Background

In the era of combination antiretroviral therapy (cART), the overall life span of people living with HIV (PLWHIV) on ART is now almost similar to that of uninfected individuals [1, 2]. However, disparities in treatment outcomes still exist in low-income and high-income countries [3]. Amongst ART-treated patients, the most prevalent comorbidities include cardiovascular disease (CVD) and hepatic disorders [4]. CVD remain a challenge in the management of PLWHIV, with a persistent gap in the reported survival outcomes. The paucity of clinical trials focusing on cardiovascular outcomes of patients on long-term ART limits the availability of evidence-based patient management strategies. In a previous meta-analysis of cardiac dysfunction in PLWHIV, left ventricular systolic and diastolic dysfunction was prevalent despite successful ART [5]. Moreover, the levels of inflammation were a strong predictor of systolic dysfunction [5]. Contradictory findings exist on how ART modifies the conventional biomarkers of inflammation [6], and dyslipidaemia [6,7,8]. These inconsistent findings further complicate efforts of delineating the precise mechanisms, which link the interplay between the immune continuum and CVDs in PLWHIV on ART. To date, reliable findings on ART induced changes on CVD-risk markers remain elusive. Notably, several studies have reported on the pivotal role that platelets play in linking inflammation and CVD.

Incongruent findings on the levels of platelets activation in PLWHIV have been reported in several observational studies [6,7,8,9]. However, the impact of ART on circulating platelets remains unclear. Notably, antiretroviral drugs have been shown to directly activate platelets, in vitro [10], while individual observational studies have reported on conflicting findings of attenuated [6, 11] and persistently elevated levels of platelet activation following initiation of ART [7,8,9, 12, 13]. However, the importance and clinical relevance of evaluating platelet activation in the thrombotic-risk stratification of PLWHIV on ART is confounded by variance in the study setting, duration of ART exposure and differences in the methods of enumerating activated peripheral blood platelets.

This systematic review and meta-analysis, therefore, aimed at providing a comprehensive synthesis of studies reporting on the levels of platelets activation in adult patients living with HIV. The primary objective of this study was to determine whether the levels of platelet activation are elevated in treatment naïve patients with HIV, while the secondary objective was to determine whether the levels of platelet activation are attenuated, following successful ART.

Methods

This systematic review and meta-analysis was prepared and conducted following the preferred reporting items for systematic reviews and meta-analysis (PRISMA) statement [14]. The review was registered on the prospective register of a systematic reviews registry (PROSPERO) (registration number: CRD42017062393), and the study protocol was published [15]. A comprehensive and systematic search of published studies was conducted to address the following research questions:

  1. 1.

    Are platelets activated in patients living with HIV?

  2. 2.

    Does successful ART attenuate the levels of activated platelets in patients with HIV?

Sources of evidence and search strategy

A comprehensive search was performed on the MEDLINE, Academic search complete, CINAHL with full-text, Health Source: Nursing/Academic edition, and APA Psycinfo databases using the EBSCOHOST search engine. We further searched the Cochrane Central Register of Clinical Trials (Wiley interface), the metaregister (www.controlled-trials.com/mrct/), ISI web of science, and the Global Index of Medicus. Two reviewers (BBN and PVD) independently searched the electronic databases from inception through the 30th of November 2019 using the following search terms: “Platelets” OR “Thrombocytes” OR “Platelet P-selectin” OR “Platelet CD40L” OR “Platelet monocyte aggregates” OR “Platelet leukocytes aggregates” AND “HIV” OR “HIV-1” AND “Antiretroviral therapy”. We further scanned the reference list of selected studies for additional relevant studies. No language restrictions were applied.

Study selection

The screening of the titles and abstracts of all studies reporting on platelet activation in HIV-infected patients on antiretroviral therapy was independently conducted by three reviewers (BBN, TMN and PVD). Reviews, case reports and pre-clinical studies were excluded. Abstracts from conference proceedings as well as grey literature were also excluded, due to the reported inconsistencies between findings reported in conference abstracts and full publications [16]. The selected studies were also scanned for potential duplicate data publications that may arise from overlapping cohorts reported in different publications. Studies were selected and included in the meta-analysis based on the availability of study-level data required for the effect size estimation.

