NONNARCOTIC ANALGESICS ARE

Transcription

1 ORIGINAL CONTRIBUTION Analgesic Use and Renal Function in Men Kathryn M. Rexrode, MD Julie E. Buring, ScD Robert J. Glynn, ScD Meir J. Stampfer, MD Linda D. Youngman, PhD J. Michael Gaziano, MD, MPH Context Several case-control studies suggest an association between analgesic use and increased risk of chronic renal disease, but few cohort studies have examined this association. Objective To determine whether analgesic use is associated with risk of renal dysfunction. Design and Setting Cohort study of analgesic use data from the Physicians Health Study, which lasted 14 years from September 1982 to December 1995 with annual follow-up. Participants A total of initially healthy men who provided blood samples and self-report of analgesic use. Main Outcome Measures Elevated creatinine level defined as 1.5 mg/dl (133 µmol/l) or higher and a reduced creatinine clearance defined as 55 ml/min (0.9 ml/s) or less, and self-reported use of acetaminophen, aspirin, and other nonsteroidal antiinflammatory drugs (never [ 12 pills]; pills; pills; and 2500 pills). Results A total of 460 men had elevated creatinine levels (4.2%) and 1258 had reduced creatinine clearance (11.4%). Mean creatinine levels and creatinine clearances were similar among men who did not use analgesics and those who did, even at total intakes of 2500 or more pills. In multivariable analyses adjusted for age; body mass index; history of hypertension, elevated cholesterol, and diabetes; occurrence of cardiovascular disease; physical activity; and use of other analgesics, the relative risks of elevated creatinine level associated with intake of 2500 or more pills were 0.83 (95% confidence interval [CI], ; P for trend=.05) for acetaminophen, 0.98 (95% CI, ; P for trend=.96) for aspirin, and 1.07 (95% CI, ; P for trend=.86) for other nonsteroidal anti-inflammatory drugs. No association was observed between analgesic use and reduced creatinine clearance. Conclusions Moderate analgesic use in this cohort study of initially healthy men was not associated with increased risk of renal dysfunction. JAMA. 2001;286: NONNARCOTIC ANALGESICS ARE among the most widely used medications and are used at least weekly by one quarter of US adults. 1-3 High acute doses of analgesics, particularly nonsteroidal antiinflammatory drugs (), have been implicated as causes of acute renal failure, 4-7 which is most commonly seen among patients with impaired renal function. It has also been suggested that prolonged use of various analgesics may cause progressive renal damage ultimately leading to chronic kidney failure. The earliest study linking chronic analgesic use and chronic renal dysfunction was reported almost 50 years ago. 8 In the ensuing half century, numerous case reports, several casecontrol studies, and 2 small cohort studies have raised the possibility of an association between use of various analgesics and chronic renal disease. With respect to different types of analgesics, the association between phenacetin and kidney disease has been quite strong In the United States, phenacetin was banned from the market in Whether other analgesics are implicated in renal dysfunction is less clear, with some studies showing no association between regular use of analgesics such as acetaminophen, aspirin, and other and chronic renal dysfunction 9,13 and other studies showing increased risk. 10,11,14-16 A recent casecontrol study, for example, reported a 2-fold increased risk of end-stage renal disease among individuals with lifetime use of more than 1000 acetaminophen pills and an 8-fold increased risk among those with a lifetime cumulative dose of more than 5000 NSAID pills. 16 These case-control studies, however, have significant methodologic limitations, including difficulty in establishing the time order of exposure, confounding by indication, selection Author Affiliations: Division of Preventive Medicine, Channing Laboratory, Department of Medicine, Brigham and Women s Hospital (Drs Rexrode, Stampfer, Gaziano, and Glynn), Department of Ambulatory Care and Prevention, Harvard Medical School (Dr Buring), Department of Epidemiology, Harvard School of Public Health (Drs Stampfer and Buring), and Department of Biostatistics, Harvard School of Public Health (Dr Glynn), Boston, Mass; Office of Research, US Food and Drug Administration, Laurel, Md (Dr Youngman); and Massachusetts Veterans Epidemiology Research and Information Center and VA Boston Healthcare System, Boston, Mass (Dr Gaziano). Corresponding Author and Reprints: J. Michael Gaziano, MD, MPH, Division of Preventive Medicine, Brigham and Women s Hospital, 900 Commonwealth Ave, Boston, MA ( American Medical Association. All rights reserved. (Reprinted) JAMA, July 18, 2001 Vol 286, No

