Independent studies are needed before causality is established

EDITOR —Lindberg et al suggested that the use of calcium channel blockers increases the risk of suicide.1 Methodological problems, however, render that conclusion uncertain. In a cross sectional ecological study they found a weak but significant correlation between rates of suicide and use of calcium channel blockers, expressed as numbers of defined daily doses dispensed by pharmacies in 152 municipalities in Sweden. The defined daily dose is, however, a technical unit for studies of use of drugs2; defined daily doses might differ twofold or more from the daily doses actually prescribed. Therefore, when used for other purposes, such as an estimate of individuals at risk (as in Lindberg et al's paper), methods based on the defined daily dose require validation.3

The authors also carried out a historical cohort study of patients with an index prescription of an antihypertensive drug. They found that “five users of calcium channel blockers (three men and two women, one with uncertain intent) and four non-users (three men and one women, none with uncertain intent) committed suicide” within seven years after they bought the index drug in 1988 or 1989. A minimum requirement for applying statistics on the outcome in nine individuals is to validate exposure as well as outcome. One misclassification in this study would mean that the difference was no longer significant. One of the “suicides” in the calcium channel blocker cohort was not even a certain suicide but an undetermined unnatural death. The remaining eight cases of alleged suicide were not validated against death certificates or medical records. It is not known whether these nine patients were taking an antihypertensive drug at the time of death, whether they were depressed, etc. Potential confounders, such as the severity of hypertension, comorbidity, concomitant drug treatment, and history of depression or use of antidepressants, were not controlled for, although such prescription data are available in the database and medical records can be made available for validation purposes in this population.4

It is vital that the non-experimental nature of pharmacoepidemiology is recognised. If the association between calcium channel blockers and suicide can be confirmed by validation of exposure and outcome, this association has to be confirmed in independent studies before any causality is established.

The authors do not have a conflict of interest.5

Drug prescriptions over longer period should have been followed up

EDITOR —Since 1988 there have been suspicions that calcium channel blockers may cause depression; these suspicions have been based on spontaneous reports, including reports of cases in which rechallenge gave a positive result.1 3Hallas recently found that calcium channel blockers and angiotensin converting enzyme inhibitors but not β blockers were more often prescribed before than after the start of antidepressant treatment4 this is another finding that needs confirmation.

In a population based prescription database Lindberg et al identified a cohort of patients who received at least one prescription of a cardiovascular drug in 1988 or 1989.5 These patients were followed up in the mortality register until the end of 1994. Nine suicides were identified; five of the patients who committed suicide (one with uncertain intent) had been exposed to calcium channel blockers and four to other cardiovascular drugs (relative risk 5.4 (95% confidence interval 1.4 to 20.5)). The authors calculated an excess risk of 1.1 suicides/1000 users/year.

We applied this estimate to sales of calcium channel blockers and the total number of suicides among men and women aged ≥45, and we calculated the proportion of all suicides in Sweden that might have been caused by calcium channel blockers (table).

It is highly implausible that such large proportions of all suicides would be caused by calcium channel blockers. More plausible is that the authors overestimated the risk. The significance of the estimated relative risk disappears if the one uncertain case is not taken as a suicide. The authors did not check the history of depression or use of antidepressants in the cohort. Since there is speculation that β blockers may cause depression, the cohort who had been prescribed calcium channel blockers may well have contained patients with prior depression. β Blockers are contraindicated in patients with obstructive lung diseases. This may be a confounding factor if such patients are at higher risk of suicide.

It is disappointing that the authors did not use the database to its full potential. They could have followed up all the patients' drug prescriptions, including prior use of antidepressants, continuously for a longer period to get more information. Such a study could have given valuable information to confirm or refute the idea that calcium channel blockers may cause depression. Lindberg et al's paper has not helped to elucidate that important question. It has, however, contributed to the confusion of prescribers and patients over the benefits and risks of calcium channel blockers.

The authors do not have a conflict of interest.

