CORRESPONDENCE
May 1992
has long been of concern; notwithstanding, the autopsy findings in our own case of INISD effectively rule out hemophagocytic lymphohistiocytosis (HLH). The absence of a significant number of histiocytes and lymphocytes in the liver, spleen, and bone marrow [not discussed in our case report] of our patient, as well as the absence of splenomegaly (the spleen weighed 7.9 g in our patient vs. 12.9 g in an average 5-week-old boy) argue against HLH. In addition, the distinctive distribution of iron in nonreticuloendothelial cells in our patient would be unusual in HLH. It should be noted that hemophagocytic syndromes may develop under a variety of circumstances such as severe viral infection, malignancy, fat overload, and following blood transfusion.’ JOHN A. BARNARD,
III, M.D.
Division of Gastroenterology/Nutrition Department of Pediatric Gastroenterology Vanderbilt University Medical Center Nashville, Tennessee 1. Henter J, Elinder G, Ost A, et al. Diagnostic guidelines for hemophagocytic lymphohistiocytosis. Semin Oncol 1991;16:2933.
Geography and Inflammatory
Bowel Disease
Dear Sir: Sonnenberg et al.’ have used national rates of hospitalization for inflammatory bowel disease (IBD) to show marked variations in hospital discharges and the standardized mortality ratios for both Crohn’s disease and ulcerative colitis. Lower ratios were observed in the southern states except for Florida and Arizona. Although their observation sdggests that environmental factors may play a major role in the etiology of IBD, a note of caution is required. Only a small proportion of IBD cases may ever be hospitalized. In California, the annual incidence of ulcerative proctocolitis in a defined health plan population was 10.9/100,000 and 7.0/100,000 for Crohn’s disease, whereas the rates of first hospitalizations were only 5.5/100,000 and 5.2/100,000, respectively.* Of 28 new IBD cases diagnosed in one year, only 21% required hospitalization in the next 31/2years. Sonnenberg et al. believe that medical practice patterns in the United States are fairly “homogeneous.” However, the work of Wennberg et al. 3.4has shown that admissions for such procedures as hysterectomy and prostatectomy have large and significant differences in medical practice patterns. In the northern states, if practice patterns were more likely to result in hospitalization for IBD, their influence on the variation of IBD hospitalization could be substantial. The similar pattern observed for mortality data certainly strengthens the authors’ conclusion that the observed regional differences are real. However, variation in death certificate coding for IBD could also vary along with medical practice. ROBERT A. HIATT, M.D.
Kaiser Permanente Medical Care Program Division of Research LEON KAUFMAN,
M.D.
Department of Medicine Kaiser Permanente Medical Center Oakland, California Sonnenberg A, McCarty DJ, Jacobsen SJ. Geographic variation of inflammatory bowel disease within the United States. Gastroenterology 1991;100:143-149. Hiatt RA, Kaufman L. Epidemiology of inflammatory bowel disease in a defined northern California population. West J Med 1988;149:541-546. Wennberg J, Gittelsohn A. Variations in medical care among small areas. Sci Am 1982;246:120-134.
1827
4. Wennberg JE, Freeman JL, Shelton RM, Bubolz TA. Hospital use and mortality among Medicare beneficiaries in Boston and New Haven. N Engl J Med 1989:321:1168-1173.
