Letters

Letters

The Editors welcome submissions for possible publication in the Letters section. Authors of letters should: • Include no more than 400 words of text, three authors, and five references • Type with double-spacing • Send with the letter a transfer-of-copyright form (see Table of Contents for location) signed by all authors • Provide a self-addressed envelope if they want to be notified that the letter was received Letters commenting on an Annals article will be considered if they are received within 6 weeks of the time the article was published. Only some of the letters received can be published. Published letters are edited and may be shortened; tables and figures are included only selectively. Authors will be notified only if their letter is accepted. Unpublished letters cannot be returned.

The Use of Placebos in Controlled Trials To the Editors: In a discussion of the rational use of the placebo, Bourne (1) has observed that "In some cases no 'placebo group' may be necessary at all. If one mode of therapy has definitely been shown to have genuine benefit, that therapy, not placebo, should be the 'control.' For example, in testing new drugs for symptomatic relief in rheumatoid arthritis, a demonstrated superiority to placebo medication is worse than useless, since (1) the placebo group has been deprived of active medication (aspirin) and (2) the real question, 'Is new drug X superior to conventional therapy?' has not been answered." These observations are directly applicable to the clinical trial reported by Sninsky and colleagues (2), in which patients with ulcerative colitis were treated with either an experimental new drug, mesalamine, or an inactive placebo. The outcome was consistent with Bourne's analysis. Patients in the placebo group were deprived of active medication and 22 (50%) experienced worsening of their disease during the study period. Also, the question, "Is mesalamine superior to conventional therapy with sulfasalazine?" was not answered. I find it disturbing that the investigators did not surmise during the planning phase of their study that some patients with ulcerative colitis would probably worsen while receiving inactive placebo. I am surprised that neither the collaborating physicians nor their institutional review boards insisted that the control group be treated with sulfasalazine rather than with placebo. Such a study could have shown whether mesalamine is superior to conventional therapy, and, more important, the suffering experienced by control patients whose disease worsened during the study could have been reduced.

In response: We appreciate the important issue raised by Dr. Phillips regarding the rational use of placebo-controlled trials. We agree that in certain clinical situations the incorporation of a placebo arm may not be necessary. However, as noted in a position paper of the American College of Gastroenterology entitled, "Placebo-Controlled Clinical Trials in Gastroenterology" (1), at least two principal circumstances exist in which inclusion of a placebo group in randomized clinical trial should present no clinical problem. The first occurs when the commonly used medication for a specific disease is of questionable or low efficacy. Table 2 of our paper (2), shows that more than 90% of the patients had been treated with sulfasalazine. Therefore, most of our patients were either allergic or interolant to sulfasalazine, or they and their physicians were frustrated enough with their inadequate response to the drug that they consented to inclusion in the study, knowing that 30% of the study patients would receive a placebo. The second situation occurs when the commonly used therapy (sulfasalazine in this case) has a high frequency of unacceptable side effects. As we pointed out in our introduction, as many as 30% of patients taking sulfasalazine experience side effects that require withdrawal of the drug or a substantial reduction in the dose of suboptimal levels. The position paper states that "a trial of a new therapy for ulcerative colitis must be placebo-controlled among the large population of patients allergic to, or intolerant of, sulfasalazine." Further, we minimized the adverse effects of placebo by continuously monitoring the patients and promptly removing from the trial any whose clinical condition was deemed unacceptable by the patient or the physician. Also of note, 23% of the patients in the placebo group were in remission or had improved by the end of the study. It is now reasonable to consider a trial comparing Asacol with sulfasalazine in patients with newly diagnosed ulcerative colitis to determine whether this new form of therapy has comparable efficacy or whether the drug should be used primarily in patients with refractory ulcerative colitis or in those intolerant to the side effects of sulfasalazine. In summary, the various institutional review boards involved in our trial and the patients giving informed consent took all these factors into account and provided the medical community with an ethical and scientifically sound trial evaluating this new oral form of mesalamine therapy for mild to moderately active ulcerative colitis. Charles A. Sninsky, MD University of Florida College of Medicine and the Gainesville Veterans Affairs Medical Center Gainesville, FL 32610

Michael Phillips, MD St. Vincent's Medical Center New York Medical College Staten Island, NY 10310

Fergus Shanahan, MD Associate Professor of Medicine UCLA Los Angeles, CA

References 1. Bourne HR. Rational use of placebo. In: Clinical Pharmacology; 2d edition. Melmon KL, Morrelli HF; eds. New York: Macmillan; 1972: 1052-62. 2, Sninsky CA, Cort DH, Shanahan F, Powers BJ, Sessions JT, Pruitt RE, et al. Oral mesalamine (Asacol) for mildly to moderately active ulcerative colitis. Ann Intern Med. 1991;115:350-5.

