J Clia Epidea~ielVol. 47, No. 10, pp. 1175-I 179, 1994 Pergamon
Copyright 0 1994 Ekevier Science Ltd Printed in Great Britain. All rights reserved 089543S6/94 $7.00 + 0.00
RANITIDINE-ASSOCIATED AUTOIMMUNE HEMOLYTIC ANEMIA IN A HEAiTH MAINTENANCE ORGANIZATION POPULATION PETER W. CHDD,‘* JOANH. GOLDBERG’and RICHARD PLATT’ ‘Charming Laboratory, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts, ‘Harvard Community Health Plan, Boston, Massachusetts and )Department of Ambulatory Care and Preventive Medicine, Harvard Community Health Plan and Harvard Medical School, Channing Laboratory, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts, U.S.A. (Received in revised form 8 February 1994)
Abstract-Reversible hematologic abnormalities including hemolytic anemia [I] with a positive direct Coombs’ test have been associated with ranitidine. In addition to the case
report cited above, the U.S. Food and Drug Administration had received five other cases of hemolysis associated with recent intake of ranitidine as of February 1991. To investigate the possible association of ranitidine with autoimmune hemolytic anemia, a study was conducted to determine how often diagnoses of hemolytic anemia or abnormal Coombs’ test results followed dispensing of ranitidine using the automated medical and pharmacy records of a large health maintenance organization. No occurrences of hemolytic anemia were identified among 12,054 individuals following 38,686 prescriptions for this medication. The 95% upper confidence bound was 3.1 cases/lO,OOOexposed persons. One abnormal direct Coombs’ test with mild anemia was discovered during routine prenatal testing of an asymptomatic patient who was dispensed ranitidine two and a half months previously. Hemolysis, however, was not demonstrated and an association with prior ranitidine use could not be confirmed. Additional analyses indicate that in only 30% of ranitidine courses was a blood count obtained. In those courses with hematocrits below 40%, less than 1% had a Coombs’ test performed. Chart review suggests that the majority of individuals with severe anemia have alternative explanations other than autoimmune hemolysis for their anemia. This analysis indicates that ranitidine is unlikely to be a common cause of clinically recognized autoimmune hemolytic anemia and demonstrates the utility of large automated medical and pharmacy data bases to conduct post-marketing studies of spontaneously reported adverse drug effects. Computerized medical records Ranitidine Autoimmune hemolytic anemia Automated record linkHealth maintenance organization Pharmacoepidemiology age system
INTRODUCTION Several reports have associated ranitidine and other Hz-receptor antagonists with various *All correspondence
should be addressed to: Dr Peter W. Choo, Channing Laboratory, 180 Longwood Avenue, Boston, MA 02115, U.S.A.
hematologic abnormalities, including granulocytopenia, thrombocytopenia and hemolytic anemia with a positive direct Coombs’ test [l-8]. By F e b ruary of 1991, the U.S. Food and Drug Administration (FDA) had received six spontaneous reports of hemolysis associated with the use of ranitidine and began specific monitoring 1175
for this possible adverse reaction [Kaufman S., personal communication]. One reported case involved a 74-year-old man on ranitidine 300 mg daily for 3 years who developed Coombs’-positive anemia with a hemoglobin of 10 mg/dl and a reticulocyte count of 15% that resolved after dechallenge. Another case reported by Pixley et al.  in 1989 occurred in a 59-year-old man with a history of alcohol abuse and polyarteritis nodosa who, after 29 days of intravenous ranitidine 100 mg daily, first developed direct Coombs’-positive anemia that resolved after stopping ranitidine and recurred on rechallenge. To determine how commonly clinically apparent autoimmune hemolytic anemia is associated with antecedent ranitidine use in an ambulatorycare population, we undertook an investigation, in two parts, using the automated medical and pharmacy records of the Harvard Community Health Plan (HCHP). The timely assessment of a possible association between ranitidine and autoimmune hemolytic anemia was made possible by the fully computerized nature of the HCHP clinical database and demonstrated the utility of such databases to perform rapid postmarketing studies of unintended drug effects.