Data extraction and management

Two independent reviewers (BBN and PVD) extracted detailed study information and characteristics using a predefined standardised data extraction form. The extracted data items included details of the author, year of publication, original language, sample size, years of follow-up, effect measures reported, gender ratio, levels of coagulation markers, platelet counts, levels of inflammatory markers, mean CD4 counts, HIV-1 RNA levels and HAART regimen. We further extracted the methods used to determine HIV-1 RNA, platelet function and the methods of platelet separation. The extracted data was cross-checked, and inconsistencies were resolved by discussion or referred to a third reviewer (MZ) for arbitration.

Outcomes

The primary outcomes assessed included levels of platelet activation: reported as the standardised mean difference in soluble platelet P-selectin (sCD62P) or P-selectin (CD62P) levels and soluble CD40L (sCD40L).

Assessment of risk of bias in included studies

The quality of the included studies was independently assessed by two reviewers (ZM and VM) using the modified Downs and Black checklist which is suitable for evaluating both RCTs and non-controlled trials [17]. The checklist assesses four domains which include (I) reporting bias, (II) external validity, (III) internal validity and (IV) selection bias. The overall scores were graded as excellent (26–28), good (20–25), fair (15–19) and poor (≤ 14). These corresponded to those previously reported [18]. Reviewer scores were cross-checked by two reviewers (TMN and VM) and discrepancies were resolved either through discussions or arbitration by BBN.

Data synthesis and statistical analysis

Descriptive data items on platelet function in adult HIV-infected patients were summarised in studies which inadequate data for a meta-analysis. The Kappa Cohen’s Kappa (ĸ) was used to assess the inter-rater agreement of the reviewer’s quality agreements and scored as no agreement (ĸ ≤ 0.00); none to slight (ĸ = 0.01–0.20); fair (ĸ = 0.41–6.0); substantial (ĸ = 0.61–0.80) and almost perfect (ĸ = 0.81–1.00).

Sensitivity analysis and publication bias

Heterogeneity was assessed using the I2 statistic and a value of 50% was considered as substantial heterogeneity [19], with a p value for heterogeneity testing set at < 0.05. Continuous outcomes were pooled as standardised mean differences (SMD), and due to the few number of studies included in the meta-analysis, the Hedges’ g statistic (g) was used to correct for small study bias. In studies that reported medians and interquartile ranges, we estimated the mean as previously described [20]. Publication bias was assessed using funnel plot analysis, the Egger’s regression test. A p value of < 0.05 represented significant levels of publication bias and the trim and fill method was used to identify and compute an adjusted effect size. The sensitivity analysis was performed to test the unexplained sources heterogeneity and robustness of the reported effect estimates. We assessed the effect of each study on the overall standardised mean difference (SMD), by performing a repeated meta-analysis following the omission of a single-study at a time. All analysis was performed using STATA 16.0 (StataCorp LP, TX, USA). All p values for associations were two-sided and the value of < 0.05 was considered as significant.

Patient and public involvement

It was not appropriate or possible to involve patients or the public in the design, or conduct, or reporting, or dissemination plans of our research.

Results

Study selection

We identified 831 citations through the electronic database search (390 in MEDLINE, 284 in Academic Search Complete, 73 in CINAHL, 72 in Health Source: Nursing/Academic edition, and 8 in APA Psycinfo and 4 through other sources). After the initial duplicate removal and screening of abstracts, 827 studies were deemed irrelevant and excluded (Fig. 1). The full texts of the remaining 323 studies were assessed for eligibility, and only 30 studies fulfilled the pre-specified inclusion criteria. The 30 studies were all published in English. These included 2 randomised control trial [21, 22] and 28 non-randomised studies [6,7,8,9, 11, 12, 22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44].

Fig. 1
figure 1

Flow diagram of the study selection process

Full size image

Characteristics and quality of eligible studies

The characteristics of the included studies are shown in Table 1. In total, the included studies comprised of 2325 participants. The risk of bias for all included studies was evaluated by two independent reviewers (TMN and VM) using the Blacks and Downs checklist [17]. (Additional file 2: Table S1). The overall inter-rater reliability was assessed using Cohen’s Kappa. The levels of agreement for the various domains were scored as substantial (ĸ = 0.81) for the reporting bias and external validity, substantial (ĸ = 0.75) for internal validity and almost perfect (ĸ = 0.81) for selection bias. Majority of the included studies were rated as poor (76%, n = 27) while only 20% of the included studies were rated as fair (Additional file 2: Table S1).