References 1. Placebo-controlled clinical trials in gastroenterology—a position paper of the American College of Gastroenterology. Am J Gastroenterol. 1984;79(12):913-7. 2. Sninsky CA, Cort DH, Shanahan F, Powers BJ, Sessions JT, Pruitt RE, et al. Oral mesalamine (Asacol) for mildly to moderately active ulcerative colitis. Ann Intern Med. 1991;115:350-5.

1 F e b r u a r y 1992 • Annals

of Internal

Downloaded From: http://annals.org/ by a National Institutes of Health User on 06/12/2013

Medicine

• V o l u m e 116 • N u m b e r 3

265

Patient-centered and Physician-centered Approaches to Interviewing To the Editors: In their recent article (1), Smith and Hoppe emphasize the patient-centered interview, yet their identification of a specific physician-centered approach is also important. They describe the latter primarily in terms of its datagathering function, but physicians are not always objective data gatherers. Just as patients have attitudes, beliefs, and emotions that influence information flow, so do physicians. Others have recently commented on the importance of physician attitudes (2) and personality (3) in patient care. It may be important for physicians to discuss these issues with patients or with colleagues; and, in some cases, the physician may need only to be aware that such issues exist. The separation of the two aspects of the interview into distinct segments seems inefficient, because events discussed from the "personal" standpoint need to be rediscussed later from the "organic disease" standpoint. The division perpetuates the mind-body dissociation of the biomedical model. The integrated approach that the authors seek might be accomplished more easily through a discussion of personal and objective aspects of a symptom or event simultaneously. Indeed, the richest personal data often emerge as patients discuss details of an event. Conversely, uncomfortable feelings may be suppressed when reference to events is made in general terms during the first part of the interview. Finally, patients may be easily confused by the artificial separation. To return the patient to the interviewer's structure when he digresses from the physician's path seems an unnecessary violation of the belief that the patient should control the direction of the conversation. Barry Egener, MD Good Samaritan Clinic at St. Johns Portland, OR 97203

In response: Dr. Egener correctly notes that we restricted our description of the physician-centered aspect to data gathering, a factor which we believe reflects the physician's conscious agenda. Physicians' attitudes, beliefs, and emotions are primarily unconscious, unrecognized determinants of the doctor-patient interaction that, unfortunately, were beyond the scope of our paper. The question is nonetheless germane. Although we recommend an integrated approach, we did not address the unrecognized personal responses that almost inevitably arise in the physician or student (1). For example, the doctor may have an unrecognized need to appear pleasing and may therefore avoid unpleasant information that the patient wants to address. Other unrecognized needs (for example, a need to control or the fear of psychosocial data or emotions) can have equally deleterious effects on physician-patient interaction and thus impair communication (1, 2). Also, unrecognized reactions are typically magnified when using a patientcentered approach because of the increased psychosocial data that it generates. Therefore, in addition to instruction in a patient-centered approach, physician awareness of their own unrecognized responses also is necessary (3). Ideally, there would be no need to separate the two aspects of the interview, because physicians would equally value and balance patient-centeredness and physician-centeredness. In actuality, however, we have found such a separation to be a useful pedagogic tool. It appropriately places the patient-centered aspect on an equal footing with the more familiar physician-centered approach. One benefit is the ability to integrate organic-disease data with the unique psychosocial contextual material. Because most biomedical data typically surface dur1 February 1992 • Annals of Internal Medicine

Robert C. Smith, MD Ruth B. Hoppe, MD Department of Medicine Michigan State University College of Human Medicine East Lansing, MI 48824 References 1. Smith RC. Teaching interviewing skills to medical students: the issue of countertransference. J Med Educ. 1984;59:582-8. 2. Smith RC, Stein HF. A topographical model of clinical decisionmaking: a heuristic for family medicine teaching. Fam Med. 1987; 19: 361-3. 3. Smith RC. Use and management of physicians' feelings during the interview. In: Lipkin M, Putnam SM, Lazare A; eds. The Medical Interview—A Textbook on Medical Interviewing. New York: Springer-Verlag; 1991 [In press!. Hyperamylasemia and Hematologic Malignancies

References 1. Smith RC, Hoppe RB. The patient's story: integrating the patientand physician-centered approaches to interviewing. Ann Intern Med. 1991;115:470-7. 2. Barsky AJ, Wyshak G, Latham KS, Klerman GL. Hypochondriacal patients, their physicians, and their medical care. J Gen Intern Med. 1991;6:413-9. 3. Nightengale SD, Yarnold PR, Greenberg MS. Sympathy, empathy, and physician resource utilization. J Gen Intern Med. 1991;6:420-3.