PATIENTS AND METHODS
The HCHP is currently the largest health maintenance organization in New England. By the end of this study, the HCHP maintained a computerized ambulatory medical record system for 11 clinical centers which contained information on approximately 750,000 past and present members. Computer-Stored-Ambulatory-Record (COSTAR)-coded diagnoses for essentially all ambulatory encounters were available through the computerized medical records of members who received care at these 11 centers. Most diagnoses made in emergency rooms and during hospital admissions were ordinarily captured in the ambulatory record. Additionally, laboratory test results were accessible by computer for members at these 11 clinical centers plus another two centers without computerized medical records. Since April 1988, pharmacy records for all clinical centers have also been computerized and about 90% of HCHP members have had prescription drug coverage provided by the health plan. In the first part of this study, the HCHP automated pharmacy records were searched for all individuals who received ranitidine dispensings between April 1, 1988 and September 30,
1990. The automated medical records were also searched for all diagnoses of hemolytic anemia (COSTAR code 4114) or a positive direct or indirect Coombs’ test (COSTAR codes Al06 or A107, respectively) that were recorded in the same time period. Unspecified anemia was not included in the list of screening diagnoses because of the likelihood that this diagnosis would include a large number of cases related to blood loss from peptic ulcer disease and other related indications for therapy with ranitidine. The full prescription records of individuals who received ranitidine and had a screening COSTAR diagnosis or positive Coombs’ test code were reviewed to find those patients with ranitidine prescriptions before the diagnosis of hemolytic anemia or a positive test result. The full medical records of cases with antecedent ranitidine prescriptions were reviewed by two of the authors (P.C. and J.G.). In the second part of the study, additional analyses were conducted in the majority of HCHP members who received care at the 11 health centers with computerized medical records and had prescription drug coverage during the entire observation period. All individuals in this subpopulation who received a ranitidine prescription between April 1988 and September 1990 were identified. Ranitidine exposure windows (or courses) for these individuals were constructed by defining the beginning of a window as the date of the first dispensing. The window was closed 30 days later unless the first dispensing was followed within 60 days by another dispensing. In this case, the exposure window was extended up to this subsequent dispensing. An exposure window continued to be extended in the same manner until no additional ranitidine dispensings occurred in the following 60 days. The exposure window was then closed 30 days after the last dispensing. All exposure windows with beginning dates before April 1, 1988 or ending dates after September 3 1, 1990 were omitted from the analyses. The number of blood counts and the distribution of hematocrits obtained during all ranitidine courses was then determined from computerized test files. Among individuals with hematocrits measured in ranitidine windows, the number of Coombs’ tests ordered in exposure windows and their results were also ascertained within strata defined by ranges of hematocrits. Individuals with diagnoses of hemolytic anemia rendered during exposure were also identified.
Autoimmune Hemolytic Anemia
To evaluate the causes of anemia detected in ranitidine windows, the full-text medical records of all cases with minimum hematocrits in the range of lo-19% and of randomly selected samples comprising half of the cases with hematocrits between 20 and 29% and 5% of cases with milder anemia between 30 and 39% were reviewed by one of the authors (P.C.). The review covered all encounter records that occurred from 1 month before the first ranitidine window opened to the month after the last exposure window was closed.
Between April 1, 1988 and September 30, 1990, 38,686 ranitidine prescriptions were dispensed to 12,054 individuals (mean 3.2 dispensings/person) at all 13 health centers; 30,368 prescriptions were filled by 9179 members (mean 3.3 dispensings/person) receiving care at those 11 centers with computerized medical records. In the subpopulation of 7582 members with prescription drug coverage during the entire study period and who received care at one of the 11 centers with computerized records, 25,177 prescriptions of ranitidine (mean 3.3 dispensings/person) were dispensed from an HCHP pharmacy. Nine individuals received ranitidine at some time during the 30-month study period and also had a diagnosis of hemolytic anemia or an abnormal Coombs’ test. Four of these patients received their first diagnosis or had an abnormal Coombs’ test noted in the encounter records after receiving a prescription for ranitidine. Review of the full-text medical records of these four patients indicated that only one patient had actually had an abnormal Coombs’ test or hemolytic anemia. This individual was a 31-year-old woman in her first trimester of pregnancy who received 45 tablets of ranitidine to be taken at bedtime for gastritis, two and one-half months before an abnormal direct Coombs’ test result was obtained. She had not been prescribed any other medications in this interval. The woman was asymptomatic and the test was apparently obtained as part of her routine prenatal care. Her blood group and type were 0, D and DU negative; and her hematocrit was 36.5% (normal 38.0-47.9%). No further evaluation was pursued and the association between ranitidine and this patient’s Coombs’ test could not be confirmed.
