ, Jones HM. Assessing collateral circulation in the hand— mission remain unknown (1, 2). Microsporidia species infect the
four methods compared. Anaesthesia. 1989;44:594-5. central nervous system in animals, but brain involvement in hu-
4. Williams T, Schenken JR. Radial artery puncture and the Allen test. mans has not been reported (2).
Ann Intern Med. 1987;106:164-5. A 33-year-old man presented with transient right-sided hemi-
paresis and left temporal headache. The patient had a 10-year
In response: Ulnar collateral circulation should be assured be- history of human immunodeficiency virus (HIV) infection (CD4
fore a procedure is done, given the reported incidence of radial count, 4 cells/mm3) and cytomegalovirus retinitis. His medica-
artery thrombosis or occlusion and the reported lack of ulnar tions included trimethoprim-sulfamethoxazole, fluconazole, and
collateral flow in 5% to 6% of patients (1). The Allen test has ganciclovir. A physical examination showed wasting and psycho-
been historically recommended for the bedside screening of ulnar motor retardation. His cerebrospinal fluid protein level was 100
flow (2). Case reports of limb ischemia after radial artery can- mg/dL. Serum and cerebrospinal fluid cryptococcal antigen and
nulation despite a negative Allen test result have been attributed rapid plasma reantigen results were negative. Magnetic reso-
to emboli (3), not inadequate ulnar collateral circulation that was nance imaging showed 24 lesions (sizes ranged from 0.5 to 1.5
undetected. cm), with minimal mass effect. The patient's serum Toxoplasma
The use of pulse oximetry has been recommended for enhanc- gondii antibody titer was less than 1:16, and his cysticercosis titer
ing the interpretation of results of this test and has been modi- was less than 1:10. He received 1 week of empirical antitoxoplas-
fied for critically ill or uncooperative patients. In pulse oximetry mosis therapy, which was discontinued because of intolerance. A
and the Allen test, the results are merely positive or negative and neurosurgical consultant believed that a biopsy was not war-
cannot be interpreted as more than that. Both tests identify ranted. The patient's condition continued to deteriorate; he de-
patients whose low flow state or undetected ulnar flow with veloped more seizures and died 2 months after his initial pre-
radial artery occlusion places them at risk. The Allen test, how- sentation.
ever, requires patient cooperation. These tests may erroneously The autopsy was limited to the brain and showed an organism
exclude some patients who have subclinical flow that could be that resembled both Pleistophora and Thelohania species. The
detected by other means. exact identification of the infecting organism awaits further anal-
Clinicians need to know whether subtler flow changes are ysis (3).
522 15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6
Toxoplasma gondii is the most common opportunistic pathogen
causing focal central nervous system lesions in patients with the
acquired immunodeficiency syndrome (AIDS). Infection with this
organism was unlikely in our patient because of trimethoprim-
sulfamethoxazole prophylaxis, a negative serologic result for T.
gondii (4), and the many central nervous system lesions.
The standard approach to central nervous system lesions in
patients with AIDS is empirical therapy for toxoplasmosis; re-
fractory cases are considered for brain biopsy. A more aggressive
approach to central nervous system lesions may be needed in a
select subset of patients with AIDS. Earlier diagnosis might lead
to administration of appropriate therapy and could prevent po-
tential toxicities from empirical therapy. Albendazole may be
used to treat central nervous system microsporidiosis. This drug
is also effective for microsporidial diarrhea (5) and in other
parasitic infections of the brain.
Recognition of central nervous system microsporidiosis may
contribute to better management of HIV-infected patients.
Johannes Berg, MD
Leslie E. Diaz, MD
Bradley S. Bender, MD
Veterans Affairs Medical Center
Gainesville, FL 32608-1197
1. Weber R, Bryan RT, Schwartz DA, Owen RL. Human microsporidial
infections. Clin Microbiol Rev. 1994;7:426-61.
2. Canning EU, Lom J. The microsporidia of vertebrates. London: Aca-
demic Pr; 1986.
