Published in the January 3, 1996 Volume of the Journal of the American Medical Association

Carcinogenicity of Lipid-Lowering Drugs

Thomas B. Newman, MD, MPH, Stephan B. Hulley, MD, MPH

Data Sources – Summaries of carcinogenicity studies published in the 1992 and 1994 Physicians? Desk Reference (PDR), additional information obtained from the US Food and Drug Administration, and published articles identified by computer searching, bibliographies , and consultation with experts

Study Sample – We tabulated rodent carcinogenicity data from the 1994 PDR for all drugs listed as ?hypolipidemics.? For comparison, we selected a stratified random sample of antihypertensive drugs. We also reviewed methods and interpretation of carcinogenicity studies in rodents and results of clinical trials in humans.

Data Synthesis – All members of the two most popular classes of lipid-lowering drugs (the fibrates and the statins) cause cancer in rodents, in some cases at levels of animal exposure close to those prescribed to humans. In contrast, few of the antihypertensive drugs have been found to be carcinogenic in rodents. Evidence of carcinogenicity of lipid-lowering drugs from clinical trials in humans is inconclusive because of inconsistent results and insufficient duration of follow-up.

Conclusions – Extrapolation of this evidence of carcinogenesis from rodents to humans is an uncertain process. Longer-term clinical trials and careful post-marketing surveillance during the next several decades are needed to determine whether cholesterol-lowering drugs cause cancer in humans. In the meantime, the results of experiments in animals and humans suggest that lipid-lowering drug treatment, especially with the fibrates and statins, should be avoided except in patients at high short-term risk of coronary heart disease.

The past decade has seen a more than 10-fold increase in prescriptions for lipid-lowering drugs, with more than 26 million prescriptions in 1992 in the United States. The number of prescriptions for these drugs is likely to continue to increase because they are being aggressively promoted by their manufacturers and because many patients who are eligible for lipid-lowering treatment according to the most widely publicized US guidelines are still not receiving them.

The duration and impact of such treatment in any one patient may extend for 30 years or more. Yet the drugs have been approved by the Food and Drug Administration (FDA) based on the findings of full-scale clinical trials lasting less than a decade - far less in the case of the hydroxymethylglutaryl.C0A reductase inhibitors (the statins). Thus, millions of asymptomatic people are being treated with medications, the ultimate effects of which are not yet known. This situation has become more controversial recently, particularly for primary prevention, because meta-analyses of randomized clinical trials have suggested that cholesterol-lowering drugs may increase noncardiovasular mortality.
In this article we turn to less direct evidence. As part of the new drug approval process, pharmaceutical companies are required to submit data from rodent carcinogenicity studies to the FDA. These studies are not generally published in scientific journals, but are summarized in the Physicians? Desk Reference (PDR) and in the prescribing information that accompanies advertisements and the medications themselves. We examine this evidence herein, comparing results of carcinogenicity studies of lipid-lowering agents with those of antihypertensive agents, another class of drugs often prescribed for decades in asymptomatic people with the goal of preventing cardiovascular disease.

Methods

Rodent toxicity date were abstracted from the product information in the 1992 and 1994 editions of the PDR for all drugs listed in the ?Product Category Index? as ?hypolipidemics.? For each drug, we abstracted data on all the studies of carcinogenesis in animals. The relative exposures listed in Tables 1 through 3 represent the lowest exposures at which statistically significant increase in the incidence of tumors was reported in treated animals when compared with control animals.

For comparison, we drew a sample of oral drugs listed in the ?Product Category Index? as cardiovascular agents with hypertension s an inedication, after stratifying these adrenergic blockers, adregenic stimulants, a/b adrenergic blockers, angiotensin-converting enzyme inhibitors, B-blockers, calcium channel blockers, diuretics, rauwolfia derivatives, and vasodilators. After excluding combination and duplicate preparations, we randomly sampled one third of the drugs in each category, increasing the sample by one if the number of drugs in the category was not divisible by three.