266

ing the patient-centered aspect, there is no need to elicit them later. Only additional biomedical details need to be added. Psychosocial or biomedical material need not be repeated. When the transition is used, there is no confusion to patients who like the chance to express themselves and to be in control. We recommend that interviewers actively and open-endedly encourage psychosocial material from a patient who is giving much organic-disease data (for example, "How did that make you feel?"). This approach gives the patient the chance to address often suppressed and ignored psychosocial information and place it on equal footing with the biomedical data. As a general rule, we suggest a few attempts at such redirections if the patient persists in an organic-disease focus. This approach allows the patient to talk about psychosocial or biomedical issues, but forces neither, thus emphasizing the essence of patient-centeredness: the patient chooses where to go.

To the Editors: Hyperamylasemia has been described in association with numerous solid malignant tumors (1, 2), but its association with hematologic malignancies has not been extensively studied. In a preliminary study, we found that 32% of patients with hematologic malignancies had hyperamylasemia (2). To gain more information regarding the incidence and significance of hyperamylasemia in patients with hematologic malignancies, we studied serum amylase activity in 35 patients and followed them over time (3, 4). Forty-one consecutive patients with various hematologic malignancies were evaluated in our hematology clinic. Six patients were excluded because of concurrent illness or misdiagnosis of a hematologic malignancy. Eighteen healthy volunteers who were similarly evaluated comprised the control group. Nine patients were retested approximately after one year. The 35 study patients included 19 women and 16 men aged between 25 and 83 years. Eleven patients had chronic lymphocytic leukemia and 9 had lymphoma. The remaining patients had various other malignancies, including multiple my-

Figure 1. Survival among patients with hyperamylasemia and those with normal amylase levels. All patients had hematologic malignancies.

• Volume 116 • Number 3


eloma, polycythemia vera, chronic myelocytic leukemia, acute myelocytic leukemia, myeloid metaplasia, acute lymphocytic leukemia, and hairy-cell leukemia. A total of 10 patients (29%) had increased amylase levels (range, 311 to 569 U/L), including 4 of the 11 patients with chronic lymphocytic leukemia. Seven patients had predominantly salivary-type isoamylase, two had predominantly pancreatic-type isoamylase, and one had macroamylasemia. Hyperamylasemia was not related to the duration of hematologic malignancy, age, sex, or type of therapy. One member of the control group had hyperamylasemia. (mean, 216 ± 42 U/L; range, 161 to 319 U/L). The 1-year mortality for patients with hyperamylasemia was 60% (6 of 10 patients) and 25% (3 of 12 patients) for patients with normal amylase levels (P < 0.05). Five of the 6 patients who died had salivary-type hyperamylasemia. This increased 1-year mortality is represented in a Kaplan-Meier curve (Figure 1). Hyperamylasemia associated with solid tumors is thought to result from secretion by malignant cells. The hyperamylasemia seen in patients with hematologic malignancies may also be related to amylase production by abnormal cells. The high incidence of hyperamylasemia in patients with hematologic malignancies should be recognized in order to avoid the misdiagnosis of pancreatitis based on routine laboratory tests. When a diagnosis of pancreatitis is suggested in a patient with a hematologic malignancy, other tests, including serum lipase determinations and abdominal imaging techniques, should be used to confirm the diagnosis. In addition, the presence of salivary-type hyperamylasemia may be a poor prognostic indicator and should alert the physician to seek other factors suggestive of increased disease activity. Leonard Stein, MD The Norwalk Medical Group Norwalk, CT 06851 Simmy Bank, MD Kanti Rai, MD Long Island Jewish Medical Center New Hyde Park, NY References 1. Berk JE, Shimamura J, Fridhandler L. Tumor associated hyperamylasemia. Am J Gastroenterol. 1977;68:572-7. 2. Buch P, Bank S, Anfang C, Lendvai S, Kranz V. The incidence of hyperamylasemia in malignant disease. Gastroenterology. 1981 ;80: A1118. 3. Stein L, Greenberg R, Bank S, Lendvai S, Owen T, Rubin J. The incidence of hyperamylasemia in hematological malignancies. Gastroenterology. 1988;94:A443. 4. Stein LB, Bank S, Dabezies MA, Jagathambal K, Rai K. "S-type" hyperamylasemia: A poor prognostic indicator in patients with hematologic malignancies? Gastroenterology. 1991; 100:A300. Methicillin-resistant Staphylococcus Homes

aureus in Nursing

To the Editors: In a well-designed study of methicillin-resistant Staphylococcus aureus (MRSA) in a long-term Veterans Affairs facility, Bradley and colleagues (1) convincingly showed that patient-to-patient transmission occurred at a relatively slow rate, despite the absence of routine cohorting and strict isolation procedures. These results are reassuring to those responsible for infection control in such facilities, and provide further evidence that patients with MRSA should not be denied hospital admission or necessary rehabilitative services. The authors note relatively few MRSA infections in their patient sample (9 infections among 341 patients). They state that their findings differed from those of our study (2), in which we reported that persistent carriage of MRSA was a significant risk factor for the development of staphylococcal infection. Calculation of the rate of MRSA infection in the two study samples shows that the supposed difference in results is more apparent than real. Using the data given by Bradley and coworkers, we calculated the crude rate of MRSA infection in their study sample to be 0.24 per thousand patient days. The