Diagnosing a Diagnostic StudyLipase/Amylase Ratio Dear Sir: We read with great interest the recent paper by Gumaste et al.’ regarding the diagnostic utility of the lipase/amylase ratio. We were intrigued with the idea of having a blood test that could readily discriminate between alcoholic and nonalcoholic causes of acute pancreatitis. Unfortunately, we are not yet willing to accept the authors’ results because their study may have introduced several potential biases recognized in the development of diagnostic tests.z*3 We are unsure what was used as the “gold standard” to determine the final diagnosis. The authors state that the diagnosis in phase one of their study was “based on a combination of clinical features, changing amylase levels, and imaging studies” and in phase two “the diagnosis was based purely on clinical and laboratory criteria.” A form of incorporation bias may have been introduced into the study if the results of the lipase/amylase ratios were used as part of the evidence to help make the final diagnosis of pancreatitis. Were the authors’ preset diagnostic criteria for pancreatitis applied in a blinded and independent manner aside from the lipase/ amylase ratio? We are not provided reassurance that the chart reviewers for the phase one component of the study were blinded to the hypothesis tested. Failure to do this could have introduced a diagnostic-review bias whereby knowledge of the lipase/amylase ratio may have influenced the chart review of the data used to establish the final diagnosis of pancreatitis. In phase two of the study, evidence that the lipase/amylase ratio was not compared with the preset diagnostic criteria in a blinded manner is given by the anecdotal case report related in the discussion section. Here the authors were about to embark on an extensive workup to determine the etiology of the pancreatitis. However, because the lipase/amylase ratio was greater than two, they “repeatedly interviewed” the patient regarding his drinking habits until he finally admitted to heavy alcohol abuse, and no further workup was done. This is an example of workup bias because knowledge of the ratio result affected the intensity of the clinical workup toward the establishment of a particular diagnosis. Ideally, the same intensity of workup should have been equally applied to all patients independent of the lipase/amylase ratio results. Finally, we question whether the spectrum of patients used to test the validity of the lipase/amylase ratio was representative of the general population. This becomes evident when one analyzes the utility of the patients’ sex in distinguishing between alcoholic and nonalcoholic pancreatitis. In phase one of the study, all 20 patients with alcoholic pancreatitis were male, whereas all 10 patients with nonalcoholic pancreatitis were female. Sex, therefore, had a sensitivity, specificity, and predictive value of loo%! In phase two of the study, all but 1 patient with alcoholic pancreatitis were male, and all but 1 patient with nonalcoholic pancreatitis were female. Sex, therefore, had a sensitivity of 92%, specificity of as%, and a positive predictive value of 92%. which is also better than the comparable figures reported for the lipase/amylase ratio in phase two of the study. We are still very intrigued and compliment the authors for their idea of a lipase/amylase ratio in distinguishing between alcoholic and nonalcoholic forms of acute pancreatitis. Such a simple noninvasive blood test, if validated, would certainly be clinically useful and practical. However, we agree with and affirm the authors final statement of the article that further testing and duplication
1828
CORRESPONDENCE
of their study results from other centers is required before the lipase/amylase ratio can be widely accepted and adopted as a useful clinical criterion. LAWRENCE K. LOO, M.D. ZEN0 L. CHARLES-MARCEL, M.D. FRANZ FISHER, M.D. TIMOTHY RICHARDSON, M.D. DANIEL CASTRO, M.D. MOUNA HADDAD, M.D.
Department of Medicine Loma Linda University Jerry L. Pettis Memorial VA Hospital Loma Linda, California 92357 Gumaste VV, Dave PB, Weissman D, Messer J: Lipase/amylase ratio. A new index that distinguishes acute episodes of alcoholic from nonalcoholic acute pancreatitis. Gastroenterology 1991;101:1381-1388. Department of Clinical Epidemiology and Biostatistics, McMaster University Health Sciences Center. How to read clinical journals: II. To learn about a diagnostic test. Can Med Assoc J 1981;124:703-710. Ransohoff DF, Feinstein AR: Problems of spectrum and bias in evaluating the efficacy of diagnostic tests. N Engl J Med 1978;299:926-930.
