19 March 2015
In their updated systematic review and meta-analysis, Wang and colleagues  identified 12 cohort studies, 5 case-control studies, and 1 randomized clinical trial (RCT). Based on an aggregate measure emanating from adjusted risk and hazard ratios, they reported that, within a meaningful degree of heterogeneity, non-steroidal anti-inflammatory drugs (NSAIDs) are statistically significantly associated with a 28% reduced risk of Alzheimer’s disease (AD) in the general population. The authors examined for the effect of moderating factors such as publication year, population origin, sample size, percentage of female participant, percentage of low education, and qualities of studies on the heterogeneity. They also posited that other factors, such as age of participant, drug dosage, and adjusted variables might potentially account for the observed heterogeneity. We agree with their approach, yet a few points remain that need to be addressed. For example, the heterogeneity could be due to inclusion of studies with mixed methodological and statistical factors. Heterogeneity could be accounted by the type of statistics included, risk ratio (RR) or hazard ratio (HR); by the diagnostic approach, Diagnostic and Statistical Manual of Mental Disorders (DSM) or the National Institute of Neurological and communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA); or the weight given to number of adjusting variables. While the authors raise and examine this last issue, the heterogeneity remained relatively unaccounted in their analysis. In this instance, two approaches could be considered: 1) run a random effect meta-regression using the number of variables included for adjusting the statistics of interest (e.g., RR and HR); or 2) examine the global analysis and include only studies that at least report consistent variables for adjustment (e.g., age, gender, education). Of note, while it is challenging to keep all studies without giving weight to each adjusting variables, by controlling or keeping adjustment consistent, this should minimize methodological variation.
We examined the studies that are included in this meta-analysis, and focused our attention to the 12 cohort studies, given that they hold the larger set of data and minimized study design variation. Systematically, by using the above-mentioned variables, we identified a sample that would be more homogenous. Thus, A) excluded studies that did not at least report on age, education, and gender as adjusting variable; B) excluded studies using DSM criteria for diagnosis of AD given the heterogeneity in criteria (version III-R and IV); and finally C) separated studies that report RR versus HR. From the 12 cohort studies, we shortened to 7 based on adjustment variables and diagnostic criteria for AD, and then to 2 groups, one consisting of 5 cohort studies reporting on adjusted HR and the second, including 2 studies reporting on adjusted RR. At this stage, accounting for more of the possible heterogeneity as a result of multiple exclusion factors, the sample size (n = number of studies included in the analysis) become so small, that examination of further moderating variables brings Type 1 error. However, at any step of the way, an analysis could potentially show, explain, and lower the heterogeneity.
Given the challenge of being unable to effectively rely on the existing data to adequately inform us of the true benefits or harms associated with the use of NSAIDs for the reduction of risk for AD, we encourage any future observational studies to take into consideration the above-mentioned factors and carefully examine for treatment duration, something that was briefly touched upon by the meta-analysis. Furthermore, in line with the Wang study, we encourage comparison between NSAIDS and other compounds via RCTs, given the limitations of the existing evidence supporting that NSAIDs protect against incidence of AD.
Amir A. Sepehry1 and Philip E. Lee2
1Division of Neurology, Department of Medicine, University of British Columbia, Vancouver, Canada; E-mail: firstname.lastname@example.org
2Division of Geriatric Medicine, Department of Medicine, University of British Columbia, Vancouver, Canada.
 Wang J, Tan L, Wang HF, Tan CC, Meng XF, Wang C, Tang SW, Yu JT (2015) Anti-inflammatory drugs and risk of Alzheimer's disease: an updated systematic review and meta-analysis. J Alzheimers Dis 44, 385-396.