infection rate in our study was 0.27 per thousand patient days for first episodes of MRSA infection; the rate is 0.31 per thousand if multiple infections in the same patient are counted. The rates of MRSA infection thus appear similar in the two facilities. Significant differences in intent and design exist between the two studies. The objective of Bradley and colleagues was to identify the frequency of patient-to-patient transmission of MRSA, whereas our objective was to determine the consequences of MRSA colonization. Our study duration (3 years) and the duration of longitudinal follow-up of patients (mean, 243 days) was considerably longer than that of Bradley and coworkers; this prolonged observation was important because infections occurred among colonized patients at a relatively constant rate of 15% for each 100 days of carriage. If we had used a shorter follow-up period, as did Bradley (mean, 3.6 months), we would have been unable to detect a significant risk for infection associated with MRSA colonization. We believe that when the two studies are considered together, they are not contradictory, but complementary. Both studies conclude that MRSA carriers remain colonized for long periods and that carriage is more likely in more severely debilitated patients. Bradley and colleagues showed that longterm care patients colonized with MRSA are infrequently the source of MRSA acquisition by other patients. We have shown that patients colonized with MRSA are at high risk for infection if colonization persists. Robert R. Muder, MD Marilyn M. Wagener, MPH Victor L. Yu, MD Veterans Affairs Medical Center and University of Pittsburgh School of Medicine Pittsburgh, PA 15240 References 1. Bradley SF, Terpenning MS, Ramsey MA, Zarins LT, Jorgensen KA, Sottile WS, et al. Methicillin-resistant Staphylococcus aureus: colonization and infection in a long-term care facility. Ann Intern Med. 1991;115:417-22. 2. Muder RR, Brennen C, Wagener MM, Vickers RM, Rihs JD, Hancock GA, et al. Methicillin-resistant staphylococcal colonization and infection in a long-term care facility. Ann Intern Med. 1991; 114:10712. To the Editors: We were impressed by the recent article by Bradley and associates (1) on methicillin-resistant Staphylococcus aureus (MRSA) colonization and infection. It is an excellent contribution to the discussion on this important topic. Their year-long observation yielded results similar to our own single-point prevalence study at the Mercy nursing facility in Chicago. In Chicago, Veterans Affairs hospitals were first affected in the early 1980s by increasing numbers of MRSA isolates. A report by Hsu and colleagues (2) from a small teaching hospital showed a markedly higher proportion of MRSA isolates (from infection or colonization) among nursing home residents than among patients from the community. Beginning in 1986 and 1987, most hospitals began to experience a rapid rise in the isolates of MRSA with reported ranges between 18% and 80% (Chicago Department of Health survey). In those institutions where this information could be obtained, one third to one half of the isolates appeared to have been obtained from patients admitted from nursing homes. An investigation of our affiliated nursing facility showed the overall Staphylococcus aureus carriage rate on a single nasal swab to be 51%. Isolates of MRSA were detected in 14% of patients and were associated almost exclusively with nonambulatory status, poor nutritional state, use of a urinary catheter, recent admission to the facility, and hospitalization within the past 6 months during which the patient received antibiotics. These results are similar to those reported by Bradley and associates (1). The colonization rate in our facility was similar to that reported by Storch and colleagues (3) and less than that reported by Bradley and associates (1), although the methods are not truly comparable because we used only a single swab on a

1 February 1992 • Annals of Internal Medicine Downloaded From: http://annals.org/ by a National Institutes of Health User on 06/12/2013

• Volume 116 • Number 3

267

single site. The points of similarity are the relatively low probable transmission within the facility with most carriers having acquired MRS A during a short-term hospital stay, and the correlation of carrier status with functional status. To prevent the spread of MRSA in the acute care setting, not all nursing home residents need to be screened for MRSA. Those with poor functional status before hospital admission within the past year would be the most likely carriers. Charles E. Cherubin, MD Salah Antar, MD Joanne Jurkovic, MS Mercy Hospital and Medical Center The Infectious Disease Research Institute 5000 East End Avenue Chicago, IL 60615 References 1. Bradley SF, Terpenning MS, Ramsey MA, Zarins LT, Jorgensen KA, Sottile WS, et al. Methicillin-resistant Staphylococcus aureus: colonization and infection in a long-term care facility. Ann Intern Med. 1991;115:417-22. 2. Hsu CS, Macaluso CP, Special L, Hubble RH. High rate of methicillin resistance of Staphylococcus aureus isolated from hospitalized nursing home patients. Arch Intern Med. 1988;148:569-70. 3. Storch GA, Radcliff JL, Meyer PL, Hinrichs JH. Methicillin-resistant Staphylococcus aureus in a nursing home. Infect Control. 1987;8: 24-9.