Specific methods for conducting rodent carcinogenicity studies are at the discretion of pharmaceutical companies rather than being explicitly regulated by the FDA. The most common approach is to use Sprague-Dawley rats and CD1 mice, with groups of 50 animals of each sex randomly assigned to treatment at one of three doses of medication for 2 years. Tumor incidence in these groups is compared with the incidence in one or two control groups at each dose separately and in a test of trend.

Results

The product information for lipid-lowering drugs indicates that all the fibric acid derivatives and statins caused cancer in rodents (Table 1). In most cases the rodent exposure at which carcinogenicity was observed was of the same order of magnitude as that observed with the maximum dose recommended for humans. Cholestyramine resin and probucol were not found to be carcinogens themselves, but cholestyramine enhanced the effects of other carcinogens. Niacin,.like many older drugs, was not tested for carcinogenicity in rodents.

The carcinogenicity data for most drugs listed in Table 1, from the 1994 PDR, present relative exposure in terms of blood levels. Before 1993, relative exposures were3 presented in terms of milligrams per kilogram of body weight. The information in the PDR changed in 1993 for the two most popular cholesterol-lowering drugs, lovostatin and gemfibrozil (Table 2). The 1992 PDR reported that lovastatin causes liver cancer in mice at 312 times the maximum recommended human dose when relative exposure is based on the dose administered in milligrams per kilogram of body weight, but the 1994 PDR reported carcinogenicity at three to four times the human exposure when relative exposure is based on blood levels. Similarly, gemfibrozil is listed as causing cancer at 10 times the human dose in the 1992 PDR when expressed as milligrams per kilogram of body weight but at 1.3 times the human dose in the 1994 PDR based on blood levels achieved.

For comparison, we examined whether rodent carcinogenicity is also reported for antihypertensive drugs (Table 3). (Note that for antihypertensive drugs most of the relative exposures are based on milligrams per kilogram administered; these are higher than the values when expressed as milligrams per meter squared of body surface area, or as blood levels.) unlike the lipid-lowering drugs, most drugs for lowering blood pressure do not cause cancer in rodents. Of the 20 hypertensive drugs sampled for Table 3, only three (15%) caused cancer and three (15%) caused benign tumors. Among all 41 antihypertensive drugs for which results were reported in the PDR, six (15%) caused cancer and seven (17%) caused benign tumors. For 28 (68%), results of carcinogenicity studies were entirely negative.

COMMENT
Extrapolating From Rodent Studies

In this article we call attention to evidence of experimental carcinogenicity of lipid-lowering medications. This class of drugs has come into widespread use in people who are currently healthy, but who have laboratory findings – high blood cholesterol levels – that place them at above-average risk for their age for the development of coronary heart disease (CHD) in the future. Because the latent period between exposure to a carcinogen and the incidence of clinical cancer in humans may be 20 years or more, the absence of any controlled trials of this duration means that we do not know whether current drug treatment of hypercholesterolemia will lead to an increased rate of cancer in coming decades.

What are the implications of this uncertainty for clinicians and patients trying to decide whether to prescribe or take cholesterol-lowering drugs? The answer depends on how we extrapolate the risk of cancer from rodents to humans and how the risk compares with the benefit of preventing CHD by lowering blood cholesterol levels.

The extrapolation of cancer risk from rodents to humans is controversial, particularly for agents such as those mentioned in this article that test negative on the Ames test or other tests for mutagenicity. Mechanisms of carcinogenicity in such drugs are incompletely understood, and sufficient empiric data to estimate what proportion represents a clear risk to humans are lacking. Most all known human carcinogens have been found to be carcinogenic in mice or rats, ie, the sensitivity of rodent studies is reasonably high. Unfortunately, what we need to know is not sensitivity, but relative predictive value, and the positive-predictive value of rodent carcinogenicity assays is not known. There are several examples of rodent carcinogenicity that were subsequently proven to be carcinogenic in humans, but for the carcinogenicity of such chemicals, data .re not adequate to establish human carcinogenicity or lack thereof.