Assuming that the patient population of all 13 clinical centers constituted the study population, the observed risk of diagnosed hemolytic anemia associated with ranitidine was 0 eases/12,054 exposed persons. The 95% upper confidence bound derived from the Poisson distribution was 3.7 eases/12,054 exposed persons or 3.1/10,000. Among patients with full prescription coverage at health centers with automated records, the degree of surveillance for anemia and hemolysis during treatment with ranitidine was assessed. There were 12,375 ranitidine windows representing 675,431 days of exposure in 7182 ranitidine users. The median number of windows per individual was one (interquartile range: l-2) and the median duration of a ranitidine exposure window was 30 days (interquartile range: 30-56 days). During windows of ranitidine exposure, 69 16 blood counts were obtained on 3729 persons. Slightly more than half of the recipients of ranitidine had at least one blood count during their ranitidine exposure windows. Blood counts were obtained in 30% of exposure windows. The mean number of blood counts obtained per window was 0.6 (f 1.7 SD, range O-75). There were 1623 (43.5%) of 3729 ranitidine users who had hematocrits less than 40% in exposure windows. The full-text medical records were reviewed of all three individuals with hematocrits between 10 and 19%, 59 randomly selected (representing 50% of all) individuals with hematocrits between 20 and 29% and 71 randomly selected (about 5% of all) with hematocrits between 30 and 39% to ascertain the causes of their anemia. Plausible explanations for the observed anemia were found in two-thirds of the cases (see Table 1). As expected, a greater proportion of cases with more severe anemia had alternative explanations for their anemia found in their medical records. Causes for anemia were discernible in 97% who had minimum hematocrits below 30% in a ranitidine window and 42% of individuals with hematocrits between 30 and 39%. In many individuals of this latter group the temporal sequence of hematocrit results was inconsistent with hemolysis associated with recent exposure to ranitidine; and in no cases was the temporal relationship between exposure and reduced hematocrit highly suggestive of hemolysis. Acute and chronic blood loss most often attributable to gastrointestinal bleeding was felt to be responsible for
W. C~oo ef al.
Table 1. Primary cause of anemia identified during ranitidine exposure in the Harvard Community Health Plan, 4/l/88 to 9131190 Minimum hematocrit 10-19 20-29 30-39 Total
Acute and chronic blood loss*
0 21 10 31
2 23 7 32
AIDS/ HIV? 0 8
Chronic renal failure
Vitamin B12 deficiency
1 1 2 4
0 0 1 1
0 2 41 43
Sample total (Population total)
3 (3) 59 (119) 71 (1501) 133 (1623)
*Includes gastrointestinal, gynecological and surgical blood loss, as well as two cases of iron-deficiency anemia without an identified source of bleeding. tAquired Immune Deficiency Syndrome/Human Immunodeficiency Virus. $One case each of idiopathic thrombocytosis, liver failure, idiopathic retroperitoneal fibrosis and autoimmune hemolytic anemia (not attributable to ranitidine). $One case each of rheumatoid arthritis, Waldenstrom’s macroglobulinemia, temporal arteritis, severe congestive heart failure, pregnancy, anorexia secondary to gastric dysmotility, diabetes mellitus; and two cases of acute appendicitis that went to surgery.