Postmenopausal Estrogen and Prevention Bias To the Editors: When citing the improved prognosis for breast cancer among estrogen-treated women, Barrett-Connor (1) offers the plausible explanation that their cancers were diagnosed earlier, presumably due to increased surveillance. I offer another plausible explanation: The increased incidence of breast cancer among estrogen-treated women is falsely inflated by overdiagnosis. The increased rate of breast cancer (and absence of an increased death rate) among these women may be caused by the more frequent diagnosis of noninvasive lesions, 15% of which would not have progressed to malignant disease if left untreated (2). Women who use estrogen have a higher rate of mammography screening (3). The National Cancer Institute's (NCI) latest report on the nation's 15-year survival statistics shows a correlation between increasing acceptance of mammography screening and the rising incidence of breast cancer (4). One reason to suspect overdiagnosis is the fact that the nation's breast cancer death rate has dropped only slightly. The NCI's statistics apply only to invasive cancer and thus would not include cases of ductal carcinoma in situ (Edwards B, Personal communication). A precedent exists, however, for the inaccurate identification of carcinoma in situ as invasive cancer. According to Ketcham and Moffat (5), "These diagnostic errors have been documented by the Surveillance Epidemiology and End Results program, the National Breast Cancer Survey of the American College of Surgeons, and the Breast Cancer Detection Demonstration projects." Maryann Napoli Associate Director Center for Medical Consumers New York, NY 10012 References 1. Barrett-Connor E. Postmenopausal estrogen and prevention bias. Ann Intern Med. 1991;115:455-6. 2. Physicians Data Query. Database of the National Cancer Institute. Bethesda, Maryland; 1991. 3. Dupont WD. Estrogen replacement therapy and the risk of breast cancer. JAMA. 1991 ;265:1824. 4. Ries LA, Hankey BF, Miller BA, Hartman AM, Edwards BK. Cancer statistics review: 1973-1988. National Cancer Institute, NIH Pub. no. 91-2789, 1991. 5. Ketcham AS, Moffat FL. Vexed surgeons, perplexed patients, and breast cancers which may not be cancer. Cancer. 1990;65:378-83.

268

Optimal Timing of Initial Breast Cancer Surgery To the Editors: In his editorial (1), Dr. McGuire concludes that he cannot be sure of any apparent benefit from timing of initial breast cancer surgery in relation to the patient's menstrual cycle because all published studies dealing with the question have been retrospective. He suggests that a properly designed prospective study be done to resolve this question. We suggested the same in our original 1989 article (2), and subsequently. Dr. McGuire's witty argument, about the inevitability of statistically significant results due only to chance, is misleading. He neglects to mention that the hypothesis that surgical timing within the fertility cycle affects outcome originated from a prospectively randomized murine study (5). The positive correlation that was observed and reported was reproduced several times, and mechanistic studies addressing fertility-cycle modulation of cellular immune function also preceded the initial retrospective clinical review (6). We are puzzled by Dr. McGuire's omission of this information from the editorial. We are also puzzled by his noting a lack of any clear consensus in the published studies. The three published series that reported a positive correlation between such timing and prognosis (2, 7, 8), representing a total of 573 patients, all agreed that primary surgery done on days 18-20 from the start of a patient's menstrual cycle is associated with a favorable prognosis. Several workers have already begun to describe observations that may lead to mechanistic explanations for this observed effect. Wicha (9) reported marked menstrual cycling of serum concentrations of mammastatin, a negative growth modulator of malignant breast epithelium; Pollack (10) noted that insulin-like growth factor I (IGF-I), a potent mitogen for breast cancer cells, may be induced by the unopposed estrogen secreted during the first half of the menstrual cycle; and increasing concentrations of progesterone, noted after ovulation, have been postulated to inhibit the growth-promoting effect of unopposed estrogen (11). The possibility that the prognosis-enhancing effect of initial breast cancer surgery between days 18 to 20 of the menstrual cycle is fortuitous must be taken seriously. Yet, the observation may be real. Because this association would represent the only prognostic variable in breast cancer management over which the clinician has a priori control, we must be careful not to dismiss prematurely a potentially valuable and simple therapeutic tool. Avium Z. Blunting, MD William J. M. Hrushesky, MD University of Southern California Encino, CA 91436 References 1. McGuire WL. The optimal timing of mastectomy: low tide or high tide? Ann Intern Med. 1991;115:401-3. 2. Hrushesky WJ, Bluming AZ, Gruber SA, Sothern RB. Menstrual influence on surgical cure of breast cancer. Lancet. 1989;2:949-52. 3. Hrushesky WJ, Bluming AZ, Gruber SA. Menstrual status and breast cancer surgery—reply. Breast Cancer Res Treat. 1990;16:121. 4. Bluming AZ, Hrushesky WJ. The effect of surgical timing within the fertility cycle on breast cancer outcome. Ann NY Acad Sci. 1991; 618:277-91. 5. Ratajczak HV, Sothern RB, Hrushesky WJ. Estrous stage influences surgical cure of a mouse breast cancer. J Exp Med. 1988;168:88-96. Additional references (6-11) are available from the authors on request.