Prescriptions for particular drug are influenced by numerous factors

EDITOR —On the basis of evidence from ecological analysis and a population cohort, Lindberg et al suggested that use of calcium channel blockers may increase the risk of suicide.1 In ecological studies, findings are limited to describing differences in populations that would signal the presence of effects worthy of further investigation; this limitation is mainly because of the unavailability of data necessary for control of confounding.2

In an attempt to adjust for cardiovascular comorbidity the authors performed analyses in which they used partial correlation coefficients that adjusted the correlation (for each tested cardiovascular drug group) with the rate of use of the other drug groups. The authors reported a significant partial correlation coefficient (r) of 0.29 (P<0.001) between calcium channel blockers and suicide rates. However, the P value is determined largely by the size of the sample. The magnitude of r itself must be evaluated too. The degree of association can most usefully be expressed asr², the proportion of total variance in a dependent variable that can be explained by the independent variable.3 Thus less than 9% (that is, 0.292 of the variance in mortality from suicide is explained by use of calcium channel blockers—a value of only moderate association, since 91% of the variation is not explained.

A major flaw in the cohort study is the lack of adjustment for confounding effects. Only sex and age were adjusted for in the multivariate models. Other potential confounding factors associated with risk of suicide, such as socioeconomic status, comorbid conditions, severity of disease, mental status, functional status, cognitive status, and social support, were not considered. Patients receiving cardiovascular drugs are heterogeneous, and prescriptions for a particular drug are influenced by numerous other factors. Furthermore, the cohort study was based on use of prescription drugs in only one Swedish municipality between 1988 and 1989. Its generalisability to other populations and to the use of more recent calcium channel blockers is questionable. Moreover, use of death by suicide as a proxy for depression is neither sufficient nor effective in addressing the potential link between calcium channel blockers and depression. Finally, as only nine deaths from suicide occurred in the cohort it is difficult to draw causal conclusions on the basis of such a small number of outcome events without additional sensitivity analyses.

The findings of the current study probably result from ecological fallacy and an inadequate adjustment for confounding effects. Further investigations, with sufficient control for confounding effects, are needed to examine the association between use of calcium channel blockers and depressive symptoms.

Robert Makuch is a paid consultant to Searle (the manufacturer of verapamil).

Authors' reply

EDITOR —Suicide is a major cause of premature death, particularly among middle aged people, so examining whether commonly used drugs such as antihypertensives influence risk of suicide is important. Our study suggested that calcium channel blockers increase that risk.

We agree that ecological analyses based on defined daily doses have numerous pitfalls. That is why we combined our ecological study with a population based cohort study. Both these authors and Wiholm et al seem to believe that the significance of the increased risk of suicide in the cohort study depends on one single case. Any reader can verify that this is wrong. A crude relative risk can easily be derived from published data. Exclusion of the fifth (uncertain) case of suicide leaves four suicides in 617 patients exposed to a calcium channel blocker versus four suicides in 2780 subjects exposed to other antihypertensives; this yields P=0.041 with Fisher's exact test (0.013 with the fifth case included). A log rank test yields P=0.012 (0.002 with the fifth case included), and Cox regression analysis with age and sex as covariates yields P=0.044 (0.013 with the fifth case included) and a relative risk of 4.22 (95% confidence interval 1.04 to 17.14).

Wiholm et al present data on the proportion of Swedish suicides that would have been attributed to calcium channel blockers if our estimates were true. They then take these calculations as evidence that our results are overestimates. However, they ignore the fact that our study population is substantially older than the national average. Of all subjects aged ≥45 in Sweden, 43% are ≥65. The corresponding proportion in our study population is 70%. The low number of suicides in the study population makes analyses of age differences uncertain, but in a new analysis that we undertook for this letter the risk of suicide related to calcium channel blockers appeared only among elderly subjects. Age adjustment of Wiholm et al's data, based on the assumption that the increased risk of suicide exists only among subjects aged ≥65, shows that Wiholm et al may exaggerate the number of suicides related to calcium channel blockers by ≥62%. Their calculations are thus deceptive.

All three letters emphasise the need for adjustments for possible confounders. We discussed this at length in our paper. It is equally important to remember that overadjustment may camouflage a genuine effect.

In the ecological study about 9% of the variance in suicide rates among municipalities seemed to be explained by use of calcium channel blockers. Unlike Chen and Makuch, we consider this to be substantial;calcium channel blockers are used by under 3% of the population.