2 ANALGESIC USE AND RENAL FUNCTION IN MEN bias, and recall bias. 17,18 Given these methodologic limitations, the existing evidence is insufficient to support a causal association. One way of minimizing confounding by indication and recall bias inherent in the case-control design is to examine the effect of long-term analgesic use on earlier, preclinical renal disease outcomes in a large cohort study. The Physicians Health Study(PHS) cohort provided such an opportunity. In this group of more than initially healthy men who provided blood samples, we evaluated the association between retrospective selfreported intake of acetaminophen, aspirin, and other and renal function, as measured by creatinine levels and creatinine clearance. METHODS The design and methods of the PHS are described in detail elsewhere. 19 In brief, the study consisted of apparently healthy male physicians without prior history of cardiovascular disease, cancer (except nonmelanoma skin cancer), current liver disease or renal dysfunction (defined as renal failure or insufficiency), or other major illnesses. They were randomly assigned to 1 of 4 groups: aspirin (325 mg on alternate days), beta carotene (50 mg on alternate days), both, or neither to test whether aspirin and/or beta carotene prevent heart disease or cancer. In January 1988, the study s aspirin arm was terminated early based primarily upon a statistically extreme 44% reduction in the occurrence of a first myocardial infarction among those randomized to aspirin. The trial s beta carotene arm was continued until its scheduled completion in December Every 6 months for the first year and annually thereafter participants were sent follow-up questionnaires. Information was collected regarding study compliance; personal characteristics, including weight and height; medical history, including history of hypertension, elevated cholesterol, and diabetes; health behaviors, including smoking status (never, past, current), frequency of physical activity (none, monthly, 1-4 times per week, 5 times per week), and use of alcohol (rarely/ never, monthly, weekly, daily, or more); and the occurrence of cardiovascular disease and any other relevant medical events during the study period. Exposure Information At the end of the randomized trial, the physicians completed a detailed retrospective history of analgesic use over the 14-year study period. Participants were asked about their use of 8 categories of analgesics: (1) aspirin and buffered aspirin alone (excluding study-drug aspirin), (2) combination pills containing primarily aspirin, (3) acetaminophen alone, (4) combinations containing primarily acetaminophen, (5) combinations containing both acetaminophen and aspirin, (6) other, (7) other pain relief agents used as single agents, and (8) other pain relief combinations. For each category, participants were asked if they had taken this type of medication 12 or more times since enrolling in the PHS in Those answering yes were asked to indicate how many days per month and how many pills per day these medicines were taken on average in the preceding year. The participants were also asked if this pattern had been constant since enrolling in the PHS. Those answering no were asked to record the number of years for which this had been the pattern of use and to describe their earlier pattern. use as part of the PHS was calculated by using the compliance data reported on each participant s annual questionnaire and his assignment to aspirin or placebo. After the aspirin arm of the PHS was terminated, participants could elect to continue receiving study aspirin; their reported compliance with this aspirin was also included. Total aspirin use was the sum of aspirin provided as part of the PHS and self-reported aspirin use on the analgesic-history questionnaire. Based on this information, analgesic consumption was categorized as acetaminophen (acetaminophen and acetaminophen-containing compounds), aspirin (aspirin and aspirin-containing compounds), and other. We did not include in these analyses the use of combination pills containing aspirin and acetaminophen nor the use of other analgesics, because the number of users of these was small (237 for aspirin and acetaminophen, 202 for other analgesic and other analgesic combinations). For each category of analgesic use, the number of pills per year was recorded and the cumulative number of pills over the study period was calculated by summing the reported use over the years of the study. In order to simplify comparisons with prior studies, 16 we created 4 categories of analgesic use: never ( 12 pills during the 14-year study period), 12 to 1499 pills, 1500 to 2499 pills, and 2500 or more pills. Very high use ( 7000 pills over the study period) was also examined in secondary analyses. To assess internal validity, information on analgesic use was obtained using a structured telephone interview from a sample of 910 randomized PHS participants without a history of renal or other major diseases 1 to 2 years after they completed the written analgesicuse questionnaire. These participants were asked about their use of over-thecounter and prescription analgesics, the average number of days per month and pills per day these medications were used in the previous year and in the 2 prior 5-year periods, and use of higher doses of analgesics for at least 1 month or more. Cumulative analgesic use assessed by telephone interviews was compared with that reported on the written questionnaire. The correlation coefficients of analgesic use during an identical follow-up period from 1996 until the return of the written questionnaire were 0.76 for acetaminophen, 0.70 for aspirin, and 0.45 for other. For the period from 1988 to 1995 the correlations between the 2 methods were 0.67 for acetaminophen, 0.40 for aspirin, and 0.46 for other. The correlations for analgesic use are of similar magnitude to correlations over several years for variables widely used in epidemiologic studies such as blood pressure 20 and serum cholesterol. 21 In comparing the respondents in the highest quartiles on both questionnaires, 316 JAMA, July 18, 2001 Vol 286, No. 3 (Reprinted) 2001 American Medical Association. All rights reserved.