To the Editors: In his recent editorial (1), Dr. McGuire suggested that chance alone "could well account for the published inconsistencies regarding the timing of mastectomy." After dividing the time since last menstrual period into 14 day intervals, he concluded that, by such methods, there was a high probability of finding a "chance" difference. We challenge Dr. McGuire on several points. First, no random analyses have been done; instead, three specific hypotheses were tested by different groups. Hrushesky and colleagues (2), after experiments conducted in mice, compared the "periovulatory" and "perimenstrual" phases. Senie and coworkers (3) compared the "follicular" with the "luteal" phases. At

1 February 1992 • Annals of Internal Medicine • Volume 116 • Number 3


ings indicate that attempts should be made to understand and explain differing results from other centers, rather than to dismiss them as chance findings. We have examined our recent data (since 1985) in which the surgical procedures have been altered to include diagnostic needle biopsies in a small number of patients (10). In addition to confirming our previous findings, we did a separate analysis on patients who had surgery during a "favorable" phase. In this group, those who had had needle biopsy during the unfavorable phase (days 3 to 12) fared worse than the remainder. This finding suggests a possible explanation for the observed discrepancies between studies from different centers. We suggest that other groups examine the timing of all surgical interventions. W. M. Gregory, MD M. A. Richards, MD I. S. Fentiman, MD Guy's Hospital London SE1 9RT England

Figure 1. Overview of 5 year survival results among patients with breast cancer. The area of the black squares represents the variance for individual studies; the lines represent 999c CI for the typical odds ratios. The diamond shows a 959c CI for all the results combined. our facility, we tested the hypothesis that unopposed estrogen (days 3 to 12 of the cycle) might result in poorer outcome (4). This interval was derived from examination of published hormone profiles and preceded the analysis. As a final check, we calculated the 10-year death rates for smaller intervals, which showed the death rates rising and falling in accordance with unopposed estrogen levels (4). Thus, the data were seen to fit the a-priori hypothesis; the suggested random element was not present. The second point concerns the level of statistical significance found among patients studied at Guy's Hospital. The reported chi-square statistic was 29.3 (P = 6 x 10" 8 ); that is, a probability of less than one in a million of this being a chance finding. We repeated Dr. McGuire's simulation using the data set from our original study (249 patients). When looking for a P value of 0.05 we also found approximately 2S9c of positive cases out of 100 sets of random assignments. However, when the P value of 6 x 10" 8 was used, no positive results were found after 10 000 further repetitions. As a more appropriate method for examining whether our result could be a chance finding we conducted a meta-analysis of 5-year survival rates for all 10 reported studies (2-5) using standard methods. The results are shown in Figure 1. Data were obtained using careful measurement of the published survival curves. The effective sample size at 5 years was estimated using the method described by Simon, (6) and the number of deaths was derived. Five-year figures were chosen because, in the studies reported as significant, the curves had diverged considerably and most had a median follow-up of greater than 5 years. One paper (7) did not report survival data, and disease-free survival was used instead. Two studies reported survival based on intervals of 0 to 14 compared with 15 to 32 days (4, 7), and these studies are also included. One study (8) only reported results based on hormone levels, dividing the cycle into perimenstrual, follicular, ovulatory, and luteal. To test the unopposed estrogen hypothesis on the Guy's Hospital sample, only data from the follicular and luteal phases were used in the meta-analysis for this study; the other two intervals span more than one estrogenic phase and were not included. The meta-analysis showed a significant overall effect of the timing of surgery (P = 0.003). However, a test for heterogeneity of the observed results of the meta-analysis (9) yielded a chi-square of 18.9 (9 degrees of freedom), giving a P value of 0.009. This shows differences between the studies that are very unlikely to be explained by chance. Thus, the findings from our center should not be dismissed on these grounds. The magnitude and statistical significance of the Guy's find-