3. Yachnis AT, Berg J, Martinez-Salazar A, Bender BS, Diaz L, Rojiani Figure 1. Transmission electron micrograph showing attaching
AM, et al. Disseminated microsporidiosis especially involving the CNS, and effacing lesions on the surfaces of colonic mucosal epithelial
heart, and kidneys: report of a newly recognized pansporoblastic species cells. Small rod-shaped bacteria tightly adhere to the surfaces of epithelial
in two symptomatic AIDS patients. Am J Clin Pathol. [In press], cells, which have lost their microvilli. Other bacteria are present in endocy-
4. Carr A, Tindall B, Brew BJ, Marriott DJ, Harkness JL, Penny R, et al. totic vacuoles within the epithelial cells. (Original magnification x 7000.)
Low-dose trimethoprim-sulfamethoxazole prophylaxis for toxoplasmic
encephalitis in patients with AIDS. Ann Intern Med. 1992;117:106-11.
5. Orenstein JM, Dieterich DT, Lew EA, Kotler DP. Albendazole as a hybridization and was negative for the EPEC-specific locus of the
treatment for intestinal and disseminated microsporidiosis due to Sep-
tata intestinalis in AIDS patients: a report of four patients. AIDS. EPEC adherence factor and bundle-forming pilus gene or with
1993;7(Suppl 3):S40-2. the E. co/i-specific enterotoxin probes (4); 2) the fluorescent
actin stain assay showed the AE interaction between the non-
lactose-fermenting isolate and HCT-8 cells; and 3) the patient's
diarrhea resolved promptly after treatment with oral ciprofloxa-
cin (no treatment was given for the microsporidia).
Enteroadherent efl&4-Positive Escherichia coli Further prospective studies are needed to elucidate the role of
Associated with Chronic AIDS-Related Diarrhea potentially diarrheagenic E. coli organisms in chronic diarrhea
and the related wasting syndrome seen in HIV-infected patients.
To the Editor: Chronic diarrhea is a substantial contributor to
morbidity in patients with human immunodeficiency virus (HIV) John H. Hii, MD
infection. Even after extensive evaluation, a pathogen often can- John G. Guccion, MD
not be identified (1). Recent studies of patients with diarrhea George Washington University Medical Center
associated with the acquired immunodeficiency syndrome have Washington, DC 20037
described histopathologic findings of enteroadherent bacteria
that produced attaching and effacing (AE) lesions on the intes- Cynthia L. Gilbert, MD
tinal cell (2, 3). Such lesions are virulent characteristics expressed Georgetown University
by the chromosomally encoded eaeA gene cluster of enteropatho- Washington, DC 20007
genic Escherichia coli (EPEC). None of the bacteria recovered in
these studies carried the eaeA gene (2, 3). References
We report the isolation of an unusual EPEC-like strain in an 1. Bartlett JG, Belitsos PC, Sears CL. AIDS enteropathy. Clin Infect Dis.
HIV-infected man with a 10-month history of chronic, intermit- 1992;15:726-35.
tent diarrhea and a weight loss of 9 kg. When diarrhea began, 2. Kotler DP, Orenstein JM. Chronic diarrhea and malabsorption associ-
the patient's CD4 count was less than 100 cells/mm3. He had ated with enteropathogenic bacterial infection in a patient with AIDS.
been receiving a combination of zidovudine and didanosine and Ann Intern Med. 1993;119:127-8.
treatment for presumptive disseminated Mycobacterium avium 3. Kotler DP, Giang TT, Thiim M, Nataro JP, Sordillo EM, Orenstein
complex disease and antipneumocystosis prophylaxis. Examina- JM. Chronic bacterial enteropathy in patient with AIDS. J Infect Dis.
tion of an ileal biopsy specimen showed intracellular Enterocyto- 4. Law D. Adhesion and its role in the virulence of enteropathogenic
zoon bieneusi sporoblasts and spores. Examination of a colonic Escherichia coli. Clin Microbiol Rev. 1994;7:152-73.
biopsy specimen showed multiple superficial V-shaped mucosal
ulcers containing aggregates of small bacillary organisms. Some
of these clearly produced AE lesions, as shown by electron
microscopic examination (Figure 1). A stool sample was guaiac
negative, and cultures yielded predominant non-lactose-ferment- Internal Medicine and Training Subspecialists
ing, indole-negative E. coli. The latter could not be typed and
was nonmotile. Strong evidence supports the pathogenic role of To the Editor: Dr. Weiner's editorial (1) on the article by
the isolate: 1) Molecular studies showed that the isolate was Langdon and colleagues (2) was thoughtful in its laudatory re-
positive for eae by polymerase chain reaction (PCR) or colony flection. However, I must take exception to his statement that
15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6 523
a related issue that also appears to have been side- body of work as shown by publication in peer-reviewed journals
stepped is a delineation of the exact role the Board and presentations at scientific meetings.
