Author: Mira Katan MD, MS, Department of Neurology, University Hospital of Zurich, Switzerland
In clinical stroke research we are looking into relationships between exposure variables such as specific treatments (e.g. statins) or risk factors (e.g. hypertension) and a subsequent outcome variable such as functional outcome, mortality or stroke recurrence.
Effect modification describes the situation where the magnitude of the effect of an exposure variable on an outcome variable differs depending on a third variable. In other words the presence or absence of an effect modifier changes the association of an exposure with the outcome of interest. In biology effect modification can also be described as synergy or antagonism depending on whether the effect modifier increases or decreases the effect1.
The clinical motivation behind the assessment of effect modification is to identify whether the effect of for example a treatment is different in groups of patients with different characteristics2. It seems reasonable to belief that not all treatment effects are homogeneous across all patients whether they are young or old, male or female etc., thus looking for effect modification is in a way a first step towards personalizing treatment. Discovering and describing effect modification is therefore desirable and leads to a more elaborate description of the effect of a treatment itself 3. Assessing effect modification helps in identifying patients who may benefit most from a treatment or may not benefit from a treatment at all. Here a theoretical example where treatment is modified by stroke severity. Patients with mild to moderate strokes benefit while those with severe strokes show no effect upon treatment.
One recently published example where effect modification occurs is in patients with symptomatic carotid stenosis treated with carotid stenting. In a recent meta-analysis of pooled patient data from four randomized trials in patients with symptomatic carotid stenosis, the effect of carotid stenting (CAS) versus carotid endarterectomy (CEA) on the risk of a subsequent stroke and death among different age groups was assessed4. They found that there was a four times greater risk of peri-procedural risk of CAS treated patients if they were older than 70 years compared to those younger than 60 years. By contrast age had little effect on the risk of patients assigned to CEA4. Another example where effect modification takes place is in patients treated with oral anticoagulants. A meta analysis of the ARISTOTEL, AVERROES, RE-LY(150mg dose) and ROCKET AF trials reported that women treated with NOACS had lower rates of major bleeding compared with men after multivariable adjustment for other vascular and demographic risk factors (OR 0.84, 95%CI 0.75-0.96 p=0.007)5.
There are many more examples where effect modification can be observed. However, RCTs adequately powered to test effect modification by for example gender are still quite rare and thus most data come from meta-analyses. Ensuring adequately powered RCTs to assess differences in treatment effects by potential effect modifiers are essential to develop an adequate evidence base to manage specific patient subgroups.
- Fletcher H.F., Fletcher S.W., Fletcher G.S. Clinical Epidemiology- THE ESSENTIALS 5th Wolters Kluwer/Lippincott Williams & Williams 2014
- Corraini P. et al. Effect modification, interaction and mediation: an overview of theoretical insights for clinical investigators. Clinical Epidemiology 2017:9; 331-338.
- Rothman KJ. Modern Epidemiology. Boston: little Brown and Co.; 1986
- Howard G. et al. Association between age and risk of stroke or death from carotid endarterectomy and carotid stenting: a meta-analysis of pooled patient data from four randomized trials. The Lancet 2016: 387; 1305-1311.
- Pancholy S.B. et al. Meta-analysis of gender differences in residual stroke risk and major bleeding in patients with nonvalvular atrial fibrillation treated with oral anticoagulants. Am. J.Cardiol.2014: 113; 485-490