Reply. We appreciate the comments made by Dr. Loo et al. with regard to our study1 and we would like to clarify the points they have raised. First, the lipase/amylase ratio was not used as a part of the evidence to make a final diagnosis of pancreatitis. The diagnosis of pancreatitis was based on a combination of clinical features, laboratory values, and imaging studies in both phases of the study. Details regarding the diagnosis are clearly outlined on page 1362, column 2, lines 9-43. Further, Loo et al. appear to have misread the paper in stating that in phase two “the diagnosis was based primarily on clinical and laboratory criteria.” In only 3 of the 21 patients in phase two was the diagnosis based primarily on clinical and laboratory criteria (p. 1362, col. 2, lines 41-43). In phase two, the diagnosis was.confirmed by ultrasonography in 15 patients, computed tomography scanning in 2, and at surgery in 1 patient @. 1362, col. 2, lines 35-41). All patients in the study were meticulously interviewed with regard to their use of alcohol. The patient who was repeatedly interviewed belonged to phase one of the study. Phase two of the study was undertaken mainly to overcome any bias that may have crept into the phase one of the trial. The patients in phase two constituted an unselected population. The etiology of the pancreatitis was ascertained purely by clinical data and imagingstudies and was performed independent of the lipase/amylase ratio. Further, the criteria we used were objective enough to overcome any bias. Eight of the 9 patients deemed to have gallstone pancreatitis had gallstones on ultrasonography. The 1 other patient had a dilated common bile duct on ultrasonography and computed tomography scan. None of these patients had a history of regular intake of alcohol. All 12 patients in the alcoholic group had a history of alcohol abuse ranging from 5 to 24 years, and none of them had evidence of gallstones on ultrasonography. With regard to the spectrum of patients, we can only state that we conducted the study on the consecutive patients who presented to us. It is a well-known fact that alcoholic pancreatitis predominates in men, whereas gallstone pancreatitis is common among women. To seek equal distribution in a population where one does not exist merely for statistical reasons is to lose sight of the purpose of a study. In conclusion, we would like to thank Dr. Loo et al. for their keen interest in our paper and would like to re-emphasize the fact
GASTROENTEROLOGY Vol. 102, No. 5
that this study needs to be duplicated can be an accepted criterion.
by other centers before it
VIVEK GUMASTE, M.D. P. DAVE, M.D. D. WEISSMAN. M.D. J. MESSER, M.D.
Gastroenterology Division Mount Sinai Services at the City Hospital at Elmhurst Elmhurst, New York 11373 1. Gumaste VV, Dave PB, Weissman
D, Messer J. Lipase/amylase ratio: a new index that distinguishes acute episodes of alcoholic from nonalcoholic pancreatitis. Gastroenterology 1991:101:1361-1366.
Sensitivity Criteria
and Specificity of the Manning
Dear Sir: The recent very useful and informative article by Talley et al.’ seems to contain a miscalculation of the sensitivity and specificity of the Manning criteria for diagnosing irritable bowel syndrome (IBS). This is unfortunate, because it makes the Manning criteria appear even less effective than they are in excluding organic diseases, thereby detracting unnecessarily from the important findings of the paper. In Table 2 on page 654 of the original paper by Manning et al.,’ the following values can be determined, as they also calculated. With two or more criteria, the sensitivity is 0.91 and the specificity is 0.70; with three or more criteria present, the sensitivity is 0.63 and the specificity is 0.85. ROBERT C. SMITH, M.D. Departments ofInternal Medicine and Psychiatry Michigan State University College of Human Medicine East Lansing, Michigan 48824 1. Talley NJ, Zinsmeister AR, Van Dyke C, Melton J. Epidemiology of colonic symptoms and the irritable bowel syndrome. Gastroenterology 1991;101:927-934. 2. Manning AP, Thompson WC, Heaton KW, Morris AF. Towards positive diagnosis of the irritable bowel. Br Med J 1978;2:653654. Reply. We thank Dr. Smith for his kind comments and are happy to have the opportunity to clear up any misunderstanding about the sensitivity and specificity of the Manning criteria. As stated in our article,’ the Manning criteria by convention include six symptoms, not four. 2d When all six symptoms are taken into account, as is presented in Table 3 of the original article by Manning et aLs the results are as we stated in our paper. However, if only four criteria are used (as described by Dr. Smith], then of course different results are obtained, as shown in Table 1.
Table 1. Manning
Criteria for Irritable 22
Sensitivity (%) Specificity (% I
of
4O
22
of
6b
Bowel Syndrome r3
of
4’
23
of
91
94
63
84
70
55
85
76
6b
aThe four symptoms are pain relief after defecation, more frequent and looser stools after onset of pain, and abdominal distention. bSix symptoms include the four above plus mucus and feeling of complete evacuation.
