Modest Association of Long-Term ACE Inhibitor Treatment With Lung Cancer: The Promise and Pitfalls of Epidemiological Drug-Safety Analyses

Abstract

Results of the carefully executed Evaluation of Treatment with Angiotensin Converting Enzyme Inhibitors and the Risk of Lung Cancer (ERACER) study, reported in this issue, echo those of several previous observational analyses of the association of long-term angiotensin-converting enzyme (ACE) inhibitor use with incident lung cancer. These epidemiological drug-safety analyses merit cautious interpretation. First, the number needed to harm (NNH) of 6667 reported in ERACER for ACE inhibitors compared with angiotensin-2 receptor blockers (ARBs) after approximately 12 years of follow-up should be balanced against therapeutic benefits. Previously reported meta-analyses of randomized controlled trials (RCTs) over a mean 4.3-year follow-up suggested number needed to treat (NNT) of 67 for all-cause mortality, 116 for cardiovascular mortality, and 86 for a composite of myocardial infarction (MI) and stroke for ACE inhibitors, compared with nonsignificant benefits for ARBs on the mortality outcomes and NNT of 157 for ARBs on the MI/stroke composite. Second, confounding by indication is possible because until 2013, ACE inhibitors, not ARBs, were first-line medications for heart failure, which is associated with incident lung cancer. Third, findings may be compromised by detection bias due to investigation of ACE inhibitor-induced cough, or by residual confounding due to influential factors not measurable in the available data, such as socioeconomic status (SES) or smoking history. The important questions raised by ERACER and similar drug-safety analyses should be addressed in long-term RCTs or in enhanced large-database pharmacoepidemiological analyses, measuring both NNH and NNT and controlling for SES, indication, medication, and dosage.

In this issue, Anderson et al share the results of the Evaluation of Treatment with Angiotensin Converting Enzyme Inhibitors and the Risk of Lung Cancer (ERACER), a well-executed time-to-event comparison of angiotensin-converting enzyme inhibitors (ACE inhibitors) with angiotensin-2 receptor blockers (ARBs) on a primary outcome of incident lung cancer and secondary composite outcome of incident lung cancer or death. ¹ Strengths of this observational study include a large (N = 187,060) sample, follow-up time exceeding those of randomized controlled trials by a mean of approximately 3 years, and reliance on an electronic data warehouse with which the authors’ institution has more than 25 years of experience. To reduce biases common in pharmacoepidemiological analysis, ² the authors employed careful design features: exclusion of prevalent users and cancers occurring during the first year of follow-up to eliminate prevalent-user and latency biases, statistical control of baseline patient characteristics to reduce selection bias, and extensive sensitivity analysis. ¹

The authors found that after a mean 7.1 (up to 20) years of follow-up on treatment for any indication, rates of the primary outcome for those treated with ACE inhibitors and ARBs, respectively, were 2.31 and 2.16 per 1000 patient-years (adjusted hazard ratio [HR] = 1.18, 95% confidence interval [CI] = 1.06-1.31). ¹ The association was modest, with a number needed to harm (NNH) of 6667, and emerged only after about 12 years of follow-up. An association of ACE inhibitor treatment with the composite secondary outcome was also modest (33.8 vs. 35.1, respectively, per 1000 person-years, HR = 1.07, 95% CI = 1.05-1.09, NNH = 769).