Matushak/Pinsky et al have argued that the concordance rate between rats and mice, 0% to 75%, provides an upper end of the concordance to be expected between rodents and humans. However, these are diverse species, probably with varying susceptibility to carcinogens. In some cases, differences in carcinogenic metabolism with the human species is larger than average differences between humans and rats. Thus, drugs that are carcinogenic in rodents may be carcinogenic in humans in general or in subsets of the human population but not others, or they may not be carcinogenic in humans at all.

Extrapolation from rodents to humans deserves special comment for the record. These drugs belong to a group of mutagenic carcinogens in which heightened carcinogenicity is associated with proliferation of peroxisomes (an organelle in liver cells involved in oxidative metabolism). Because humans and other primates are much less susceptible to this peroxisomal proliferation than rodents, some authors have questioned the relevance of the rodent carcinogenicity of these drugs. However, some peroxisome proliferation has been reported in humans on therapeutic doses of fibrates, and other histologic and biochemical changes in the liver are common, especially with long-term use. Thus, while it is unlikely that the fibrates are as carcinogenic in people as in rodents, it would be unwise to dismiss the rodent studies entirely.

Interpretation of rodent findings commonly involves extrapolating not only from rodents to humans, but also from high dose to low dose. However, the dose extrapolation is less problematic for cholesterol-lowering drugs because humans are exposed to doses similar to those that cause cancer in rodents. Gold et al have emphasized the importance of relating concentrations of chemicals that cause cancer in rodents to the concentrations to which humans are typically exposed. In their ranked list of 80 possible carcinogenic hazards to humans, the average daily dose of the cholesterol-lowering drug clofibrate was ranked second, exceeded only by high levels of daily occupational exposure to the fumigant ethylene dibromide.

In sum, rodent carcinogenicity studies can provide only a suggestion of carcinogenic risk to humans. However, the consistent carcinogenicity of the cholesterol-lowering drugs, coupled with their intended pattern of long-term use in healthy people, sets them apart from the antihypertensives and other drugs and is a matter of concern. If this were not so, there would be no point in requiring that drugs be tested for carcinogenicity in rodents. As stated by the World Health Organization?s International Agency for Research on Cancer, ?. . . in the absence of adequate data on humans, it is biologically plausible and prudent to regard agents and mixtures for which there is sufficient evidence of carcinogenicity in experimental animals as if they presented carcinogenic risk to humans?.

Estimating Relative Exposure

As shown in Table 2, the apparent riskiness of a drug depends on how relative exposure is measured. Because the patterns of drug absorption, metabolism, and exretion are different in rodents and humans, comparisons of drug exposures based on blood levels over time are now considered more valid thatn comparisons based on dose administrere per unit of body weight. The FDA policy on how to express dose equivalency has changed accordingly. When doses are compared using blood levels rather than dose administered, the animal dose that causes cancer appears closer to the dose prescribed to humans. For hepatic carcinogens the blood levels may underestimate exposure, both in rodents and in humans, because the liver may be exposed through portal circulation to much higher levels of drugs given orally than measurements in the peripheral blood would indicate.

A draft guideline recently issued by the FDA, developed as part of an international effort to harmonize drug testing requirements, suggests testing drugs at systemic exposures (based on the area under the curve of blood levels vs time) at least 25 times those observed at the maximum dose recommended for humans. Table 1 shows that most of the lipid-lowering drugs cause cancer at exposures well below this suggested margin of safety.

Why Were These Drugs Approved?