May 1992
has long been of concern; notwithstanding, the autopsy findings in our own case of INISD effectively rule out hemophagocytic lymphohistiocytosis (HLH). The absence of a significant number of histiocytes and lymphocytes in the liver, spleen, and bone marrow [not discussed in our case report] of our patient, as well as the absence of splenomegaly (the spleen weighed 7.9 g in our patient vs. 12.9 g in an average 5-week-old boy) argue against HLH. In addition, the distinctive distribution of iron in nonreticuloendothelial cells in our patient would be unusual in HLH. It should be noted that hemophagocytic syndromes may develop under a variety of circumstances such as severe viral infection, malignancy, fat overload, and following blood transfusion.’ JOHN A. BARNARD,
III, M.D.
Division of Gastroenterology/Nutrition Department of Pediatric Gastroenterology Vanderbilt University Medical Center Nashville, Tennessee 1. Henter J, Elinder G, Ost A, et al. Diagnostic guidelines for hemophagocytic lymphohistiocytosis. Semin Oncol 1991;16:2933.
Geography and Inflammatory
Bowel Disease
Dear Sir: Sonnenberg et al.’ have used national rates of hospitalization for inflammatory bowel disease (IBD) to show marked variations in hospital discharges and the standardized mortality ratios for both Crohn’s disease and ulcerative colitis. Lower ratios were observed in the southern states except for Florida and Arizona. Although their observation sdggests that environmental factors may play a major role in the etiology of IBD, a note of caution is required. Only a small proportion of IBD cases may ever be hospitalized. In California, the annual incidence of ulcerative proctocolitis in a defined health plan population was 10.9/100,000 and 7.0/100,000 for Crohn’s disease, whereas the rates of first hospitalizations were only 5.5/100,000 and 5.2/100,000, respectively.* Of 28 new IBD cases diagnosed in one year, only 21% required hospitalization in the next 31/2years. Sonnenberg et al. believe that medical practice patterns in the United States are fairly “homogeneous.” However, the work of Wennberg et al. 3.4has shown that admissions for such procedures as hysterectomy and prostatectomy have large and significant differences in medical practice patterns. In the northern states, if practice patterns were more likely to result in hospitalization for IBD, their influence on the variation of IBD hospitalization could be substantial. The similar pattern observed for mortality data certainly strengthens the authors’ conclusion that the observed regional differences are real. However, variation in death certificate coding for IBD could also vary along with medical practice. ROBERT A. HIATT, M.D.
Kaiser Permanente Medical Care Program Division of Research LEON KAUFMAN,
M.D.
Department of Medicine Kaiser Permanente Medical Center Oakland, California Sonnenberg A, McCarty DJ, Jacobsen SJ. Geographic variation of inflammatory bowel disease within the United States. Gastroenterology 1991;100:143-149. Hiatt RA, Kaufman L. Epidemiology of inflammatory bowel disease in a defined northern California population. West J Med 1988;149:541-546. Wennberg J, Gittelsohn A. Variations in medical care among small areas. Sci Am 1982;246:120-134.
1827
4. Wennberg JE, Freeman JL, Shelton RM, Bubolz TA. Hospital use and mortality among Medicare beneficiaries in Boston and New Haven. N Engl J Med 1989:321:1168-1173.