As noted by Anderson et al, the findings of ERACER are consistent with those of 2 previous observational studies. One by Hicks et al, an analysis of a United Kingdom database covering about 15 million patients seen in general practice for a mean follow-up of 6.4 (range 1-20) years, identified incident lung cancer rates of 1.6 versus 1.2 per 1000 for those treated with ACE inhibitors and ARBs, respectively. ³ Another by Lin et al, a propensity-matched cohort analysis of the National Health Insurance database in Taiwan (follow-up minimum of 28 days, mean of 6 years), produced similar results, with rates of 1.22 and 1.66 per 1000 person-years, respectively, for those treated with ACE inhibitors and ARBs. ⁴ More recently, a nested case-control analysis of Danish national registries indicated an odds ratio of 1.33 (95% CI = 1.08-1.62) for lung cancer, comparing patients treated with >3650 defined daily doses, roughly corresponding to 10 years of treatment at a standard adult dosage, of ACE inhibitors instead of ARBs. ⁵ All 4 author teams attributed their results to a biologically plausible connection of ACE inhibitor treatment with lung accumulation of bradykinin, a vasodilator that acts on B₂ receptors expressed in various adenocarcinomas including those of the lung, ⁶ and substance P, a peptide that acts on the neurokinin-1 receptor system in tumor cells to induce proliferation, migration, and angiogenesis. ⁷

What should we make of these findings, which are qualitatively consistent across evaluations of 4 geographically disparate populations? ^{1,3 -5} A need for careful and critical assessment is apparent from the scope and content of popular press coverage of the study by Hicks et al after its publication in 2018. ^{8 -14} Although much of the coverage appropriately expressed cautious interpretation because of the study’s nonrandomized design, some reports adopted a more unquestioning view, with one Forbes medical commentator declaring that “ACE inhibitors are a risk factor for lung cancer…[that] should be contraindicated in smokers and ex-smokers.” ¹⁰ Similarly, an article posted on a health information website for consumers noted that the “millions of Americans [who] take [ACE] inhibitors to lower their blood pressure…might also increase their odds of developing lung cancer.” ¹⁴ Given the possibilities of similar less-than-careful presentation of ERACER in the popular press, or of insufficient consumer understanding of study limitations, how should clinicians and academicians prepare to address questions or concerns raised by patients? Several factors suggest a need for circumspection.

Mortality Benefit of ACE Inhibitors

Foremost, as noted in an editorial that accompanied the publication by Hicks et al, ¹⁵ the extremely small risk of lung cancer observed after long-term treatment with ACE inhibitors should be balanced against the much greater mortality benefit of these medications. For example, in a meta-analysis of 18 randomized controlled trials including 68,343 patients treated for hypertension with renin-angiotensin-aldosterone system (RAAS) inhibitors, Brugts et al found that the numbers needed to treat (NNT) over a mean 4.3 years of treatment with ACE inhibitors were 67 for all-cause mortality, 116 for cardiovascular mortality, and 86 for a composite of myocardial infarction and stroke. ¹⁶ For each of these outcomes, NNT values for ARBs were considerably larger, and 2 were nonsignificant (NS): 335 (NS) for all-cause mortality, 409 (NS) for cardiovascular mortality, and 157 for the composite of myocardial infarction and stroke. Although none of these randomized studies compared ACE inhibitors and ARBs head-to-head, and their mean follow-up time was shorter than that of 3 of the 4 observational studies, ^1,3,5 considerable mortality benefit for ACE inhibitors in hypertension is evident.

Residual Confounding

An additional reason for circumspection, also noted in the editorial that accompanied the work of Hicks et al, ¹⁵ is the likely possibility that despite efforts to mitigate selection bias, residual confounding remains. ^2,17 The “e-value,” defined as the minimum association of a single unmeasured confounder and outcome that would explain study results (i.e., if the associations were not truly causal), ¹⁸ helps illustrate the problem. A standard e-value formula (available as an online calculator at https://www.evalue-calculator.com/) indicates that the ERACER findings would be explained by an unmeasured confounder that multiplies the HR of both ACE inhibitor use and outcome by 1.64; an HR of only 1.27 would explain the composite outcome of lung cancer and mortality. Although the e-value itself should be interpreted cautiously because it assumes only a single confounder, whereas multiple confounders may exert influence on observational analyses, ¹⁹ the calculation does highlight the need for extra care in interpreting weak associations.

