Acetaminophen (paracetamol; N-acetyl-p-aminophenol), a widely used antipyretic and analgesic, is the active metabolite of phenacetin. 1 Phenacetin use has been associated with an increased risk of renal cell carcinoma in a number of case-control studies, 2–5 whereas an association between acetaminophen use and renal cancer has been suggested in some 4,6 but not in other reports. 5,7 Phenacetin use has also been associated with an increased risk of bladder cancer in most 4,8,9 but not all 10 case-control studies; in contrast, acetaminophen use has generally not been found to increase the risk of bladder cancer. 8,9 Most of these studies of the risk of urothelial cancers among users of phenacetin and acetaminophen have relied on interviews or questionnaires to ascertain drug exposure.
Our group previously studied the risk of renal and bladder cancers related to acetaminophen exposure in a case-control study using data from Group Health Cooperative of Puget Sound, a large U.S. health maintenance organization that maintains computer records of its members’ prescriptions and hospitalizations. 11 We found an increased risk of renal cancer among subjects who filled 40 or more prescriptions for acetaminophen or acetaminophen-containing products (OR 2.6, 95% CI 1.1–6.0), but the risk of bladder cancer with heavy acetaminophen exposure was only slightly elevated (OR 1.3; 95% CI 0.6–2.8). To evaluate whether these results could be confirmed in another large, automated database in which accurate prescription information and outcome data are routinely recorded, we conducted a similar nested case-control study within the U.K. General Practice Research Database. 12,13 We also evaluated the effect of non-steroidal anti-inflammatory drugs, which are not known to increase the risk of either bladder or renal cancer.
Subjects and Methods
The base population comprised subjects age 50 and older with data recorded in the General Practice Research Database.
Selection of Cases
Using an automated search strategy, we identified subjects age 50 years or older with a first-time recorded diagnosis of renal cancer in 1995–1998 (N = 187) or bladder cancer in 1995 (N = 243). We reviewed the potential cases’ computer profiles without information on drug exposures and excluded those who were first diagnosed at death (23 renal, 14 bladder), had recurrent disease (6 renal, 28 bladder), or had a benign tumor, an uncertain diagnosis, or another cancer (29 renal, 12 bladder). In addition, we distinguished 20 potential renal cancer cases whose tumors had transitional cell histology or were located in the renal pelvis (which we designate “renal pelvis cancer”) and analyzed these cases separately from the others with renal cancer.
Selection of Controls
We matched up to four controls to each case. The matching factors were year of birth (within 1 year), sex, general practice, year of first recorded prescription information in the GPRD (within 1 year), and index date. We applied the same exclusions to the controls as to the cases (other than those relating to the diagnosis of renal or bladder cancer).
For the primary analysis of acetaminophen exposure, we included prescriptions for acetaminophen itself as well as combinations of acetaminophen with codeine, dihydrocodeine, propoxyphene, or metoclopramide (“acetaminophen-containing” medicines). For each subject, we tabulated the number of prescriptions recorded during various periods before their index date (<1 year; 1 to <2 years; 2 to <3 years; 3 to <5 years; and 5 years or more). Because medicines taken within a year before the diagnosis of renal or bladder cancer are unlikely to be etiologically relevant, we restricted our definition of exposure to acetaminophen prescriptions recorded before the year before the index date. Also, to avoid potential bias attributable to variable durations of recorded prescription information (“left censoring”), the exposure definition excluded prescriptions recorded before 5 years before the index date.
For the analysis of non-steroidal anti-inflammatory drug use, we classified exposure as the number of prescriptions for the same periods before the index date as described above for acetaminophen. We included all non-steroidal anti-inflammatory drugs except aspirin in this analysis.
All cases and controls had at least 3 years of recorded prescription history before their index date, and most had more than 5 years of recorded prescription history before their index date (82% of renal cancer cases and their controls and 70% of bladder cancer cases and their controls).