How did it happen that cholesterol-lowering agents were approved by the FDA for long-term use in spite of their animal carcinogenicity? To address this question, we obtained minutes of the Endocrinologic and Metabolic Drugs Advisory Committee meetings (under the Freedom of Information Act) at which lovastatin and gemfibrozil were discussed. For lovastatin, part of the answer may be that the doses for the carcinogenicity data were presented in milligrams per kilogram of body weight. As shown in Table 2, this presentation gives the impression that carcinogenicity occurred only at very high doses. In any case, the only reported discussion of animal carcinogenicity studies at the FDA advisory committee meeting on lovastatin (February 19 and 20, 1987) was by the representative of Merck Sharp & Dohme (makers of Mevacor brand of lovastatin), who downplayed the importance of the studies.

In contrast, carcinogen city appears to have been a greater concern to participants of the meeting on gemfibrozil (October 17, 1988). The minutes state, ?Dr. Gloria Troendle (deputy director, Division of Metabolism and Endocrine Drug Products for the FDA) noted that gemfibrozil belongs to a class of drugs that has been shown to increase total mortality. It has been shown to have animal carcinogenicity and she does not believe the FDA has ever approved a drug for long-term prophylactic use that was carcinogenic at such low multiples of the human dose as gemfibrozil.?

Elizabeth Barbehenn, PhD, a pharmacologist with the FDA, expressed similar concerns at the same meeting. After summarizing animal carcinogenicity studies and reviewing the hypothesis that liver carcinogenicity is related to (missing text in original copy of report), she pointed out that ?peroxisomal proliferation has been found in the liver of all species studied? and that in any case peroxisomal proliferation in the liver would not explain tumors in other tissues. She concluded that ?fibrates must be considered as potential human carcinogens and their carcinogenic potential should be part of the risk-benefit-equation for evaluating gemfibrozil.? The rebuttal to these comments came from representatives of Parke-Davis (makers of the Lopid brand of gemfibrozil), who ?noted that peroxisomes in man are not the same as peroxisomes in rodents,? peroxisomal proliferation although florid in rodents ?has not been proven unequivocally in man? and that ?unlike the vast majority (85% to 90%) is not mutagenic and not genotoxic.?

When asked to vote at the end of the meeting, the minutes state that only ?three of the nine members (of the advisory committee) believed that the potential benefit of using gemfibrozil for prevention of coronary heart disease out weighed the potential risk associated with such use.? However, such votes are only advisory to the FDA, which decided to approve the re-labeling of gemfibrozil for prevention of CHD, although only for the narrow group of patients in whom the Helsinki Heart Study suggested the greatest likelihood of benefit ?(Type IIb) patients with low HDL-C (high-density lipoprotein cholesterol), elevated LDL-C (low-density lipoprotein cholesterol) and triglycerides who have had an inadequate response to weight loss, diet, exercise, and other pharmacologic agents such as bile acid sequestrants and nicotinic acid? (Lopid advertisement in the New England Journal of Medicine, June 15, 1989). Unfortunately, the subsequent popularity of gemfibrozil suggests that its use has not been restricted to this small groiup; it was the second most popular lipid-lowering drug in the United States in 1992, the most recent year for which data are available.

Risks and Benefits of Lipid-Lowering Drug Treatment

How should the worrisome but uncertain risk of cancer be weighted against demonstrated benefits of cholesterol lowering? The answer depends on both the patient and the class of drug being considered. For patients with known CHD, the recent Scandinavian Simvastin Survival Study strongly suggests that the benefits of cholesterol lowing exceed the risks, at least6 in men an din the short term (5 years). This result is consistent with earlier meta-analyses of trials of other classes of cholesterol-lowering drugs, which also suggested a mortality benefit in those at high short-term risk of CHD. Given the strength of this evidence, it is reasonable to treat high blood cholesterol with drugs in patients with coronary or other atherosclerotic disease.

On the other hand, for patients not at high (>1% per year) short-term risk of CHD death, especially patients with life expectancies of more than 10 to 20 years, pharmacologic treatment probably should be avoided. For this group, the benefits of treatment are smaller and the potential risk of increased cancer in the decades after treatment is of greater concern. More significantly, meta-analysis of cholesterol-lowering trials have suggested that in this large group the risks of cholesterol lowering may exceed the benefits.