References 1. McGuire WL. The optimal timing of mastectomy: low tide or high tide? Ann Intern Med. 1991:115:401-3. 2. Hrushesky WJ, Bluming AZ, Gruber SA, Sothern RB. Menstrual influence on surgical cure of breast cancer. Lancet. 1989;2:949-52. 3. Senie TJ, Rosen PP, Rhodes P, Lesser ML. Timing of breast cancer excision during the menstrual cycle influences duration of diseasefree survival. Ann Intern Med. 1991:115:337-42. 4. Badwe RA, Gregory WM, Chaudary MA, Richards MA, Bentley AE, Rubens RD, et al. Timing of surgery during menstrual cycle and survival of premenopausal women with operable breast cancer. Lancet. 1991:337:1261-4. 5. Rageth JC, Wyss P, Unger C, Hochuli E. Timing of breast cancer surgery within the menstrual cycle: influence of lymph-node involvement, receptor status, postoperative metastatic spread and local recurrence. Ann Oncol. 1991:2:269-72. Additional references (6-10) available from the authors on request. Transplants from the Same Donor To the Editors: Spitzer, Zwiebel and Jacobson (1) recently reported the case of a patient who received an HLA-identical bone marrow transplant after receiving a renal allograft from the same donor. The patient developed severe graft-versushost disease with persistent leukemia and died 40 days after transplantation. We report a similar instance of grade-Ill graftversus-host disease treated successfully. A 37-year-old white man with end-stage renal disease secondary to rapidly progressive glomerulonephritis had a kidney transplant from an HLA-identical sibling donor on 31 July 1980. The patient received azathioprine and prednisone therapy until September 1988, when he developed pancytopenia. The azathioprine therapy was discontinued, and his hematocrit normalized during the following 6 weeks with continued prednisone therapy (10 mg/d). In February 1989, the patient again presented with pancytopenia, and a bone marrow evaluation suggested a myelodysplastic syndrome. In March 1989, a repeat bone marrow evaluation, including morphology, immunophenotyping, and histochemistries, showed acute myelogenous leukemia, FAB M4. Cytogenetics revealed a deletion in the long arm of chromosome 8 (8q-). The patient was successfully treated (induced) with cytarabine and daunorubicin and the remission was maintained (consolidated) with two courses of cytarabine and daunorubicin. In December 1989, bone marrow examination for progressing pancytopenia revealed a leukemic relapse. The patient received a conditioning regimen of busulfan, 16 mg/kg body weight, and cyclophosphamide, 120 mg/kg. On 18 December 1989, the patient received an allogeneic bone marrow transplant (3.3 x 108 mononuclear cells/kg) from the same sibling. Prophylaxis for graft-versus-host disease consisted of cyclosporin and methylprednisolone. The immediate post-transplant course was complicated by slow hematopoietic engraftment, but residual leukemia was not evident. The patient received a second bone marrow transplant from the same donor (1.72 x 108 mononuclear cells/kg) without further chemotherapy on 9 January 1990 (day 22). He showed neutrophil recovery (more

1 February 1992 • Annals of Internal Medicine • Volume 116 • Number 3 Downloaded From: http://annals.org/ by a National Institutes of Health User on 06/12/2013

269

than 0.5 x 109/L) on day 7 and platelet independence (more than 2 x 109/L) on day 15 after the second bone marrow transplantation. The patient developed grade-Ill graft-versushost disease manifested as erythrodermatous skin rash (biopsy proved), profuse bloody diarrhea (rectal biopsy findings were consistent with graft-versus-host disease), and hyper-bilirubinemia (maximum bilirubin level, 6.4 mg/dL). He was aggressively supported, and the immunosuppressants were optimized. The patient subsequently developed chronic graft-versus-host disease with skin and liver involvement, which has been controlled with 10 mg/d of prednisone. He remains in clinical remission more than 22 months after bone marrow transplantation, with good performance status and a functional renal allograft (serum creatinine level, 1.5 mg/dL). We concur with Spitzer and colleagues (1) that previous sensitization to minor histocompatibility antigens may enhance the risk for graft rejection and that "tolerance" to donor alloantigens may increase the risk for acute graft-versus-host disease. In contrast with their findings, our patient had slow engraftment but eventually had complete hematologic reconstitution without evidence of relapse. Further, the development of grade-Ill graft-versus-host disease probably enhanced the graft-versus-leukemia effect that contributed to the continued relapse-free survival in this patient who, historically, would be at high risk for relapse because he received his transplant while in leukemic relapse and because he developed leukemia secondary to prolonged alkylator treatment (2). These two cases illustrate a problem that potentially may be encountered with allogeneic bone marrow transplantation in patients who are also recipients of organs from the same HLAidentical sibling. In our case, the presence of the renal allograft neither made the recipient more tolerant to the subsequent bone marrow transplant nor produced graft failure. In this small high-risk group, a combination of regimens, including therapy with immunosuppressants (cyclosporine, methotrexate, glucocorticoid combinations [3]) or preemptive therapy with immunomodulators (4, 5), or both, may be required to alleviate life-threatening graft-versus-host disease. D. H. Vesole, MD, PhD S. Jagannath, MD University of Arkansas for Medical Sciences Little Rock, AR 72205 References 1. Spitzer TR, Zwiebel J, Jacobson RJ. Transplants from the same donor [Letter]. Ann Intern Med. 1991 ;115:498. 2. Auclerc G, Jacquillat C, Auclerc MF, Weil M, Bernard J. Posttherapeutic acute leukemia. Cancer. 1979;44:2017-25. 3. Yau JC, LeMaistre CF, Zagers GK, et al. Methylprednisolone, cy and methotrexate for prophylaxis of acute graft-versus-host disease. Bone Marrow Transpl. 1990;5:269-72. 4. LaMaistre CF, Meneghetti C, Yau JC, et al. Pre-emptive therapy of graft vs. host disease (GVHD) with a pan-T-cell immunotoxin (IT) associated with accelerated engraftment [Abstract]. Antibody, Immunoconjugates and Radiopharmaceuticals. 1990;3:70. 5. Blaise D, Olive D, Hirn M, et al. Prevention of acute GVHD by in vivo use of anti-interleukin-2 receptor monoclonal antibody (33B3.1): a feasibility trial in 15 patients. Bone Marrow Transpl. 1991;8:105-11. Anabolic Steroids and Muscle Strength To the Editors: The article by Elashoff and colleagues (1) provided a comprehensive assessment of the status of our current knowledge of the possible effects of androgens on muscular strength. In addition to the numerous deficiencies of the reported studies noted by Elashoff, we point out that none of the studies provided any measurement of the steroid levels obtained in blood or tissue. This factor may result from the difficulty of measuring in serum the variety of anabolic agents taken by athletes. However, considering the multiple routes of androgen metabolism it is difficult to determine an androgen response without some measure of the steroid level achieved. In our study, we report the relation between change in muscle strength using Cybex testing and change in serum androgen levels in normal, young, physically active men receiving testosterone enanthate, either 100 mg/wk (five subjects) or