We used conditional logistic regression to estimate crude odds ratios and 95% confidence intervals for the risk of renal and bladder cancer associated with acetaminophen and non-steroidal anti-inflammatory drug exposure. The number of recorded prescriptions was evaluated as both a continuous and a categorical variable. We estimated adjusted odds ratios and 95% confidence intervals for renal and bladder cancer associated with various levels of acetaminophen and non-steroidal anti-inflammatory drug exposure controlling for smoking, body mass index (BMI), and (for renal cancer only) history of hypertension and diuretic use. We used SAS, Release 8.01 (SAS Institute Inc., Cary, NC) to perform the statistical calculations.
The study protocol was approved by the Scientific and Ethical Advisory Group of the GPRD.
We identified 109 cases of renal cancer, 20 cases of renal pelvis cancer, and 189 cases of bladder cancer who were diagnosed during the study period and had at least 5 years of prescription history recorded in the General Practice Research Database before their index date. Characteristics of the cases and controls are listed in Table 1.
We matched 434 controls to the 109 cases by age (within 1 year), sex, duration of drug history in the GPRD (within 1 year), general practice, and index date. Sixty-six of the cases (61%) were male. Forty-six of the cases (42%) were age 50–64 years, 35 (32%) were age 65–74 years, and 28 (26%) were age 75 years or older (see Table 1).
The crude odds ratio for renal cancer in subjects who had at least one prescription for acetaminophen recorded between 1 and 5 years before the index date was 1.6 (95% CI 1.1–2.5) compared with those who had no prescription recorded for acetaminophen. The strength of the crude association increased with increasing number of prescriptions (see Table 2). Using conditional logistic regression, we estimated the relative risk of renal cancer associated with varying levels of acetaminophen exposure while adjusting for smoking, body mass index, history of hypertension, and history of diuretic use, as well as for the factors on which subjects were matched (age, sex, general practice, duration of drug history in GPRD, and index date). In this analysis, the estimated odds ratio for renal cancer among subjects with any acetaminophen use during the period from 1 to 5 years before the index date was essentially the same as in the crude analysis (OR 1.6, 95% CI 1.0–2.6). Again, the strength of the association increased with increasing number of prescriptions (see Table 2). Presented in Figure 1 are the model-adjusted estimated odds ratios for renal cancer by number of acetaminophen prescriptions. The graph shows an increase in the odds of renal cancer with increasing use of acetaminophen.
Among the 23 cases of renal cancer with 6 or more acetaminophen prescriptions, the primary indications for treatment were arthritis in 7, neck pain in 3, back pain alone in 2, sciatica in 2, unknown in 2, and miscellaneous conditions in 7 (1 hip pain, 1 femur fracture, 1 herpes zoster, 1 disseminated sclerosis, 1 carpel tunnel syndrome, 1 headache and leg pain, and 1 gout). Twenty of these 23 cases received acetaminophen prescriptions over a period of 3 years or more before their index date.
The use of non-steroidal anti-inflammatory drugs was not associated with an increase in the crude estimated risk of renal cancer (OR 0.9, 95% CI 0.6–1.4), and in the adjusted analysis, we found no evidence for an association between renal cancer and non-steroidal anti-inflammatory drug use at any exposure level (data not shown).
We could not separately estimate adjusted relative risks for acetaminophen use among 20 additional cases of renal pelvis or transitional cell cancer of the kidney owing to low count frequencies in some categories; in an unadjusted analysis we found an OR of 1.2 (95% CI = 0.4–3.1) related to acetaminophen use from 1 to 5 years before the index date.
We matched 744 controls to the 189 cases by age, sex, general practice, duration of drug history in the GPRD, and index date. One hundred forty of the cases (74%) were male. Forty-five of the cases (24%) were age 50–64 years, 69 (37%) were age 65–74 years, and 75 (40%) were age 75 years or older (see Table 1).
There was little overall association between the incidence of bladder cancer and acetaminophen use from 1 to 5 years before the index date (OR 0.9; 95% CI 0.7–1.3). We found no evidence of increasing risk with increasing exposure to acetaminophen (see Table 3). In conditional logistic regression adjusting for smoking and body mass index, there was similarly no evidence of any increase in risk associated with any acetaminophen use between 1 and 5 years before the index date (OR 0.9, 95% CI 0.6–1.3) or with increasing acetaminophen exposure (see Table 3).
