Correspondence

Kidney Failure and Analgesic Drugs

N Engl J Med 1995; 332:1514-1516June 1, 1995

Article

To the Editor:

The case–control study by Perneger et al. (Dec. 22 issue)1 on the association of end-stage renal disease (ESRD) with analgesic drugs reaches inappropriate conclusions that are based on flawed methods. Fundamental to the validity of case–control studies is the requirement that the control population be similar to the case-patient population except for the case-defining characteristic.2 The authors compared predominantly black (54 percent) and male (58 percent) case patients with mainly nonblack (86 percent) and female (65 percent) control subjects. The adjustments made do not deal with the probable differences in the socioeconomic, educational, and residential (urban vs. rural) status of the patients and the control subjects, and analgesic-drug use undoubtedly varies with these factors.

A second flaw is confounding by indication.2 To avoid gastrointestinal side effects and drug interactions, patients with many chronic diseases are more likely to take acetaminophen than other analgesic drugs. Since some of these patients are at risk for ESRD — for example, those with hypertension or diabetes mellitus — it should come as no surprise that they might have greater exposure to acetaminophen than patients without chronic disease. In other words, chronic disease is likely to be independently associated with both acetaminophen use and ESRD, thus confounding any causal relation based on the observed association. To avoid this “innocent bystander” effect the authors should have studied a control group of patients with chronic disease who did not have ESRD.

Lastly, it is inappropriate to extrapolate from a case–control study with relatively modest odds ratios, which, by the authors' own admission, may not demonstrate a causal relation to population-attributable risks and costs. Although some characteristics of the case patients may be similar to those of the national population of patients with ESRD, differences in racial and other features could well make the study nonrepresentative.3

Gerald Faich, M.D., M.P.H.
Pharmaceutical Safety Assessments, Narberth, PA 19072

3 References

1. 1

   Perneger TV, Whelton PK, Klag MJ. Risk of kidney failure associated with the use of acetaminophen, aspirin, and nonsteroidal antiinflammatory drugs. N Engl J Med 1994;331:1675-1679
   Full Text | Web of Science | Medline

2. 2

   Collet JP, Boivin JF, Spitzer O. Bias and confounding. In: Strom BL, ed. Pharmacoepidemiology. 2nd ed. New York: John Wiley, 1994:609-28.
   

3. 3

   Gaylin DS, Held PJ, Port FK, et al. The impact of comorbid and sociodemographic factors on access to renal transplantation. JAMA 1993;269:603-608
   CrossRef | Web of Science | Medline

To the Editor:

The case–control study by Perneger et al. of the risk of ESRD associated with the use of acetaminophen, aspirin, and other nonsteroidal antiinflammatory drugs (NSAIDs) provides no valid basis to support the authors' conclusion that “People who often take acetaminophen or NSAIDs have an increased risk of ESRD, but not those who often take aspirin.” Although a temporal association between the use of acetaminophen and ESRD was reported, it cannot be interpreted as causal because the time of acetaminophen use relative to the onset of renal disease was not determined. Patients with renal dysfunction are commonly advised to avoid taking aspirin and other NSAIDs and to take acetaminophen if they require an analgesic. This is sound advice, since NSAIDs may decrease renal blood flow and the glomerular filtration rate in asymptomatic patients with renal failure.1 Therefore, in patients in whom ESRD develops, the use of acetaminophen would be expected to exceed that of aspirin and other NSAIDs. This is a classic example of confounding by indication. A similar correlation might be found between ESRD and a low-protein diet or therapy with furosemide. A causal relation can be inferred only if greater use of acetaminophen is documented before the onset of the underlying renal disease that ultimately progressed to ESRD. The authors' conclusions, which could lead many patients with renal disease, as well as the general population, to change analgesic therapy from acetaminophen to aspirin, require greater substantiation and could result in a serious disservice to the public. The risk of such a switch, with its potential for fatal gastrointestinal side effects, is well established.2 In summary, this study does not support the authors' conclusions and recommendations related to acetaminophen, aspirin, or NSAIDs and the causes of ESRD.

Edward B. Nelson, M.D., Ph.D.
McNeil Consumer Products, Fort Washington, PA 19034

2 References

1. 1

   Whelton A, Stout RL, Spilman PS, Klassen DK. Renal effects of ibuprofen, piroxicam, and sulindac in patients with asymptomatic renal failure: a prospective, randomized, crossover comparison. Ann Intern Med 1990;112:568-576
   Web of Science | Medline

2. 2

   Strom BL. Adverse reactions to over-the-counter analgesics taken for therapeutic purposes. JAMA 1994;272:1866-1867
   CrossRef | Web of Science | Medline

To the Editor:

We wish to point out four sources of bias in the study by Perneger et al. that cast doubt on their conclusions. First, the number of pills taken does not equal the dose. The variable used to estimate analgesic-drug ingestion was the number of pills taken, rather than the estimated dose in grams. The amount of aspirin in different formulations varies more than sixfold; thus, a patient can take 405 mg of aspirin in five pills or 500 mg in one pill. Similarly, two common formulations of acetaminophen, 325 mg and 500 mg, were available during the study period. Over-the-counter preparations of ibuprofen and naproxen are available in fixed-dose formulations; hence, the number of pills ingested is proportional to the dose. The wide range of aspirin and acetaminophen doses available to patients introduces a bias that makes any assessment of dose–response based on pill counts uninterpretable.