Although, as discussed herein, the risk of causing or promoting cancer probably needs to be considered separately for the different classes of cholesterol-lowering drugs (and possibly for different members within a class), it is worth addressing the possibility that it is the cholesterol lowering itself, rather than an adverse effect of the drugs, that might be carcinogenic. Persons with low cholesterol levels have higher cancer death rates in cohort studies, but evidence for causality for this association is weak. At least for some cancer sites there is good evidence that preexisting cancer or confounding may be responsible. Clinical trial evidence on this question is also inconclusive. In a recent meta-analysis of primary and secondary prevention trials, Law et al found an odds ration for cancer death of 1.07(95% confidence interval (CI). 0.90 to 1.26), decreasing to 0.85% (95% CI, 0.74 to 1.05) when cancer deaths from six clinical trials with extended follow-up were pooled. This finding occurred because increased cancer deaths during or soon after cholesterol lowering in some randomized trials tend to be balanced by fewer cancer deaths than expected later. This occurrence may be due to chance or a tendency of cholesterol lowering to promote rather than cause cancer. While trials with extended follow-up are in some ways reassuring, we must remember that what happens after cholesterol-lowering interventions have been discontinued may not reflect what will happen if medications are continued for many more years.

What can be said specifically about the likely risks and benefits of different classes of drugs? The greatest concern is with the fibrates, because in the two largest primary prevention trials of these drugs, there was a higher total mortality in the intervention group. Current labeling for gemfibrozil also cautions against using it for secondary prevention because of the trend toward increased CHD and total mortality in the Helsinki Hart Study secondary prevention arm. Thus, quite apart from their animal carcinogenicity, the fibrates should have limited use for either primary or secondary prevention of CHD.

Risks and benefits are much less clear for the statins, currently the most popular group of cholesterol-lowering drugs, because for this class of drugs clinical trial experience, while more favorable, has involved too few people for too short a time. There were only 68 cancer deaths in the 5.4 years of the Scandinavian Simvastatin Survival Study, evenly divided between the simvastatin and placebo groups (relative risk=0.94%;95%CI, 0.6 to 1.5). If drugs from this class caused a rapid increase in a particular cancer (especially an otherwise uncommon cancer), this adverse effect could be discovered from post marketing surveillance or tumor registries. On the other hand, if (as is the case for smoking) an increase in cancer was delayed for decades, or if (as discussed herein) there was a more subtle or more diffuse increase in overall risk of cancer, the statins? carcinogenicity could have important public health consequences, but would be difficult to detect. Thus, it seems prudent to reserve the statins for people at high short-term risk of heart disease and to be wary about their long-term use.

Niacin and cholestyramine, although harder to take than other hypolipidemic drugs because of frequent side effects, may be safer. (Colestipol hydrochloride may e similar to cholestyramine, but data on it are much more limited.) although the PDR provides little data on rodent carcinogenicity of either of these drugs, they are older, and long-term (12 to 15 years) follow-up of clinical trials in which they were used reveals more favorable results than were seen with the fibrates. In the Coronary Drug Project, a secondary prevention trial, niacin decreased all-cause mortality, although this did not occur until after the participants had stopped taking the drug. Cholestyramine has not been shown to decrease all-cause mortality and may be a cancer promoter, but the follow-up of the large Lipid Research Clinics trial did not show any trend toward increased cancer deaths or total mortality.

Conclusion

Most cholesterol-lowering drugs cause or promote cancer in rodents. Patents to whom these drugs are prescribed, either singly or in combination, are exposed throughout many years to doses approaching those shown to be carcinogenic in animals. Although consideratlbe uncertainty is involved in extrapolating results of carcinogenicity studies fropm rodents to humans, the implication of these findings matches that of meta-analyses of clinical trials in humans. Use of cholesterol-lowering drugs should be restricted to those at high risk of short-term CHD death, such as those with prior CHD, in whom the short-term.

Table 1