3 ANALGESIC USE AND RENAL FUNCTION IN MEN there was a concordance of 51.8% for acetaminophen, 63.8% for aspirin, and 48.4% for other. Among those in the highest quartile of analgesic use on the written questionnaire, the concordance for being in the 2 highest quartiles on the telephone questionnaire was 75.3% for acetaminophen, 86.2% for aspirin, and 64.8% for other. Blood Collection and Analysis Blood kits with vacuum tubes containing EDTA, complete instructions for blood draws, and coldpacks were mailed to all enrolled physicians between December 1995 and October Each was asked to have a blood sample drawn into the vacuum tube and to return the sample in the coldpack by prepaid overnight courier. Upon receipt, each sample was centrifuged, divided into aliquots, and stored at 82 C. None of the samples inadvertently thawed during storage. Follow-up samples were received from participants; creatinine analyses were available for of these participants. Blood samples were analyzed at Oxford University, England, using an automated Jaffe rate method on a SYNCHRON LX20 autoanalyzer (Beckman Coulter, Fullerton, Calif) for quantification of creatinine. Plasma creatinine is stable in chilled next-day whole blood samples preserved with EDTA. 22 To assess quality control, splits were submitted; the coefficient of variation for these samples was 7.07%. The difference in means between the study samples and the repeat quality control samples was mg/dl (SD, 0.67). Intrabatch coefficients of variation on internal quality control runs were 1.4% to 3.6%. Outcomes The primary outcomes of interest were an elevated creatinine level, defined as 1.5 mg/dl (133 µmol/l) or higher at the time of the follow-up blood collection, and a reduced creatinine clearance, defined as 55 ml/min (0.9 ml/s) or less. Creatinine clearance was calculated using the Cockcroft-Gault equation: CrCl= [(140 age) (weight in kilograms)] / [72 (creatinine in mg/dl)]. 23 Table 1. Mean Creatinine Level and Creatinine Clearance Among Participants in the Physicians Health Study Over a 14-Year Period, Stratified by Age at Baseline Baseline Age Categories, y No. Mean (SD) Creatinine Level, mg/dl* (n = ) Mean (SD) Creatinine Clearance, ml/min (n = ) (0.21) 95.6 (22.8) (0.24) 82.1 (20.7) (0.28) 64.6 (17.3) (0.35) 47.5 (13.6) P for trend *To convert to µmol/l, multiply by Creatinine data were available for participants. To avoid recall bias and confounding by indication, we excluded 71 participants who reported renal insufficiency or renal failure prior to the completion of the analgesic assessment questionnaire, and 3 participants with creatinine levels greater than 5.0 mg/dl (442 µmol/l) at the time of creatinine measurement, leaving participants for analysis. Statistical Analyses Characteristics of participants with elevated creatinine level and reduced creatinine clearance were compared with t test or 2 test. Comparisons of mean creatinine level and creatinine clearance were computed by the general linear models procedure (SAS version 6.12, SAS Institute, Cary, NC), adjusted for age. These linear models also provided estimates of statistical significance for categories of analgesic intake controlling for age. Multiple logistic regression analysis was performed to adjust simultaneously for significant risk factors for elevated creatinine levels or reduced creatinine clearance, including age; body mass index (BMI); history of hypertension, elevated cholesterol, or diabetes at baseline or during the study period; occurrence of cardiovascular disease during the study period; physical activity; and other analgesic use. For each relative risk (RR), 2-tailed P values and 95% confidence intervals (CIs) were calculated. Values of P for trend were calculated across categories. Logistic regression models were validated using the method of Hosmer and Lemeshow, 24 and there was no evidence of a lack of fit according to the goodness-of-fit statistic. Because of concerns about residual confounding by BMI, analyses using creatinine clearance were additionally stratified by BMI categories (BMI 30 and 30 kg/m 2 ). Subgroup analyses comparing men aged 60 years and older vs those younger than 60 years, as well as those with a history of hypertension and/or diabetes at baseline, were also performed. RESULTS The mean creatinine levels and creatinine clearance of the men with no prior history of renal failure or insufficiency for whom follow-up creatinine measures were available are presented by age categories in TABLE 1. As expected, age was strongly associated with both outcomes. In blood samples collected an average of 14 years after study enrollment, mean creatinine level was 19% higher and mean creatinine clearance was 50% lower among men aged 70 to 84 years at baseline than among men aged 40 to 49 years. Due to the profound, nonlinear effect of age on renal function, all subsequent data have been adjusted for age by 5-year age categories. The age-adjusted baseline characteristics of the study cohort according to creatinine levels and creatinine clearance categories are presented in TABLE 2. Elevated creatinine levels, defined as 1.5 mg/dl (133 µmol/l) or greater, was noted in 460 participants (4.2% of the study cohort). Participants with elevated creatinine levels were significantly older, had higher BMIs, and were more likely to have had a history of hypertension, elevated cholesterol, or the occurrence of cardiovascular disease during the study. Mean ageadjusted use of acetaminophen, aspirin, and other did not differ signifi American Medical Association. All rights reserved. (Reprinted) JAMA, July 18, 2001 Vol 286, No

