We retrieved 3432 records from three databases (Fig. 1). After duplicate removal, 2647 records were screened, of which 291 were assessed for eligibility. A total of 36 different primary studies (published in 36 publications/reports) were included in the systematic review [11, 12, 29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62]. At the end of the screening phase, the kappa coefficient was 0.715 (95% CI 0.670–0.760). Upon completion of the selection by full-text reading, the kappa coefficient reached 0.951 (95% CI 0.896–1).

PRISMA flow diagram for primary study selection

A comprehensive description of the included primary studies is presented in Table 1, and their risk-of-bias assessment is detailed in Table S4 (see ESM). A graphical representation of the risk-of-bias assessment has been illustrated in the summary and traffic-light plots, in Fig. 2 and Fig. S1 (see ESM), respectively.

Global summary plot for risk of bias based on the Newcastle-Ottawa Scale (NOS)

The included studies were performed in 14 different countries, mainly in Asia (n = 22; 61.1%) [11, 29,30,31, 35, 39, 43, 44, 47,48,49,50,51,52,53,54,55,56,57, 59, 61, 62] and Africa (n = 9; 25.0%) [12, 32, 34, 36,37,38, 40, 45, 46]. Additionally, there were studies from Europe (n = 2; 5.6%) [41, 58], South America (n = 1; 2.8%) [33], and Africa (n = 1) [60]. Based on the World Bank’s country classifications by income level for 2022–2023, 16 (44.4%) studies were reported in high-income countries, 12 (33.3%) in lower-middle-income countries, six (16.7%) in low-income countries, and two (5.6%) in upper-middle-income countries. One study was conducted in “resource-limited countries”, without specifying [42]. All studies were published post-2007, except for one from the year 2000 [41]. All included studies had a prospective cohort design, with one of them employing a mixed design that also incorporated a retrospective approach [36].

The studies were conducted in a diverse range of settings, including academic and research centres, hospitals, clinics, and other practice sites. Eight studies did not provide sufficient details regarding the monitoring setting [43, 50,51,52,53,54, 59, 62]. Most studies encompassed a wide variety of therapeutics regimens, reporting on at least one drug from the categories of ART, direct-acting antivirals (DAA), integrase inhibitors (INI), nucleoside reverse transcriptase inhibitors (NRTI), non-nucleoside reverse transcriptase inhibitors (NNRTI), and protease inhibitors (PI) [11, 12, 29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49, 60,61,62].

Most of the studies employed DEM as their active pharmacovigilance strategy, with only three utilising sentinel sites [12, 59, 61] and one a registry [42]. Twelve studies used one clinical data source [35, 40, 42,