Diagnosing a Diagnostic StudyLipase/Amylase Ratio Dear Sir: We read with great interest the recent paper by Gumaste et al.’ regarding the diagnostic utility of the lipase/amylase ratio. We were intrigued with the idea of having a blood test that could readily discriminate between alcoholic and nonalcoholic causes of acute pancreatitis. Unfortunately, we are not yet willing to accept the authors’ results because their study may have introduced several potential biases recognized in the development of diagnostic tests.z*3 We are unsure what was used as the “gold standard” to determine the final diagnosis. The authors state that the diagnosis in phase one of their study was “based on a combination of clinical features, changing amylase levels, and imaging studies” and in phase two “the diagnosis was based purely on clinical and laboratory criteria.” A form of incorporation bias may have been introduced into the study if the results of the lipase/amylase ratios were used as part of the evidence to help make the final diagnosis of pancreatitis. Were the authors’ preset diagnostic criteria for pancreatitis applied in a blinded and independent manner aside from the lipase/ amylase ratio? We are not provided reassurance that the chart reviewers for the phase one component of the study were blinded to the hypothesis tested. Failure to do this could have introduced a diagnostic-review bias whereby knowledge of the lipase/amylase ratio may have influenced the chart review of the data used to establish the final diagnosis of pancreatitis. In phase two of the study, evidence that the lipase/amylase ratio was not compared with the preset diagnostic criteria in a blinded manner is given by the anecdotal case report related in the discussion section. Here the authors were about to embark on an extensive workup to determine the etiology of the pancreatitis. However, because the lipase/amylase ratio was greater than two, they “repeatedly interviewed” the patient regarding his drinking habits until he finally admitted to heavy alcohol abuse, and no further workup was done. This is an example of workup bias because knowledge of the ratio result affected the intensity of the clinical workup toward the establishment of a particular diagnosis. Ideally, the same intensity of workup should have been equally applied to all patients independent of the lipase/amylase ratio results. Finally, we question whether the spectrum of patients used to test the validity of the lipase/amylase ratio was representative of the general population. This becomes evident when one analyzes the utility of the patients’ sex in distinguishing between alcoholic and nonalcoholic pancreatitis. In phase one of the study, all 20 patients with alcoholic pancreatitis were male, whereas all 10 patients with nonalcoholic pancreatitis were female. Sex, therefore, had a sensitivity, specificity, and predictive value of loo%! In phase two of the study, all but 1 patient with alcoholic pancreatitis were male, and all but 1 patient with nonalcoholic pancreatitis were female. Sex, therefore, had a sensitivity of 92%, specificity of as%, and a positive predictive value of 92%. which is also better than the comparable figures reported for the lipase/amylase ratio in phase two of the study. We are still very intrigued and compliment the authors for their idea of a lipase/amylase ratio in distinguishing between alcoholic and nonalcoholic forms of acute pancreatitis. Such a simple noninvasive blood test, if validated, would certainly be clinically useful and practical. However, we agree with and affirm the authors final statement of the article that further testing and duplication
1828
CORRESPONDENCE
of their study results from other centers is required before the lipase/amylase ratio can be widely accepted and adopted as a useful clinical criterion. LAWRENCE K. LOO, M.D. ZEN0 L. CHARLES-MARCEL, M.D. FRANZ FISHER, M.D. TIMOTHY RICHARDSON, M.D. DANIEL CASTRO, M.D. MOUNA HADDAD, M.D.
Department of Medicine Loma Linda University Jerry L. Pettis Memorial VA Hospital Loma Linda, California 92357 Gumaste VV, Dave PB, Weissman D, Messer J: Lipase/amylase ratio. A new index that distinguishes acute episodes of alcoholic from nonalcoholic acute pancreatitis. Gastroenterology 1991;101:1381-1388. Department of Clinical Epidemiology and Biostatistics, McMaster University Health Sciences Center. How to read clinical journals: II. To learn about a diagnostic test. Can Med Assoc J 1981;124:703-710. Ransohoff DF, Feinstein AR: Problems of spectrum and bias in evaluating the efficacy of diagnostic tests. N Engl J Med 1978;299:926-930.