4 ANALGESIC USE AND RENAL FUNCTION IN MEN cantly between those with and without elevated creatinine level. Reduced creatinine clearance was present in 1258 participants (11.4%). Men with reduced creatinine clearance were significantly older and leaner and were more likely to have had a history of hypertension at baseline and the occurrence of cardiovascular disease during the study. Mean ageadjusted use of acetaminophen was lower in men with reduced creatinine clearance while age-adjusted means did not differ for aspirin or other NSAID use. We examined mean creatinine levels and creatinine clearance according to 4 predetermined categories of analgesic exposure over the 14-year follow-up (TABLE 3). Compared with nonusers, men who used 12 to 2499 acetaminophen pills had lower mean creatinine levels, although the absolute differences were small (0.01 mg/dl to 0.02 mg/dl). For users of aspirin and other, mean creatinine level did not differ significantly across analgesic use categories. Mean creatinine clearance increased significantly across increasing use categories of both acetaminophen and other, but did not differ significantly across aspirin use categories. Because weight is part of the formula for calculating creatinine clearance and analgesic use tends to increase with increasing weight, we conducted analyses stratified by BMI to evaluate possible confounding by weight. Stratifying by BMI minimized confounding by weight and attenuated the protective inverse trends for acetaminophen and other. We performed multivariable analyses that adjusted for potential renal risk factors (age; BMI; history of hypertension, elevated cholesterol, and diabetes; occurrence of cardiovascular disease during the study; and physical activity) as well as for use of other analgesics (TABLE 4). use was associated with a slightly de- Table 2. Baseline Characteristics and Cumulative Analgesic Intake Among Men (N = ) in the Physicians Health Study According to Creatinine Level and Creatinine Clearance in 1997, Adjusted for Age* Baseline Characteristics Creatinine Level 1.5 mg/dl (n = ) Creatinine Level 1.5 mg/dl (n = 460) P Value Creatinine Clearance 55 ml/min (n = 9774) Creatinine Clearance 55 ml/min (n = 1258) P Value Age, mean (SD), y 53.1 (8.4) 58.6 (9.9) (7.6) 63.9 (7.9).001 BMI, mean (SE), kg/m (0.03) 25.9 (0.16) (0.03) 23.9 (0.11).001 Hypertension, % Never Baseline Follow-up Elevated cholesterol, % Never Baseline Follow-up Diabetes, % Never Baseline Follow-up Cardiovascular disease, % Smoking, % Never Past Current Alcohol intake, % Rarely/never Monthly/weekly Daily or more Physical activity, % None time/mo-4 times/wk times/wk Mean (SE) No. of pills (21.2) (103.2) (22.5) (72.1).03 Mean (SE) No. of pills (21.0) (102.3) (22.3) (71.2).40 Mean (SE) No. of pills (21.0) (101.8) (22.2) (70.0).49 *Analgesic intake was reported as cumulative number of pills reported taken between study entry in 1982 and Creatinine clearance was calculated using the equation of Cockcroft and Gault. 23 All values (except age) were adjusted for age at baseline (categories 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, and 70 years). BMI indicates body mass index;, nonsteroidal anti-inflammatory drugs; and SE, age adjusted standard error. To convert to µmol/l, multiply by P values from general linear model for continuous variables and Mantel-Haenszel 2 test using row mean score differences for categorical variables. 318 JAMA, July 18, 2001 Vol 286, No. 3 (Reprinted) 2001 American Medical Association. All rights reserved.

5 ANALGESIC USE AND RENAL FUNCTION IN MEN Table 3. Mean Creatinine Level and Creatinine Clearance by Analgesic Use Categories, for All Men and for Nonobese Men, Adjusted for Age* Cumulative Analgesic Intake, No. of Pills No. Mean Creatinine Level, mg/dl All Men Men With BMI 30 kg/m 2 P Value Mean Creatinine Clearance, ml/min P Value No. Mean Creatinine Clearance, ml/min P Value Never use ( 12) Referent 82.7 Referent Referent Never use ( 12) Referent 82.6 Referent Referent Other Never use ( 12) Referent 82.0 Referent Referent *Values adjusted for age at baseline (categories 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, and 70 years). Restricted to men with body mass index (BMI) 30 kg/m 2 at time of follow-up blood sample. Missing values: 1201 acetaminophen, 1181 aspirin, and 646 nonsteroidal anti-inflammatory drugs (). To convert to µmol/l, multiply by Estimated by general linear regression equation comparing the distribution of each characteristic to that of the reference category, the first category in each exposure group. Table 4. Multivariable Relative Risks of Elevated Creatinine Levels ( 1.5 mg/dl) in 1997, According to Categories of Cumulative Analgesic Exposure* Analgesic Categories, by No. of Pills Age-Adjusted RR (95% CI) Model 1 Model 2 None ( ) 0.71 ( ) 0.70 ( ) ( ) 0.73 ( ) 0.73 ( ) ( ) 0.77 ( ) 0.83 ( ) P for trend None ( ) 1.03 ( ) 0.98 ( ) ( ) 1.04 ( ) 1.03 ( ) ( ) 0.99 ( ) 0.98 ( ) P for trend None ( ) 0.94 ( ) 1.01 ( ) ( ) 1.10 ( ) 0.87 ( ) ( ) 0.91 ( ) 1.07 ( ) P for trend *RR indicates relative risk; CI, confidence interval; and, nonsteroidal anti-inflammatory drugs. Model 1: Adjusted for age (40-44, 45-49, 50-54, 55-59, 60-64, 65-69, and 70 years), body mass index ( 25, , and 30 kg/m 2 ); history of hypertension (none, baseline, follow-up), elevated cholesterol (none, baseline, followup), or diabetes (none, baseline, follow-up); cardiovascular disease during follow-up; and physical activity. Model 2: Adjusted for all factors in model 1, plus other analgesics used. Fewer than 12 pills ingested between study entry in 1982 and creased risk of elevated creatinine levels (P for trend=.05), and intake of 2500 or more acetaminophen pills over the study period was associated with an RR of elevated creatinine levels of 0.83 (95% CI, ). For aspirin, no significant trend for creatinine levels was observed (P for trend=.96), and intake of 2500 or more pills was associated with an RR of 0.98 (95% CI, ). For other, no significant trend was observed (P for trend=.86), and intake of more than 2500 pills was associated with an RR of 1.07 (95% CI, ). When we examined men who consumed 7000 or more pills of each analgesic during the study ( 500 pills/ year) the RRs of elevated creatinine levels were 0.51 (95% CI, ) for acetaminophen, 1.35 (95% CI, ) for aspirin, and 1.61 (95% CI, ) for other. Results did not differ when the 74 men with reported renal disease were included. To