(and other internal medicine groups) should have in Persons may not fall neatly into categories and may change
the downsizing process. As the total number of train- fields during the course of their careers. The American Board of
ing slots decreases, exactly how should accreditation be Internal Medicine should evaluate an investigator involved in
used to ensure that only the best programs remain patient care in the same manner in which it evaluates other
within each subspecialty? clinicians. Although the Board is to be congratulated for encour-
aging research, evaluation of research competence requires dif-
I fear that Dr. Weiner has missed the point. The American ferent methods and expertise than are currently within the pur-
Board of Internal Medicine and other certification boards have a view of the Board.
single, critical function: to ensure that certified specialists and
subspecialists meet certain criteria of excellence in the areas of Stephen J. Seligman, MD
medical knowledge and skill (within the limitations imposed by a State University of New York Health Science Center
tough examination). They should never be called on to become Brooklyn, NY 11203
the instruments of "downsizing" teaching programs. The only
role they should play is an indirect one, by indicating the success Reference
of specific programs in generating board-certified graduates. To 1. Langdon LO, Toskes PP, Kimball HR, the American Board of Internal
do more would represent a dreadful conflict of interest and Medicine Task Force on Subspecialty Internal Medicine. Future roles
would be counterproductive to the mission of the boards. and training of internal medicine subspecialists. Ann Intern Med. 1996;
As Langdon and colleagues state, "the primary interest of the 124:686-91.
American Board of Internal Medicine is the quality of training of
internists; ways to finance graduate medical education are be- To the Editor: I read with interest the insightful position paper
yond the Board's purview." So it should remain. by Langdon and colleagues (1) on the future roles and training
It is painfully evident, however, that unless the U.S. Congress of internal medicine specialists. Certainly, strong subspecialist
has sufficient wisdom to create fire-walled funding, the enlight- basic scientists will always be needed alongside subspecialist clin-
ened program envisioned by the Task Force will never come to ical investigators and subspecialist clinicians. It is also clear that
fruition. This would jeopardize the entire structure of basic re- each requires more specific and differing emphasis on certain
search, clinical investigation, and teaching—the total contempo- aspects of training.
rary medical enterprise—in the foreseeable future. One could A problem can develop when the subspecialist basic scientist
expand this dream to suggest that necessary revenue could be or subspecialist clinical investigator is required to adopt a more
derived from insurers and managed care organizations; from a clinical role. This sequence of events seems to occur more com-
levy on medical licenses and relicensure; or from a surtax on monly than it has in the past; if the physician has had insufficient
cigarettes, alcohol, handguns, and ammunition. This revenue clinical experience during his or her training program, substantial
could ensure the survival and growth of this carefully and pains- gaps may exist. An example would be the physician who takes a
takingly wrought intellectual treasure. 3-year fellowship program and completes only 1 clinical year. If,
I have faith that many capable and visionary legislators will at a later date, that physician is required to be more clinically
appreciate the enormous stakes involved. I believe that after all active, a strong basic infrastructure may be lacking.
the tumult and shouting die, they will do the right thing. As subspecialty care becomes more complex, intensive, and (at
times) highly technical, a well-grounded fundamental period
Robert M. Moser, MD spent treating patients is essential. If only 1 of 3 years of a
Canyon Consulting Corp. training program has been spent in the care of patients, a phy-
Chama, NM 87920 sician changing or being forced to change the direction of his or
her career may find that clinical training inadequate.