Reply. We appreciate the comments made by Dr. Loo et al. with regard to our study1 and we would like to clarify the points they have raised. First, the lipase/amylase ratio was not used as a part of the evidence to make a final diagnosis of pancreatitis. The diagnosis of pancreatitis was based on a combination of clinical features, laboratory values, and imaging studies in both phases of the study. Details regarding the diagnosis are clearly outlined on page 1362, column 2, lines 9-43. Further, Loo et al. appear to have misread the paper in stating that in phase two “the diagnosis was based primarily on clinical and laboratory criteria.” In only 3 of the 21 patients in phase two was the diagnosis based primarily on clinical and laboratory criteria (p. 1362, col. 2, lines 41-43). In phase two, the diagnosis was.confirmed by ultrasonography in 15 patients, computed tomography scanning in 2, and at surgery in 1 patient @. 1362, col. 2, lines 35-41). All patients in the study were meticulously interviewed with regard to their use of alcohol. The patient who was repeatedly interviewed belonged to phase one of the study. Phase two of the study was undertaken mainly to overcome any bias that may have crept into the phase one of the trial. The patients in phase two constituted an unselected population. The etiology of the pancreatitis was ascertained purely by clinical data and imagingstudies and was performed independent of the lipase/amylase ratio. Further, the criteria we used were objective enough to overcome any bias. Eight of the 9 patients deemed to have gallstone pancreatitis had gallstones on ultrasonography. The 1 other patient had a dilated common bile duct on ultrasonography and computed tomography scan. None of these patients had a history of regular intake of alcohol. All 12 patients in the alcoholic group had a history of alcohol abuse ranging from 5 to 24 years, and none of them had evidence of gallstones on ultrasonography. With regard to the spectrum of patients, we can only state that we conducted the study on the consecutive patients who presented to us. It is a well-known fact that alcoholic pancreatitis predominates in men, whereas gallstone pancreatitis is common among women. To seek equal distribution in a population where one does not exist merely for statistical reasons is to lose sight of the purpose of a study. In conclusion, we would like to thank Dr. Loo et al. for their keen interest in our paper and would like to re-emphasize the fact
GASTROENTEROLOGY Vol. 102, No. 5
that this study needs to be duplicated can be an accepted criterion.
by other centers before it
VIVEK GUMASTE, M.D. P. DAVE, M.D. D. WEISSMAN. M.D. J. MESSER, M.D.
Gastroenterology Division Mount Sinai Services at the City Hospital at Elmhurst Elmhurst, New York 11373 1. Gumaste VV, Dave PB, Weissman
D, Messer J. Lipase/amylase ratio: a new index that distinguishes acute episodes of alcoholic from nonalcoholic pancreatitis. Gastroenterology 1991:101:1361-1366.
Sensitivity Criteria
and Specificity of the Manning
Dear Sir: The recent very useful and informative article by Talley et al.’ seems to contain a miscalculation of the sensitivity and specificity of the Manning criteria for diagnosing irritable bowel syndrome (IBS). This is unfortunate, because it makes the Manning criteria appear even less effective than they are in excluding organic diseases, thereby detracting unnecessarily from the important findings of the paper. In Table 2 on page 654 of the original paper by Manning et al.,’ the following values can be determined, as they also calculated. With two or more criteria, the sensitivity is 0.91 and the specificity is 0.70; with three or more criteria present, the sensitivity is 0.63 and the specificity is 0.85. ROBERT C. SMITH, M.D. Departments ofInternal Medicine and Psychiatry Michigan State University College of Human Medicine East Lansing, Michigan 48824 1. Talley NJ, Zinsmeister AR, Van Dyke C, Melton J. Epidemiology of colonic symptoms and the irritable bowel syndrome. Gastroenterology 1991;101:927-934. 2. Manning AP, Thompson WC, Heaton KW, Morris AF. Towards positive diagnosis of the irritable bowel. Br Med J 1978;2:653654. Reply. We thank Dr. Smith for his kind comments and are happy to have the opportunity to clear up any misunderstanding about the sensitivity and specificity of the Manning criteria. As stated in our article,’ the Manning criteria by convention include six symptoms, not four. 2d When all six symptoms are taken into account, as is presented in Table 3 of the original article by Manning et aLs the results are as we stated in our paper. However, if only four criteria are used (as described by Dr. Smith], then of course different results are obtained, as shown in Table 1.
Table 1. Manning
Criteria for Irritable 22
Sensitivity (%) Specificity (% I
of
4O
22
of
6b
Bowel Syndrome r3
of
4’
23
of
91
94
63
84
70
55
85
76
6b
aThe four symptoms are pain relief after defecation, more frequent and looser stools after onset of pain, and abdominal distention. bSix symptoms include the four above plus mucus and feeling of complete evacuation.