identify a possible association with analgesic use among older men, who might be most susceptible to renal dysfunction, we examined the association between analgesic use and risk of elevated creatinine levels among men aged 60 years or older at baseline. Even in this high-risk group, no significant trends of increased risk of elevated creatinine levels were observed for any of the analgesic use categories. When the 2422 men with a diagnosis of hypertension or diabetes at baseline were examined separately, the RRs of elevated creatinine levels with consumption of 2500 or more pills were 0.70 (95% CI, ) for acetaminophen, 0.57 (95% CI, ) for aspirin, and 0.63 (95% CI, ) for other. Multivariable models were also fitted to examine the association between analgesic exposure and reduced creatinine clearance while adjusting for potential confounders. No significantly increased risks were ob American Medical Association. All rights reserved. (Reprinted) JAMA, July 18, 2001 Vol 286, No

6 ANALGESIC USE AND RENAL FUNCTION IN MEN Table 5. Multivariable Relative Risk of Reduced Creatinine Clearance ( 55 ml/min) in 1997, According to Categories of Analgesic Exposure* Analgesic Categories, by No. of Pills Age-Adjusted RR (95% CI) Model 1 Model ( ) 0.73 ( ) 0.74 ( ) ( ) 0.86 ( ) 0.94 ( ) ( ) 0.84 ( ) 0.78 ( ) P for trend ( ) 1.18 ( ) 1.27 ( ) ( ) 1.25 ( ) 1.40 ( ) ( ) 1.26 ( ) 1.40 ( ) P for trend ( ) 1.00 ( ) 1.01 ( ) ( ) 1.08 ( ) 1.06 ( ) ( ) 0.92 ( ) 1.01 ( ) P for trend *RR indicates relative risk; CI, confidence interval; and, nonsteroidal anti-inflammatory drugs. See Table 4 footnote for variables in the adjusted analyses. Fewer than 12 pills ingested between study entry in 1982 and served for any of the 3 analgesic categories (TABLE 5). Results were similar for men aged 60 years or older, as well as for those with a history of hypertension or diabetes at baseline. Additional multivariable analyses were also performed (see online Tables A, B, C, D, and E at issues/v286n3/abs/joc02209.html), which include the 74 men with prior history of renal disease. COMMENT In this large cohort study of initially healthy male physicians followed up for an average of 14 years, we found no significant associations between the use of acetaminophen, aspirin, or other and either elevated creatinine levels or reduced creatinine clearance even among men who used 2500 or more pills (an average of 3 to 4 pills/ week) over the study period. Established predictors of renal dysfunction, including age and hypertension, were generally associated with elevated creatinine levels and reduced creatinine clearance. Diabetes rates tended to be higher among those with elevated creatinine levels and reduced creatinine clearance. Our study is the largest cohort analysis to date of the association between analgesic use and risk of renal dysfunction. Two small cohort studies 25,26 both found increased risks of renal dysfunction, mortality, and cancer associated with use of phenacetin. Data from both of these investigations, however, indicate that the analgesic users had more underlying illnesses than the nonusers, and these illnesses may have led both to increased analgesic use and to the wide range of outcomes identified in respective studies. 17 Our data contrast with findings from most of the earlier case-control studies. Since 1980, at least 8 case-control studies have examined the association between analgesic use and various measures of chronic renal failure, including end-stage renal disease, 11,13-16 renal papillary necrosis, 9 and first diagnosis of chronic renal disease. 10,27 Two of these 9,13 showed no clear associations, while the others found positive associations. Limitations in the case-control design make it particularly difficult to establish a causal relationship between analgesic use and renal dysfunction. Studies that included primarily patients with established renal failure cannot determine whether analgesic use predated renal failure or was a consequence of renal dysfunction. Once renal disease is discovered, patients are generally counseled to use acetaminophen for their analgesic needs instead of aspirin or other. In our study, the use of a preclinical end point, the collection of exposure data prior to assessment of renal function, and the exclusion of participants with a diagnosed history of renal disease should substantially minimize the likelihood of confounding by indication, selection bias, and recall bias. Measuring long-term use of over-thecounter medications is a challenging task. The structured questionnaire we used to assess analgesic consumption was similar to the methods used by Perneger et al 16 and other investigators in case-control studies that found significant increased risks of renal failure with analgesic exposures. Furthermore, we found moderate reliability and internal validity when comparing estimates of analgesic use obtained by written questionnaires with those obtained via structured telephone interviews, making it unlikely that measurement error explains the null association we observed. The large size of the PHS cohort made it possible to examine and detect even modest associations between analgesic use and the risk of elevated creatinine levels or reduced creatinine clearance. Furthermore, while most of the PHS participants reported moderate use of analgesics, this cohort study included more individuals reporting high use of analgesics ( 2500 pills) than any prior case-control study. Over the 14-year study period, 474 men reported high use of acetaminophen, 2360 reported high use of aspirin, and 816 reported high use of other. In contrast, the numbers of high users (cumulative lifetime use of 1000 pills) in the largest casecontrol study to date were 139 for acetaminophen, 173 for aspirin, and 52 for other. 16 Given the large number of high users in our cohort, and the relatively large risks suggested by casecontrol studies, 16 we should have been 320 JAMA, July 18, 2001 Vol 286, No. 3 (Reprinted) 2001 American Medical Association. All rights reserved.

