Influence of control selection in nested case‐control studies: the example of exposure to antiretroviral treatment (ART) and the risk of myocardial infarction (MI) in the French Hospital Database on HIV (FHDH‐ANRS CO4)

Abstract

Soon after availability of protease inhibitors (PIs), a duration-related effect relationship between PI and myocardial infarction (MI) was shown. New antiretroviral treatments (ARTs) have allowed more individualized regimens. To study their influence established risk factors of MI and additional therapeutic options such as lipid-lowering drugs will have to be taken into account. A nested case-control is an interesting alternative raising the choice of controls. With the previous full cohort analysis as reference, we investigated the influence of control selection in nested case-control studies sampled in this cohort by testing nine sampling scenarios. During the period 1996-1999, 49 MI occurred among male patients exposed to PI and followed-up in the French Hospital Database on HIV (FHDH-ANRS CO4). For each case, controls were selected using incidence-density sampling. The influence of additional matching criteria was tested. Random sampling and analysis was repeated 100 times with varying control-case ratios. When controls were randomly selected among patients of the same age who were free of MI at the date MI was diagnosed in the case, we observed a duration-related effect relationship between PI and MI in agreement with the results of the full cohort analysis. The use of four controls per case was sufficient. Estimates obtained with simple sampling were more precise than those obtained when controls were also matched for year of enrollment, initial CD4 cell count and HIV transmission group. To study ARTs as MI risk factors, nested case-control using incidence-density sampling without additional matching is one appropriate option.