Table 1 Characteristics of included studies (n = 30)
Full size table

Geographic mapping of the included studies

The included studies comprised of a total of 18 studies conducted in Europe [7, 8, 11, 12, 25,26,27,28, 33, 34, 36, 38,39,40,41, 43, 45] and 12 studies from the Americas [6, 9, 21, 22, 24, 31, 32, 37, 46]. Studies from Italy [8, 26, 33, 34, 43], Ireland [11, 36] and USA [6, 9, 21, 22, 24, 30, 32, 46] make the highest contribution (Supplementary Figure 1). The bibliometric analysis based on co-authorship links was performed to assess the potential duplication of published datasets.

Data synthesis and publication bias analysis

In total, the included studies (n = 30) comprised of 1725 HIV-infected patients and 600 HIV-negative individuals. Most of the patients with HIV (71%, n = 1230) were on various antiretroviral drugs (Table 2). The included studies also reported on various effect measures of platelet activation which included soluble P-selectin (sCD62P), soluble glycoprotein VI (sCD36), soluble CD40L and chemokine ligand 5 (RANTES) (Table 3). In nine of the studies, the included cohort of patients with HIV were virologically suppressed [6, 7, 31,32,33, 39] with only seven studies reporting on patients with CD4 counts < 500 cells/mm3 [8, 12].

Table 2 A qualitative summary of the findings on platelet function in antiretroviral-treated individuals in the included studies (n = 30)
Full size table
Table 3 Reported markers of platelet activation in treatment-naïve and ART-treated HIV-infected patients
Full size table

The Egger’s regression test suggested evidence of publication bias (p < 0.001) (Additional file 1: Table S2). In addition, the funnel plot analysis indicated asymmetry as the trim-and-fill method imputed no missing study for the primary platelet activation outcome (Additional file 1: Table S2 and Figure S1). Notably no missing study was imputed (based on the trim and fill method) for comparisons of platelet activation between ART-treated treatment-naïve (Additional file 1: Table S2).

Platelet activation in HIV-infected compared to uninfected patients

In the majority of the studies (61%, n = 17), the levels of platelet activation were elevated in HIV-infected patients compared to uninfected controls. The included studies reported on various surrogate markers of platelet activation; these included markers of alpha granule secretion (RANTES) and markers of platelet adhesion and activation (CD62P, CD40L and platelet-monocyte aggregates) (shown in Table 3). In order to assess whether the varying methods used modified the reported effect estimate, we performed a subgroup analysis based on the type of method used in each study. In our subgroup meta-analysis, the primary outcome of platelet activation in HIV-infected patients was compared to uninfected controls. We observed a quantitative effect based on the methodology used to enumerate the levels of platelet activation, whereby in studies using the Luminex, the largest effect size (Hedges’ g 3.94 [95%CI 2.68, 5.19], p < 0.001) was compared to ELISA-based technology (Hedges’ g 1.55 [95%CI 0.65, 2.44], p = 0.001) (Fig. 2b). Notably, both methods reported on similar direction of the effect and significantly increased levels of platelet activation in HIV-infected patients.

Fig. 2
figure 2

The association between platelet activation and HIV infection. a The forest plot shows the pooled effect estimate of platelet activation in treatment-naïve HIV-infected patients compared to uninfected controls. b The subgroup effects based on the methodological differences in the included studies