Eugene M. Bozymski, MD
1. Weiner JP. Internal medicine at the crossroads: training subspecialists
for the next century [Editorial]. Ann Intern Med. 1996;124:681-2. University of North Carolina
2. Langdon LO, Toskes PP, Kimball HR, the American Board of Internal Chapel Hill, NC 27599
Medicine Task Force on Subspecialty Internal Medicine. Future roles
and training of internal medicine subspecialists. Ann Intern Med. 1996; Reference
124:686-91. 1. Langdon LO, Toskes PP, Kimball HR, the American Board of Internal
Medicine Task Force on Subspecialty Internal Medicine. Future roles
and training of internal medicine subspecialists. Ann Intern Med. 1996;
Future Roles of Subspecialists In response: We agree with Dr. Seligman that subspecialty
investigators must have skills that differ from those of their
To the Editor: Langdon and colleagues (1) have assessed the clinician colleagues. For this reason, the Board has recom-
future roles and training of internal medicine subspecialists. Of mended that subspecialty training encompass different training
particular interest to those of us in academic medicine who are pathways focused on these two career goals. The investigator
interested in basic laboratory investigations is that the authors pathway, which is longer than the clinical pathway, must allow
advocate the formation of a category of "subspecialist investiga- individual investigators to develop their own unique expertise.
tor" with a subdivision into "basic scientist" and "clinical inves- The length of time required for this varies—a lifetime is some-
tigator." I hope that their ideas will stimulate support for the times not long enough. However, the Board believes that 3 years
nurturing of physicians in research. of research training comprise the minimum needed to launch a
The authors also propose periods of training and a suggested career as an investigator. In addition, program directors must be
curriculum; these are a source of concern. Persons who are allowed flexibility in arranging appropriate training experiences
effective in research have skills that differ from those of effective to benefit each investigator, and the Board has no intention of
teachers and clinicians. These skills include the ability to gener- meddling in these arrangements. Each research training experi-
ate research ideas, to access sufficient research resources, to ence, however, should include a mentor, specific performance
obtain satisfaction from long-term results that is sufficient to goals, and regular feedback to the trainee. To the degree that a
continue frustrating tasks, and to persist in publishing and dis- given knowledge base can be common to investigators (that is,
seminating their results. These skills are more closely associated research methodology, research ethics, biostatistics, or clinical
with attributes desirable in a PhD candidate than with those epidemiology), appropriate course work should be efficiently or-
associated with patient care or teaching. The process is neither ganized and included. We believe that this knowledge is testable.
time limited nor associated with a testable knowledge base. Cri- The Board shares Dr. Bozymski's concern that the clinical
teria for success are achieved goals, including completion of a training of investigators who change careers may be less than
524 15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6
optimal. We foresee that subspecialty clinicians will probably population. JAMA. 1956;1617-24.
need a broader range of clinical skills and will be called on to act 3. Langdon LO, Toskes PP, Kimball HR, the American Board of Internal
as principal care providers. For this reason, the investigator Medicine Task Force on Subspecialty Internal Medicine. Future roles
pathway requires at least 12 to 24 months of clinical training, and training of internal medicine subspecialists. Ann Intern Med. 1996;
appropriate for each subspecialty. Nevertheless, the Board would 124:686-91.
agree that a subspecialist who changes his or her career direction
from investigator to clinician is likely to require additional clin- To the Editor: Dr. Saultz (1) makes realistic comparisons and
ical knowledge and experience. If such a career change occurs contrasts between family medicine and internal medicine in terms
during training, the Board will require that additional clinical of educational experience and practice style and philosophy.
training be obtained before admission to the subspecialty certi- Concerns about practical aspects of cross-coverage, redistribution
fication examination. If a career change occurs after training, of patient demographic characteristics, and internists caring for
appropriate continuing medical education becomes the profes- children within these practices are relevant and remotely feasible.
sional responsibility of the individual physician. The prospect of involving pediatricians in this collaboration, how-
ever, is less than likely. Pediatrics has never had a close rela-
Lynn O. Langdon, MS tionship with family medicine, and it diverged on its own primary
Harry R. Kimball, MD care path 15 years before the American Academy of General
American Board of Internal Medicine Practice was formed in 1947.
Philadelphia, PA 19104 In 1967, the boards for internal medicine and pediatrics rec-
ognized the collaborative training of medicine-pediatrics physi-
Phillip P. Toskes, MD cians. As managed care evolved from the changing medical en-
University of Florida vironment, the scope of care among primary care specialists
Gainesville, FL 32610 narrowed (2). This evolution contributed to a 20-fold growth in
the number of medicine-pediatrics programs from 1980 to 1990.