7 ANALGESIC USE AND RENAL FUNCTION IN MEN able to detect an increased risk of renal dysfunction. Our study has several limitations. The PHS cohort is composed of relatively healthy men, most of whom are white. Thus, we cannot extend these results to women or minority populations. Retrospective assessment of analgesic use within an ongoing cohort study made it impossible to obtain exposure information from participants who died before the analgesic assessment. However, given that only 14 deaths with renal failure as a primary or contributing cause of death were confirmed over the 14-year study, this small loss of follow-up is unlikely to alter the results of our analyses. We used a retrospective assessment of analgesic exposure that was not designed to assess the short-term effects of high analgesic use but to examine the association with cumulative use that had been raised by prior studies. Error in measuring both analgesic use and creatinine would tend to minimize a relationship. In addition, our questionnaire assessed pill use rather than dose. However, data from our validation study indicate that most men used standard over-thecounter doses (50.7% of participants took 325-mg acetaminophen tablets, 78.6% took 325-mg aspirin tablets, and 70.3% took 200-mg ibuprofen tablets). The use of a preclinical end point is another possible limitation. While this has the advantage of avoiding recall bias and changes in medication based on knowledge of renal disease, it assumes that the relationship between analgesic use and renal disease is cumulative and linear, rather than requiring a high threshold effect. Furthermore, we specifically excluded individuals with a history of renal dysfunction to avoid problems of recall and indication bias, so we cannot directly examine the impact of analgesic exposure among individuals with preexisting renal impairment. However, we found no increased risk among higher-risk groups such as those aged 60 years or older or those with hypertension or diabetes at baseline, and our results did not differ when the 74 men with prior history of renal disease were included. Because the PHS tested the effects of aspirin, this cohort study included fewer never users of aspirin. Nevertheless, we should have been able to detect an adverse effect of this agent. Because analgesic use is so common, even small increases in the RR of renal dysfunction could have a significant impact on rates of renal disease in the United States. This study, however, provides reassuring evidence that there does not appear to be a strong association between chronic analgesic use and chronic renal dysfunction among a large cohort of men. The findings from prior studies suggesting increased risk of renal failure with high cumulative acetaminophen use may have been due to limitations inherent in the case-control design. While more research is needed, we believe that these findings provide reassurance to clinicians and patients that moderate analgesic use is unlikely to contribute to increased risk of renal dysfunction among individuals without a history of renal impairment. Author Contributions: Study concept and design: Rexrode, Buring, Gaziano. Acquisition of data: Rexrode, Buring, Youngman, Gaziano. Analysis and interpretation of data: Rexrode, Buring, Glynn, Stampfer, Gaziano. Drafting of the manuscript: Rexrode, Gaziano. Critical revision of the manuscript for important intellectual content: Rexrode, Buring, Glynn, Stampfer, Youngman, Gaziano. Statistical expertise: Glynn, Stampfer, Gaziano. Obtained funding: Buring, Gaziano. Administrative, technical, or material support: Rexrode, Stampfer, Youngman, Gaziano. Study supervision: Rexrode, Youngman, Gaziano. Funding/Support: This work was supported by grants CA-34944, CA-40360, HL-26490, and HL from the National Institutes of Health, and by a grant from McNeil Consumer Healthcare. Dr Gaziano initiated the project and approached McNeil Consumer Products for funding. The authors designed the questionnaire and collected, coded, and entered all data, which were stored at the Division of Preventive Medicine. The authors had complete control over the design, conduct, analysis, interpretation, and description of the study. Online Tables: For online Tables A-E, see the JAMA Web site at /abs/joc02209.html. Acknowledgment: We are indebted to Charles H. Hennekens, MD, the principal investigator of Physicians Health Study, the participants for their outstanding commitment and cooperation, the entire staff for their expert and unfailing assistance, and to Eleanor Danielson, MIA, Gail Borkowski, BA, Matt McGonagle, BS, and Bonnie Zapolin, BA. We thank Vadim Bubes, PhD, for programming support and P. J. Skerrett, MS, for assistance in preparing and editing the manuscript. REFERENCES 1. Chrischilles EA, Foley DJ, Wallace RB, et al. Use of medications by persons 65 and over. J Gerontol. 