270

Table 1. Correlation between Serum Androgen Measurements and Change in Muscle Strength* Serum Hormone Measurement

R Value

P Value

Change in Testosterone Change in T:SHBG Change in 3a-diol glucuronide

0.755 0.756 0.809

< 0.02 < 0.02 < 0.01

* SHBG = sex hormone binding-globulin, T = testosterone. 300 mg/wk (4 subjects) intramuscularly, for 6 weeks. The mean age of the men was 27 ± 5 (SD) years. They ate self-selected diets that tended to be low in fat and high in carbohydrate during the course of the study without nutritional supplements. Participants and investigators were blinded to the testosterone dose throughout the study. Additional data on the study design has been reported previously. Serum testosterone levels and sex-hormone binding-globulin (SHBG) were measured as previously described (2). Three-alpha-androstanediol glucuronide (3a-diol) was measured by Nichols Laboratory (Capistrano, California). Although 3a-diol glucuronide is not produced to any substantial degree in muscle, it does provide a measure of peripheral androgen action at the cellular level (3, 4). A significant positive correlation was seen between the changes in serum testosterone, testosterone-to-SHBG ratios, and 3a-diol glucuronide on treatment and changes in strength as measured by isometric flexion at the elbow (Table 1). No significant correlation was noted between serum androgen levels and flexion at the knee or extension at the elbow or knee. In summary, this study shows a positive correlation between changes in serum androgens and muscle strength in a group of physically trained men receiving testosterone. However, these changes in strength are small (7.6%). Stephen R. Plymate, MD McGuire Veterans Affairs Medical Center Medical College of Virginia Richmond, VA 23249 Karl E. Friedl, PhD U.S. Army Research Institute of Environmental Medicine Natick, MA 01760 References 1. Elashoff J, Jacknow A, Shain S, Braunstein G. Effects of anabolicandrogenic steroids on muscular strength. Ann Intern Med. 1991 ;115: 387-93. 2. Friedl K, Jones R, Hannan C, Plymate S. The administration of pharmacological doses of testosterone or 19-nortestosterone to normal men is not associated with increased insulin secretion or impaired glucose tolerance. J Clin Endocrinol Metab. 1989;68:971-5. 3. Horton R, Hawks D, Lobo R. 3-Alpha, 17-beta-androstanediol glucuronide in plasma: a marker of androgen action in normal healthy Caucasian versus Chinese subjects. J Clin Invest. 1982;69:1203-6. 4. Lookingbill D, Demers L, Wang C, Leung A, Rittmaster R, Santen R. Clinical and biochemical parameters of androgen action in normal healthy Caucasian versus Chinese subjects. J Clin Endocrinol Metab. 1991;72:1242-8. In response: We appreciate the comments of Drs. Plymate and Friedl. Their study illustrates two of the findings that were noted in several of the studies that we reviewed. First, the changes in muscle strength were small. Second, much variation exists between muscle groups; some show an apparent increase in strength, whereas others do not. It should be emphasized that their 6-week study does not reflect the "street" use of anabolic steroids either in duration or dose, both of which tend to be several fold greater than those that they tested. Glenn D. Braunstein, MD Janet D. Elashoff, PhD Sara G. Shain, MS Cedars-Sinai Medical Center—UCLA School of Medicine Los Angeles, CA 90048

1 February 1992 • Annals of Internal Medicine • Volume 116 • Number 3