23% of all cases of anemia and 34% of cases with hematocrits below 30%. Of those individuals with hematocrits below 40%, six (0.4%) had a Coombs’ test in an exposure window. All six Coombs’ tests were obtained in persons with minimum hematocrits between 30 and 39. One additional Coombs’ test was obtained on a person with a minimum hematocrit of 40-49. None of the Coombs’ tests were abnormal. No persons received their first hemolytic anemia diagnosis in a ranitidine exposure window. The incidence rate of autoimmune hemolysis associated with ranitidine based on these ranitidine exposure windows was estimated to be 0 eases/675,431 person days of exposure with an upper 95% confidence bound of 5.5 cases/l,OOO,OOOperson days of exposure.
This study suggests that clinically recognized autoimmune hemolytic anemia associated with ranitidine is rare in the ambulatory care population. Only one instance of a positive direct Coombs’ test with mild anemia after a ranitidine prescription was found in an asymptomatic patient. Her work-up was insufficient to establish a direct relationship between her use of ranitidine and the abnormal Coombs’ test. Additionally, there is no confirmation that she actually took the medication just prior to the abnormal test result. No cases of symptomatic hemolysis associated with ranitidine use were found in either of the two parts of this investigation. The risk of autoimmune hemolytic anemia observed among all individuals was 0112,054 with an upper 95% confidence bound of 3.1/10,000. The rate of hemolytic anemia observed in patients with more complete data was 0 eases/675,431 person-days of exposure
(95% upper confidence bound: 5.5/1,000,000 days). The main focus of this investigation was the occurrence of clinically apparent (as asymptomatic) autoimmune opposed to hemolytic anemia because we believed that this outcome was most relevant to practitioners and policy makers and also because surveillance for mild hemolytic anemia was bound to be incomplete. Consequently, this evaluation cannot provide an estimate of the total frequency of ranitidine-associated anemia or autoimmune hemolysis. This observational dataset, however, provided a much larger experience than would be feasible in the context of a conventional clinical evaluation. Although surveillance for anemia was incomplete, a majority of individuals did have hematocrits measured during ranitidine exposure and we presume that these more likely would have been obtained for individuals who had symptomatic anemias. A satisfactory explanation was identified for nearly all severe anemias, strengthening our belief that no serious anemias caused by hemolysis were missed. The approach utilized in our study is sufficient, therefore, to exclude the common occurrence of clinically important hemolytic reactions to ranitidine. This study also illustrates how this and other similar databases can be used to follow up spontaneous reports of possible new unintended drug effects, particularly when the goal is to obtain a rapid estimate of the potential frequency of suspected associations.
A~knon,ledgemenrs-The authors thank Emily Cain, Claire Canning, and James Livingston for assistance with data collection and processing. This research was supported in part by the U.S. Food and Drug Administration (FDU 0003 15).
Autoimmune Hemolytic Anemia
REFERENCES Pixley JS, MacKintosh FR, Sahr EA, Zanjani ED. Mechanism of ranitidine associated anemia. Am J Med Sci 1989; 369-371. Gafter U, Komlos L, Weinstein T, Zevin D, Levi J. Thrombocytopenia, eosinophilia, and ranitidine. Ann Intern Med 1987; 106: 477. Amos RJ, Kirk B, Amess JAL, Jones AL, Hinds CJ. Bone marrow hypoplasia during intensive care: bone marrow culture studies implicating ranitidine in the suppression of haemopoiesis. Human Toxic01 1987; 6: 503-506. List AF, Beaird DH, Kummet T. Ranitidine-induced
granulocytopenia: recurrence with cimetidine administration. Ann Intern Med 1988; 108: 566-567. 5. Bader A, Carrigan T, Gopalswamy N, Trulxsch DV. Reversible hematologic suppression during ranitidine treatment. DICP 1989; 23: 508-509. 6. Aymard JP, Aymard B, Netter P, Bannwarth B, Trechot P, Streiff F. Haematological adverse effects of histamine H,-receptor antagonists. Med Toxicol Adverse Drag Exp 1988; 3: 430-448. 7. McGuigan JE. A consideration of the adverse effects of cimetidine. Gastroenterology 1981; 80: 181-192. 8. Rotoli B, Formisano S, Alfinito F. Autoimmune haemolytic anaemia associated with cimetidine [Letter]. Lancet 1979; 2: 583.