1992; 47:M137-M Giovannucci E, Rimm EB, Stampfer MJ, Colditz GA, Ascherio A, Willett WC. use and the risk for colorectal cancer and adenoma in male health professionals. Ann Intern Med. 1994;121: Giovannucci E, Egan KM, Hunter DJ, et al. and the risk of colorectal cancer in women. N Engl J Med. 1995;333: Blakely P, McDonald BR. Acute renal failure due to acetaminophen ingestion. J Am Soc Nephrol. 1995; 6: D Agati V. Does aspirin cause acute or chronic renal failure in experimental animals and in humans? Am J Kidney Dis. 1996;28:S24-S Perez Gutthann S, Garcia Rodriguez LA, Raiford DS, et al. Nonsteroidal anti-inflammatory drugs and the risk of hospitalization for acute renal failure. Arch Intern Med. 1996;156: Griffin MR, Yared A, Ray WA. Nonsteroidal antiinflammatory drugs and acute renal failure in elderly persons. Am J Epidemiol. 2000;151: Spuhler O, Zollinger HU. Die Chronisch- Interstitielle Nephritis. Z Klin Med. 1953;151: McCredie M, Stewart JH. Does paracetamol cause urothelial cancer or renal papillary necrosis? Nephron. 1988;49: Sandler DP, Smith JC, Weinberg CR, et al. Analgesic use and chronic renal disease. N Engl J Med. 1989; 320: Morlans M, Laporte JR, Vidal X, Cabeza D, Stolley PD. End-stage renal disease and non-narcotic analgesics. Br J Clin Pharmacol. 1990;30: Federal Register (October 5, 1983). 13. Murray TG, Stolley PD, Anthony JC, et al. Epidemiologic study of regular analgesic use and end-stage renal disease. Arch Intern Med. 1983;143: Pommer W, Bronder E, Greiser E, et al. Regular analgesic intake and the risk of end-stage renal failure. Am J Nephrol. 1989;9: Steenland NK, Thun MJ, Ferguson CW, Port FK. Occupational and other exposures associated with male end-stage renal disease. Am J Public Health. 1990; 80: Perneger TV, Whelton PK, Klag MJ. Risk of kidney failure associated with the use of acetaminophen, aspirin, and nonsteroidal anti-inflammatory drugs. N Engl J Med. 1994;331: Delzell E, Shapiro S. A review of epidemiologic studies of nonnarcotic analgesics and chronic renal disease. Medicine (Baltimore). 1998;77: McLaughlin JK, Lipworth L, Chow WH, Blot WJ. Analgesic use and chronic renal failure. Kidney Int. 1998;54: Findings from the aspirin component of the ongoing Physicians Health Study. N Engl J Med. 1988; 318: Rosner B, Hennekens CH, Kass EH, Miall WE. Agespecific correlation analysis of longitudinal blood pressure data. Am J Epidemiol. 1977;106: Shekelle RB, Shryock AM, Paul O, et al. Diet, serum cholesterol, and death from coronary heart disease. N Engl J Med. 1981;304: Youngman LD, Lyon V, Collins R, Peto R. Problems with mailed blood in large-scale epidemiological studies and methods of correction. FASEB J. 1993; 7: Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron. 1976; 16: Hosmer DW, Lemeshow S. A goodness-of-fit test for the multiple logistic regression model. Commun Stat. 1980;A10: Dubach UC, Rosner B, Sturmer T. An epidemiologic study of abuse of analgesic drugs. N Engl J Med. 1991;324: Elseviers MM, De Broe ME. A long-term prospective controlled study of analgesic abuse in Belgium. Kidney Int. 1995;48: Sandler DP, Burr FR, Weinberg CR. Nonsteroidal anti-inflammatory drugs and the risk for chronic renal disease. Ann Intern Med. 1991;115: American Medical Association. All rights reserved. (Reprinted) JAMA, July 18, 2001 Vol 286, No

8 Table A. Mean Creatinine Level and Creatinine Clearance by Analgesic Use Categories, for All Men and for Nonobese Men, Adjusted for Age, Including the 74 Men Who Developed Renal Disease Prior to Completing the Analgesic Assessment* Cumulative Analgesic Intake, No. of Pills No. Mean Creatinine Level, mg/dl All Men Men With BMI 30 kg/m 2 P Value Mean Creatinine Clearance, ml/min P Value No. Mean Creatinine Clearance, ml/min P Value Never use ( 12) Referent 82.4 Referent Referent Never use ( 12) Referent 82.0 Referent Referent Other Never use ( 12) Referent 81.7 Referent Referent *Values adjusted for age at baseline (categories 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, and 70 years). indicates nonsteroidal anti-inflammatory drugs. Restricted to men with body mass index (BMI) 30 kg/m 2 at time of follow-up blood sample. To convert to µmol/l, multiply by Estimated by general linear regression equation comparing the distribution of each characteristic to that of the reference category, the first category in each exposure group American Medical Association. All rights reserved.