Specialty acceptance is evidenced by the 1400 graduates in pri-
vate practice, the 1000 residents in training, and national repre-
Internal Medicine and Family Medicine sentation through membership in the Primary Care Organiza-
This unique primary care specialist also cares for families and
To the Editor: In his article on the differences between internal provides the ideal collaborative partner with family medicine by
and family medicine practitioners (1), Dr. Saultz did not mention overcoming the concerns acknowledged by Dr. Saultz. Managed
a potential way to facilitate collaboration between the two disci- care utilization data and coding demographic characteristics for
plines. The American Board of Internal Medicine and the Amer- family medicine/medicine-pediatrics collaborative practices were
ican Board of Family Medicine co-sponsor an examination that recently presented (3). Collaborative practices showed a 19% to
leads to a Certificate of Added Qualification in Geriatric Medi- 64% reduction in utilization variables compared with community-
cine. Because residency review committees for both disciplines based practices; family physicians made 40% of internal referrals
require each discipline to provide geriatric medical education, to collaborative medicine-pediatrics physicians that would have
internal medicine and family medicine faculty can join in house- otherwise been made to subspecialists. In effect, the competitive
staff training. Exposure to the multidisciplinary model necessary system of generalist physicians in separate practices results in
to manage the increasing number of functionally impaired elderly
family physicians feeding the imbalance between subspecialists
is very important for trainees in both specialties. This approach
can serve as an academic model of collaboration.
These results show that collaborative practice partnerships are
efficient. Continued discussion between family medicine and in-
Clifford F. Feiner, DO ternal medicine would benefit from input from medicine-pediat-
Mount Sinai Services at Queens Hospital Center rics. Thirty years of medicine-pediatrics experience offers an
Jamaica, NY 11432 effective collaborative perspective in primary care. This is partic-
ularly relevant if the gap in pediatric training within the family
Reference medicine-internal medicine collaboration is to be reduced.
1. Saultz JW. Reflections on internal medicine and family medicine. Ann
Intern Med. 1996;124:600-3. Gary M. Onady, MD, PhD
Wright State University
To the Editor: In his recent paper (1), Dr. Saultz suggested that Dayton, OH 43409
rotations in family medicine might be established for residents in
internal medicine. In fact, that suggestion was anticipated more References
than 45 years ago by Dr. Ellsworth L. Amidon, who at that time 1. Saultz JW. Reflections on internal medicine and family medicine. Ann
was Chairman of the Department of Internal Medicine at the Intern Med. 1996;124:600-3.
University of Vermont College of Medicine. Long before family 2. The future of family practice implications of the changing environment
medicine was recognized as an academic discipline, first-year of medicine. Council on Long Range Planning and Development in
residents in internal medicine were assigned to serve as "city Cooperation with the American Academy of Family Practice. JAMA.
physicians" in Burlington, Vermont, for 6 months. In this capac- 3. Onady GM. Community based Med/Peds—family medicine collabora-
ity, the residents were responsible for the immediate health tive practice experiences [Abstract]. In: Proceedings of the Rush Pri-
needs of patients of all ages, and care was provided without any mary Care Institute. The Challenge of Collaborative Primary Care
cost to patients. Documentation of 1 year of this experience, Practice: Making It Work. 20 October 1995.
done for epidemiologic and socioeconomic reasons, has been
published (2). Although those residents pursued careers in aca- To the Editor: Dr. Saultz (1) raised an important topic in his
demic medicine and general and subspecialty internal medicine, article on the relationship between internal medicine and family
recent changes in health care have dictated a more structured medicine. Although internists always teach family medicine resi-
approach to the training of residents in subspecialty internal dents, it is unusual for family physicians to teach internal med-
medicine (3). icine residents.
General internal medicine might learn from family medicine
Richard B. Davis, MD, PhD experience without duplicate resources. For example, primary
Williamsburg, VA 23188 care and psychosocial skills are important components of the
curriculum of general internal medicine residencies, and these
References skills are essential components of the family medicine curriculum
1. Saultz JW. Reflections on internal medicine and family medicine. Ann (2).
Intern Med. 1996;124:600-3. Primary care skills in the internist are essential for effective
2. Davis RB, Kresge DJ. Domiciliary medical practice in an indigent ambulatory care (for example, management of common diseases
15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6 525
and health maintenance) (3). Similarly, family medicine empha- Dr. Onady raises an important issue on the role of physicians
sizes primary care skills for managing the needs of ambulatory who are trained in both pediatrics and internal medicine. His
patients. These skills include an epidemiologic approach (such as letter seems to suggest a role for the physician trained in medi-
prevalent diseases and problems), prevention (such as screening cine-pediatrics as an on-site consultant in primary care practice.