Although somewhat fewer subjects in the bladder cancer study had at least 5 years of recorded prescription history before their index date compared with those in the renal cancer study (see Subjects and Methods), a greater proportion of subjects had at least one recorded prescription for acetaminophen in the bladder cancer study (447/933 [47.9%]) than in the renal cancer study (221/543 [40.7%)]. Moreover, the proportion of subjects at each level of exposure was greater in the bladder cancer study than in the renal cancer study (264/933 [28.3%] compared with 146/543 [26.9%] for 1–5 prescriptions; 103/933 [11.0%] compared with 40/543 [7.4%] for 6–19 prescriptions; and 80/933 [8.6%] compared with 35/543 [6.5%] for 20 or more prescriptions).
We found no increase in risk of bladder cancer at any level of exposure to non-steroidal anti-inflammatory drugs in either crude or adjusted analyses (data not shown).
The present report of a study conducted in the General Practice Research Database confirms the results from our previous study carried out using data from Group Health Cooperative of Puget Sound, 11 and supports the conclusion that heavy use of acetaminophen is associated with an approximately 2-fold increase in the risk of renal cancer, although its use is not associated with the risk of bladder cancer. These studies have an advantage compared with other studies that have examined this question in that both ascertained acetaminophen exposure from prescription records that were generated in a general practitioner’s office or a pharmacy. Thus, the exposure data in these studies are less likely to be affected by recall bias than in studies that use interviews or questionnaires to characterize drug exposure. Occasional use of nonprescription acetaminophen is also unlikely to affect our results substantially. Meier (Meier CR, Schmitz S, Jick H. Use of paracetamol and non-steroidal anti-inflammatory drugs and the risk of developing ovarian, breast, or colon cancer. Submitted for publication, 2001) found in an interview-based self-assessment study administered by general practitioners that women are unlikely to use acetaminophen regularly (61–120 tablets per year) or chronically (121 or more tablets per year) without a prescription from their general practitioner.
Although in the present study we excluded cases diagnosed at the time of death (who may have received fewer medical services from their general practitioners than the cases we included), we doubt that the association we observed between acetaminophen use and renal cancer is due to selection bias because we applied the same exclusion to cases of bladder cancer and we did not observe an association between acetaminophen use and bladder cancer. Also, if selection bias had produced a spurious association between acetaminophen use and renal cancer, we would expect a similar association with the use of non-steroidal inflammatory drugs, which we did not find.
To reduce possible bias attributable to left censoring (that is, the tendency for subjects with index dates later in the study period to have more exposure data in the GPRD than subjects with earlier index dates), we disregarded exposure to acetaminophen before 5 years before the index date. Depending on the true induction period for renal cancer associated with acetaminophen use, we may have underestimated the quantity of etiologically relevant exposure. If earlier acetaminophen exposure were important, however, this nondifferential misclassification would have tended to weaken the association between acetaminophen use and renal cancer in our study. We also excluded from consideration any recorded exposure to acetaminophen during the year before the index date because we did not think recent exposure could have a causal role in carcinogenesis, and inclusion of recent exposure may have induced a spurious association with cancer because analgesics are often prescribed for cancer-related pain. If we over- or under-estimated the time during which this may have occurred, our estimate of the relative risk of acetaminophen exposure could be subject to some residual bias in either direction.
Still, for several reasons, we consider it unlikely that the association we observed between acetaminophen use and renal cancer could be due entirely to “reverse causation” (that is, the prescription of acetaminophen-containing drugs to treat symptoms of renal cancer). First, we did not find any association between the use of non-steroidal anti-inflammatory drugs and renal cancer, and it seems likely that general practitioners using analgesics to treat nonspecific pain that is a manifestation of undiagnosed renal cancer would not preferentially prescribe acetaminophen or non-steroidal anti-inflammatory drugs. Second, our review of the cases of renal cancer with heavy acetaminophen exposure revealed that back pain, which is the most common location of pain as a symptom of renal cancer, was not a frequent indication for acetaminophen use. Third, the duration of treatment of these cases with acetaminophen is longer than would be expected if their pain had been attributable to undiagnosed renal cancer. And finally, we excluded from our analysis prescriptions recorded during the year before each subject’s index date, when prescription of acetaminophen-containing products for pain attributable to undiagnosed cancer would be most likely. Nevertheless, the possibility that reverse causation accounts partially for the association we observed between acetaminophen use and the risk of renal cancer cannot be excluded entirely.