9 Table B. Multivariable Relative Risks of Elevated Creatinine Levels ( 1.5 mg/dl) in 1997, According to Categories of Cumulative Analgesic Exposure, Including the 74 Men Who Developed Renal Disease Prior to Completing the Analgesic Assessment* Analgesic Categories, by No. of Pills Age-Adjusted RR (95% CI) Model 1 Model 2 None ( ) 0.74 ( ) 0.73 ( ) ( ) 0.74 ( ) 0.69 ( ) ( ) 0.88 ( ) 0.88 ( ) P for trend None ( ) 1.00 ( ) 0.95 ( ) ( ) 1.00 ( ) 0.97 ( ) ( ) 0.97 ( ) 0.93 ( ) P for trend None ( ) 0.94 ( ) 1.01 ( ) ( ) 1.02 ( ) 0.81 ( ) ( ) 0.89 ( ) 1.02 ( ) P for trend *RR indicates relative risk; CI, confidence interval; and, nonsteroidal anti-inflammatory drugs. Model 1: Adjusted for age (categories 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, and 70 years), body mass index ( 25, , and 30 kg/m 2 ); history of hypertension (none, baseline, follow-up), elevated cholesterol (none, baseline, follow-up); cardiovascular disease during follow-up; and physical activity. Model 2: Adjusted for all factors in model 1, plus other analgesics used. Fewer than 12 pills ingested between study entry in 1982 and American Medical Association. All rights reserved.

10 Table C. Multivariable Relative Risk of Reduced Creatinine Clearance ( 55 ml/min) in 1997, According to Categories of Analgesic Exposure, Including the 74 Men Who Developed Renal Disease Prior to Completing the Analgesic Assessment* Analgesic Categories, by No. of Pills Age-Adjusted RR (95% CI) Model 1 Model ( ) 0.73 ( ) 0.73 ( ) ( ) 0.81 ( ) 0.87 ( ) ( ) 0.92 ( ) 0.83 ( ) P for trend ( ) 1.08 ( ) 1.14 ( ) ( ) 1.11 ( ) 1.24 ( ) ( ) 1.13 ( ) 1.23 ( ) P for trend ( ) 0.96 ( ) 1.00 ( ) ( ) 1.04 ( ) 1.03 ( ) ( ) 0.90 ( ) 0.98 ( ) P for trend *RR indicates relative risk; CI, confidence interval; and, nonsteroidal anti-inflammatory drugs. See Table B footnote for variables in the adjusted analyses. Fewer than 12 pills ingested between study entry in 1982 and American Medical Association. All rights reserved.

11 Table D. Multivariable Relative Risks of Elevated Creatinine Levels ( 1.5 mg/dl) in 1997, According to Categories of Cumulative Analgesic Exposure* Analgesic Categories, by No. of Pills Age-Adjusted RR (95% CI) Model 1 Model ( ) 0.70 ( ) 0.70 ( ) ( ) 0.73 ( ) 0.73 ( ) ( ) 0.83 ( ) 0.93 ( ) ( ) 0.62 ( ) 0.51 ( ) P for trend ( ) 1.03 ( ) 0.99 ( ) ( ) 1.04 ( ) 1.04 ( ) ( ) 0.97 ( ) 0.95 ( ) ( ) 1.17 ( ) 1.35 ( ) P for trend ( ) 0.92 ( ) 1.00 ( ) ( ) 1.10 ( ) 0.86 ( ) ( ) 0.75 ( ) 0.78 ( ) ( ) 1.24 ( ) 1.61 ( ) P for trend *RR indicates relative risk; CI, confidence interval; and, nonsteroidal anti-inflammatory drugs. See Table B footnote for variables in the adjusted analyses. Fewer than 12 pills ingested between study entry in 1982 and American Medical Association. All rights reserved.

12 Table E. Multivariable Relative Risk of Reduced Creatinine Clearance ( 55 ml/min) in 1997, According to Categories of Cumulative Cumulative Analgesic Exposure* Analgesic Categories, by No. of Pills Age-Adjusted RR (95% CI) Model 1 Model ( ) 0.72 ( ) 0.73 ( ) ( ) 0.82 ( ) 0.92 ( ) ( ) 0.95 ( ) 0.97 ( ) ( ) 0.65 ( ) 0.51 ( ) P for trend ( ) 1.21 ( ) 1.33 ( ) ( ) 1.25 ( ) 1.44 ( ) ( ) 1.26 ( ) 1.43 ( ) ( ) 1.13 ( ) 1.35 ( ) P for trend ( ) 0.95 ( ) 0.98 ( ) ( ) 1.08 ( ) 1.04 ( ) ( ) 0.80 ( ) 0.79 ( ) ( ) 1.15 ( ) 1.35 ( ) P for trend *RR indicates relative risk; CI, confidence interval; and, nonsteroidal anti-inflammatory drugs. See Table B footnote for variables in the adjusted analyses. Fewer than 12 pills ingested between study entry in 1982 and American Medical Association. All rights reserved.