methods), patient education (such as self-care), office adminis- Theoretically, this might improve efficiency in the use of medical
tration, and team work and resources use (such as family support specialists, but current practice patterns indicate that referrals
and community resources). from family physicians to other generalists are rare. On a na-
Psychosocial skills are essential for an adequate general inter- tional basis, there are too few providers trained in medicine-
nist-patient relationship (4). Similarly, family medicine empha- pediatrics to fill this role on more than a limited scale. Further-
sizes the development of psychosocial skills and the improvement more, our health care system currently has an oversupply of
of the physician-patient relationship. For example, by using the consulting specialists. In a highly penetrated managed care mar-
biopsychosocial model or systems approach, the patient is not a ket such as my own community, specialty consultation is increas-
disease but a person, and health is influenced by many nonmed- ingly being provided in a cost-effective and high-quality manner
ical factors (such as context and lifestyle) and biochemical inter- by specialists who come to the primary care setting to see pa-
actions. In addition, understanding the patient as a person and tients when needed. It seems unlikely that physicians trained in
developing psychosocial skills help to improve an internist's hu- medicine-pediatrics will function as consultants to family physi-
manistic qualities (5). cians beyond a limited basis. Dr. Onady's comment that the
Finally, both disciplines might work together in education and physician trained in medicine-pediatrics might be an ideal col-
practice with similar goals, enrich each other, and improve the laborator with family physicians as primary care providers is well
care of their patients. taken. Unfortunately, collaboration between medicine-pediatric
training programs and family medicine programs has been even
Carlos A. Reyes-Ortiz, MD more rare than collaboration between internal medicine and
McGuire Veterans Affairs Medical Center family practice residencies. It is hoped that this can be changed
Richmond, VA 23249 in the future.
References John Saultz, MD
1. Saultz JW. Reflections on internal medicine and family medicine. Ann Oregon Health Sciences University
Intern Med. 1996;124:600-3. Portland, OR 97201
2. Graduate Medical Education Directory 1994-1995. Chicago: American
Medical Association; 1994.
3. Wartman SA, O'Sullivan PS, Cyr MG. Ambulatory-based residency
education: improving the congruence of teaching, learning, and patient
care. Ann Intern Med. 1992;116:1071-5. Correction: Duplicate Publication
4. Robbins AS, Cope DW, Campbell L, Vivell S. Expert ratings of primary
care goals and objectives. J Gen Intern Med. 1995;10:429-35. In the 15 March 1996 issue of Annals, we published a letter
5. Evaluation of humanistic qualities in the internist. American Board of submitted by Dr. Charles L. Loprinzi of Mayo Clinic commenting
Internal Medicine. Ann Intern Med. 1983;99:720-4. on an article that we had previously published on the induction
of adrenal insufficiency by megestrol acetate in patients with
In response: The responses to my paper on internal medicine AIDS (1). Earlier, we had erroneously sent Dr. Loprinzi notifi-
and family medicine underscore the substantial national interest cation of rejection of this letter, and he submitted a version of it
in the relationship between the two disciplines. The above letters to the Journal of Clinical Oncology. This version was accepted by
reflect the content of other correspondence I have received. that journal and subsequently published. We wish to apologize to
Dr. Feiner points out the history of collaboration between the Dr. Loprinzi for our error and to notify our readership of our
two disciplines in geriatrics training. If the basis of this collabo- responsibility for the appearance of this letter in two biomedical
ration could be extended into primary care training, much benefit journals.
might be realized. I am gratified to hear of Dr. Davis' report of
a successful experience at the University of Vermont many years Reference
ago. Similar stories exist on a local basis, but these models have 1. Loprinzi CL, Fonseca R, Jensen MD. Induction of adrenal suppression
not been replicated in a national model. by megestrol acetate [Letter]. Ann Intern Med. 1996;124:613.
526 15 September 1996 • Annals of Internal Medicine • Volume 125 • Number 6