Get started
Work

Risk of Mortality in Patients with HIV and Depression: A Systematic Review and Meta-Analysis of a Non-Common Outcome

Year: 2017
Type: review
Abstract: Meta-analysis (MA) is increasingly being utilized to combine results of several studies to derive refined estimates. However, great difficulty is commonly encountered when non-common outcomes (NCO) are involved. This study is concerned with the MA of mortality risk attributed to depression in HIV patients to identify the challenges and solutions regarding statistical methods in MA of NCO. DerSimonian-Laird (DL) and Inverse variance (IV) techniques were used in the MA to pool estimates of mortality risk through Random Effect Models (REM) and Fixed Effect Models (FEM) depending on presence or absence of heterogeneity respectively. Series of sensitivity analyses, multi-level subgroup analyses and I-squared (I2)statistics tests were done to identify sources of bias, methodological and statistical heterogeneity respectively. From 17 studies that met the inclusion criteria, the mortality Hazard Ratio (HR) and 95% confidence interval (CI) among depressed HIV patients was 1.80 and (1.23 - 2.61) respectively, with significant statistical heterogeneity (I2 = 92.8%). Multiple population characteristics were found in subgroup analyses as sources of methodological heterogeneity (Table 3). Publication bias was present, as detected by both Egger's and Begg's tests, some studies were excluded in succession as sensitivity analysis was carried out sequentially. The final estimate was 1.56 (HR) and CI (1.33 - 1.83) without statistical heterogeneity (I2 = 0.0%). As the studies were being excluded there was a 64% reduction in the width of the CI. From the results of this paper, MA of NCO is associated with publication bias, significant statistical and methodological heterogeneity.
Language: English
Open Access status: diamond
OpenAlex ID: W2739240961
FWCI: 0.048
Cites: 24
Cited by: 1
Related to: 10