Our previous experience with the GPRD has indicated that diagnoses of cancer are confirmed in approximately 95% of cases by medical records kept in the general practitioners’ offices, such as pathology reports and death certificates. 14,15 We have also found that for breast cancer, incidence rates found among women in the GPRD are similar to those reported by the U.K. Office for National Statistics, 16 adding further to the validation of the GPRD as a source of reliable population-based information on incident cancer diagnoses and risk factors.
We thank the participating general practitioners for their ongoing contributions to the General Practice Research Database, and we thank Susan S. Jick, Alexander M. Walker, and two anonymous reviewers for their helpful comments on an earlier draft of the manuscript.
1. Clissold SP. Paracetamol and phenacetin. Drugs 1986; 32 (suppl 4): 46–59.
2. McLaughlin JK, Mandel JS, Blot WJ, Schuman LM, Mehl ES, Fraumeni JF Jr. A population-based case-control study of renal cell carcinoma. J Natl Cancer Inst 1984; 72: 275–284.
3. McLaughlin JK, Gao YT, Gao RN, Zheng W, Ji BT, Blot WJ, Fraumeni JF Jr. Risk factors for renal-cell cancer in Shanghai, China. Int J Cancer 1992; 52: 562–565.
4. McCredie M, Stewart JH, Day NE. Different roles for phenacetin and paracetamol in cancer of the kidney and renal pelvis. Int J Cancer 1993; 53: 245–249.
5. Kreiger N, Marrett LD, Dodds L, Hilditch S, Darlington GA. Risk factors for renal cell carcinoma: results of a population-based case-control study. Cancer Causes Control 1993; 4: 101–110.
6. Chow WH, McLaughlin JK, Linet MS, Niwa S, Mandel JS. Use of analgesics and risk of renal cell cancer. Int J Cancer 1994; 59: 467–470.
7. Rosenberg L, Rao RS, Palmer JR, Strom BL, Zauber A, Warshauer ME, Stolley PD, Shapiro S. Transitional cell cancer of the urinary tract and renal cell cancer in relation to acetaminophen use (United States). Cancer Causes Control 1998; 9: 83–88.
8. Piper JM, Tonascia J, Matanoski GM. Heavy phenacetin use and bladder cancer in women aged 20 to 49 years. N Engl J Med 1985; 313: 292–295.
9. Castelao JE, Yuan JM, Gago-Dominguez M, Yu MC, Ross RK. Non-steroidal anti-inflammatory drugs and bladder cancer prevention. Br J Cancer 2000; 82: 1364–1369.
10. Pommer W, Bronder E, Klimpel A, Helmert U, Greiser E, Molzahn M. Urothelial cancer at different tumour sites: role of smoking and habitual intake of analgesics and laxatives. Results of the Berlin Urothelial Cancer Study. Nephrol Dial Transplant 1999; 14: 2892–2897.
11. Derby LE, Jick H. Acetaminophen and renal and bladder cancer. Epidemiology 1996; 7: 358–362.
12. Jick H. A database worth saving. Lancet 1997; 350: 1045–1046.
13. Walley T, Mantgani A. The UK General Practice Research Database
. Lancet 1997; 350: 1097–1099.
14. Jick H, Jick SS, Derby LE, Vasilakis C, Wald Myers M, Meier CR. Calcium-channel blockers and risk of cancer. Lancet 1997; 349: 525–528.
15. Meier CR, Derby LE, Jick SS, Jick H. ACE-inhibitors, calcium channel blockers and breast cancer. Arch Intern Med 2000; 160: 349–353.
16. Kaye JA, Derby LE, Melero-Montes MM, Quinn M, Jick H. The incidence of breast cancer in the General Practice Research Database
compared with national cancer registration data. Br J Cancer 2000; 83: 1556–1558.