As is common in observational analyses of administrative data even after statistical adjustment, numerous unmeasured confounders are possible. ^2,17 The list of patient characteristics for which the ERACER Cox regression models controlled (age, sex, body mass index, smoking, chronic obstructive pulmonary disease, and alcohol misuse), excluded potentially important factors. Among these is socioeconomic status disparity, known to be associated with lung cancer, ²⁰ that could have arisen from substantial out-of-pocket drug cost differences between generic and branded products during the study period. ¹⁵ After accounting for the sampling protocol’s minimum pre-treatment observation, patients could have initiated treatment as early as 1997, 2 years after availability of generic captopril but 13 years prior to availability of the first generic ARB, losartan, in 2010. ²¹ Detection bias associated with diagnostic workups of ACE inhibitor-induced cough and lack of information about smoking history (e.g., pack years) also could have contributed to results. ¹⁵ Finally, effects of specific drug, dosage, and indication for use were not reported in any of the 4 studies, although Lin et al did successfully propensity match cohorts on hypertension and several comorbidities. ⁴

Possible Confounding by Heart Failure Indication

Failure to measure indication may be the most important of these omissions because of possible between-group differences in the prevalence of heart failure, a condition linked to incident cancer in several studies. ²² Notably, ACE inhibitors, not ARBs, were guideline-recommended medications of choice for heart failure at the start of ERACER’s observation period. ²³ Although trial results began in 2000 to indicate comparable mortality outcomes for ACE inhibitors and ARBs in patients with heart failure, the 2002 Optimal Trial in Myocardial Infarction with Angiotensin II Antagonist Losartan (OPTIMAAL) study documented a nonsignificant trend toward higher mortality with losartan compared with captopril. ^23,24 ARBs were not recommended for use in heart failure, except for ACE inhibitor-intolerant patients, until 2013, ²⁵ the final year of treatment initiation in ERACER. This history indicates a potential confounding by indication not addressed in the designs of ERACER or the other 3 observational analyses. Even the study by Lin et al, which propensity-matched on indication, included no measurement of heart failure prevalence. ⁴

Summary Assessment and Directions for Future Research

With these caveats, we are left with needs for both cautious interpretation and, as observed by Anderson et al, ¹ for further investigation. Ideally, we need high-quality long-term randomized trials, a gold standard for elimination of selection bias, ²⁶ but the feasibility of conducting trials with the necessary minimum follow-up time of 10-15 years is questionable. The optimal approach may be continued leveraging of large administrative databases. These already have a recognized role in signaling potential drug-safety issues not identified in randomized trials, as highlighted by a recent pharmacoepidemiological analysis of cardiac arrhythmias recorded in the U.S. Food and Drug Administration Adverse Event Reporting System (FAERS) database. ²⁷

Such observational work could improve on existing evidence with the addition of key design elements to address methodological concerns raised here and by others. ^15,28 These include measurement and statistical controls for drug dosage; drug-specific rather than solely class-specific effects, as pharmacokinetics vary among ACE inhibitors; cardiovascular events, to assess both NNT and NNH in a single sample followed over a uniform time period; original indication for treatment; changes in indication (e.g., infarction and heart failure in a patient originally treated for hypertension); and patient out-of-pocket cost-sharing or enrollment source (e.g., Medicaid versus private insurance) as a proxy for socioeconomic status.

Although an admittedly inadequate substitute for randomized evidence, this approach would provide physicians and patients with increasingly specific answers to the drug-safety question raised by the 4 epidemiological analyses of RAAS inhibitors reported to date. That the answers offered by ERACER and the other observational studies are incomplete, perhaps even misleading, does not negate the importance of the question, which affects more than 100 million U.S. adults considering the hypertension indication alone. ²⁹ Meanwhile, for the common indication of hypertension, currently available randomized evidence indicates clearly that the benefits of ACE inhibitor use far outweigh the potential long-term risks. ¹⁶