Second, patients misidentify aspirin and acetaminophen. The Rocky Mountain Poison Center provides approximately 4400 and 900 telephone consultations each year concerning the ingestion of acetaminophen and aspirin, respectively. We have found that many patients cannot differentiate between these two common analgesic drugs, even when the ingestion is relatively recent. This effect would be magnified if the time of the ingestion was more remote.

Third, many over-the-counter combinations of aspirin and acetaminophen were probably omitted from consideration. Although histories regarding drug mixtures were taken, it is not clear whether the study subjects were asked about mixtures of analgesic medications exclusively, or whether information about the use of such products as cold preparations, antidiarrheal preparations, and sleep aids was also sought. During a typical week, approximately 70 percent of persons take at least one nonprescription drug.1 When this is extrapolated to the study group, it is apparent that failure to include these sources of analgesic medication could introduce a large bias.

Fourth, the measurement tool, a patient's estimate of ingestion over a period of many years, was not validated. Without evidence that a patient's estimate of the number of pills ingested is a valid measure of the number actually ingested, a large bias is possible. The important questions raised in this study regarding the role of long-term analgesic-drug ingestion and renal failure merit further research. However, the serious limitations outlined above caution against premature changes in practice or in regulatory policy, such as mandating a warning label or requiring a prescription for the purchase of large quantities of acetaminophen.

Rivka S. Horowitz, M.D., Ph.D.
Vickie L. Wilson, R.Ph.
Richard C. Dart, M.D., Ph.D.
Rocky Mountain Poison and Drug Center, Denver, CO 80204

1 References

1. 1

   Gagnon JP, Salber EJ, Greene SB. Patterns of prescription and nonprescription drug use in a southern rural area. Public Health Rep 1978;93:433-437
   Web of Science | Medline

Author/Editor Response

The authors reply:

To the Editor: Dr. Faich claims that in case–control studies, case patients and control subjects must be similar in all respects except the case-defining characteristic. This is not quite correct: the case patients and control subjects must be drawn from the same population, but they may, and indeed should, differ with respect to all characteristics that are related to the disease of interest — in this case, ESRD. These variables do not necessarily confound the relation between the use of analgesic drugs and ESRD. We adjusted for known confounders including age, race, and sex and failed to find confounding by other variables. The potential for confounding by indication remains a concern. We doubt that this mechanism explains an increased risk of ESRD in persons who regularly consume more than two acetaminophen pills per day. The chronic diseases cited (hypertension and diabetes) do not cause chronic pain. As suggested, a study using a control group of patients with chronic disorders would address this issue. There is nothing wrong with computing population-attributable risks for associations that are only suspected of being causal. Such computation may well indicate whether further exploration of the issue is worth the investment. We believe that this is the case for acetaminophen and ESRD.

As pointed out by Dr. Nelson, case–control studies cannot dissect the temporal relation between exposure to analgesic drugs and ESRD. However, the median duration of use of acetaminophen in those who reported using this analgesic >1 day per month was similar in the case patients and control subjects (10 years in either group). Thus, a recent switch to acetaminophen was no more likely in the case patients than in the controls. We also computed the odds ratios of ESRD for the number of pills of acetaminophen consumed over time before the initiation of treatment for ESRD in the case patients and before the interview in the control subjects. There was an excess risk of ESRD related to the intake of 1000 to 4999 pills at each of the three periods studied (Table 1Table 1Odds Ratios for ESRD Associated with Intake of Acetaminophen at Three Different Times, Adjusted for Sex, Age, Race, and the Use of Other Analgesic Drugs, in 716 Case Patients with ESRD and 361 Control Subjects.). An intake of >5000 pills of acetaminophen was associated with ESRD only in the five-year period closest to the initiation of treatment; this may indicate that ESRD occurs fairly rapidly when very large amounts of acetaminophen are consumed on a regular basis.

Defining the temporal relation between the ingestion of analgesic drugs and the onset of renal insufficiency is important and should be evaluated in prospective studies. However, if analgesic drugs lead to ESRD by aggravating preexisting renal disease, such studies may not be helpful. We agree that advising patients or the general public to switch analgesic drugs is not a trivial issue. As we stated in the article, this should take into account all available evidence (not just our report), the benefits and good safety record of acetaminophen, and the risks associated with using other drugs.

The four possible sources of bias cited by Horowitz et al. are in our opinion valid: our assessment of the use of analgesic drugs lacked precision and most likely underestimated actual consumption. However, these sources of bias are nondifferential — that is, all are likely to affect both the case patients and control subjects in a similar way. Such misclassification would bias the odds ratios toward unity (i.e., absence of association). This arouses concern that in reality, the risk of ESRD associated with the use of analgesic drugs is higher than we found.

Thomas V. Perneger, M.D., Ph.D.
Université de Genève, CH-1211 Geneva, Switzerland

Paul K. Whelton, M.D., M.Sc.
Michael J. Klag, M.D., M.P.H.
Johns Hopkins University, Baltimore, MD 21205

Citing Articles (1)

Citing Articles

1. 1

   Johannes M. Fox, Ute Siebers. (2003) Caffeine as a promoter of analgesic-associated nephropathy - where is the evidence?. Fundamental & Clinical Pharmacology 17:3, 377-392
   CrossRef