Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Interviews for drug exposure can be incorrect

Study
Results
Comment

It is often the case that in epidemiological studies associating drug use to later outcome, drug use is determined by an initial interview. Patients are asked about the drugs they are taking, as well as many other factors, and these are then used to assess whether there is or is not an association between drug use and outcome.

Patients can be mistaken. Over time drug use might change, especially as patients get older and new conditions emerge and are treated. A new study [1] demonstrates that information from initial interviews can be incorrect, and may change the results.

Study

The study was of inhabitants of a suburb of Rotterdam aged 55 years or older living there in 1990-1993. All obtained prescription medicines from pharmacies sharing one automated computer system. Cohort members were from a previous study in this population concerning a link between calcium channel blockers and cancer.

This earlier study collected information on calcium channel blocker use from an initial interview. Pharmacy records were used to assess whether the information at the initial interview was correct according to prescriptions filled. Nonusers filled no prescriptions, non-chronic users filled less than 180 days of prescriptions, and chronic users filled at least 180 days of prescriptions.

Results

There were 2,487 persons with a median follow up of 7.6 years. Users of calcium channel blockers (by interview) were much more likely than declared nonusers to have ischaemic heart disease, stroke, or angina, and to be using diuretics, -blockers, ACE inhibitors or statins.

At baseline, 206 persons said they were using a calcium channel blocker, and 2,281 said they were not (Table 1). Of the declared users, three did not have a prescription filled, and eight were non-chronic users. Of the declared nonusers, 204 were or became chronic users, and 150 non-chronic users.



Table 1: Use of calcium channel blockers by initial interview and by pharmacy records. Shaded areas show complete agreement



Prescription records
User status by interview
Total
Chronic
Non-chronic
Nonuser
User
206
195
8
3
Nonuser
2281
204
150
1927


The association between calcium channel blocker use and cancer using baseline assessment at interview had a relative risk of 1.4 (95% CI 0.9 to 2.1). The age and sex adjusted relative risk based on chronic exposure by pharmacy prescriptions was lower, at 0.9 (0.6 to 1.2). Use of interview data overestimated the risk by about 36%.

Comment

There was a clear misclassification of use of calcium channel blockers between initial interview and pharmacy prescription records. It could be argued that in this case it made no difference, because neither with interview nor pharmacy records was there a significant association between calcium channel blockers and cancer.

But the study serves to make a point: that there are potential biases that can occur in observational studies, and we probably know neither what they all are, nor therefore can we adjust for them all. Of course, if we have a large study, with lots of events, and big differences it probably makes no big difference. But if we have bare statistical significance, or small effects, or small numbers, then we should be cautious.

Reference:

  1. AB Beiderbeck et al. Misclassification of exposure is high when interview data on drug use are used as a proxy measure of chronic use during follow-up. Journal of Clinical Epidemiology 2004 57: 973-977.

previous or next story