Full size image

In all, elevated levels of the surrogate markers of platelet activation, CD40L, P-selectin (CD62P) and the soluble form (sCD62P and sCD40L), were consistently reported in HIV-infected patients [7,8,9, 11, 12, 23, 24, 27, 30, 31, 33, 37,38,39, 41, 43, 45]. While a few studies (17%, n = 2) reported on comparable levels of sCD40L [8], RANTES [26], sCD62P and CD62P [6, 47]. However, the random-effects model, which included 8 studies reporting on 409 treatment-naïve HIV-infected patients and 138 uninfected controls, demonstrated an overall increase in the levels of platelet activation in HIV-infected patients compared to uninfected controls (Fig. 2). The overall bias-corrected standardised mean difference (Hedges’ g) in the levels of platelet activation in treatment-naïve HIV-infected patients compared to uninfected controls was 2.00 [95%CI 1.05, 2.94]; z = 4.12, p < 0.001. Notably, the levels of heterogeneity were high (I2 = 95.91%, p < 0.01). Therefore, we performed a subgroup analysis to explore the sources of heterogeneity based on the reported marker of platelet activation. The test for subgroup effects showed a significant subgroup effect (p < 0.001) (Fig. 2a). This suggests that the reported pooled effect estimate was influenced by the various markers of platelet activation reported in the included studies.

Platelet P-selectin levels in treatment-naïve HIV-infected patients

A total of 16 included studies reported on the levels of P-Selectin (CD62P) in untreated HIV-1-infected patients compared to ART-treated and uninfected controls [6,7,8, 11, 12, 21, 28, 29, 31, 32, 35, 37, 39,40,41]. However, only two studies [12, 41] had adequate study-level data and were included in the subgroup meta-analysis. The pooled effect estimate included 146 participants comprising 101 HIV-infected patients and 45 uninfected controls. Interestingly, the sCD62P levels were elevated in patients with HIV when compared to uninfected controls (Hedges’ g 0.57 [95%CI 0.22, 0.93], p = 0.002). Moreover, there were low levels of heterogeneity (I2 = 0%) (Fig. 2).

Soluble CD40L levels in treatment-naïve HIV-infected patients

A total of 384 participants were included in the subgroup meta-analysis of 6 studies. This comprised of 272 treatment-naïve HIV-infected patients and 112 uninfected controls. The levels of sCD40L were elevated in treatment-naïve HIV-infected patients compared to controls (Hedges’ g 1.89 [95%CI 0.78, 3.01], p = 0.001) (Fig. 2b). Notably, the levels of heterogeneity were high amongst the included studies (I2 = 94.38%). We, therefore, explored other sources of heterogeneity which included differences in the study design, and the methodology used to measure the levels of platelet activation (Fig. 2b).

The test for group differences showed a significant interaction effect based on the reported technique or methodology used (p < 0.01) and reported marker of platelet activation (p < 0.001) (Fig. 3b). Notably, there were no significant between-study differences on the reported pooled estimates based on the varying study designs (p = 0.10).

Fig. 3
figure 3

Platelet activation in ART treated compared to uninfected individuals. The forest plot (a) shows the pooled effect estimates of platelet activation in treated compared to uninfected patients. b The subgroup interactions based on the study design, method used to determine the levels of platelet activation and the reported platelet-associated marker

Full size image

RANTES levels in treatment-naïve HIV-infected patients

A total of 123 participants from 2 studies [43, 45] were included in this subgroup analysis (Fig. 2). In all, this included 96 treatment-naïve HIV-infected patients and 27 uninfected controls. The pooled estimate showed significantly increased levels of RANTES in treatment-naïve HIV-infected patients compared to uninfected controls (Hedges’ g 3.78 [95%CI 2.36, 5.21], p = < 0.001) (Fig. 2b). However, the levels of heterogeneity were high amongst the studies included in this subgroup (I2 = 76.21%).

Platelet activation in HIV-infected patients on antiretroviral therapy compared to treatment-naïve HIV-infected patients

Elevated levels of platelet activation, exhaustion and apoptosis persist even during successful ART [6,7,8,9, 11, 12, 21, 22, 24, 27,28,29, 31,32,33,34,35,36,37,38,39,40,41, 43, 45,46,47]. The use of NNRTIs was associated with elevated sCD40L levels [41], while others reported that ART did not affect the elevated levels of sCD40L, RANTES and urinary 11-dehydrothromboxane B2 levels [7, 21, 24, 34, 47]. In addition, plasma sCD62P and sCD40L levels were 2–3-fold higher in treated HIV-infected patients compared with treatment-naïve HIV-infected patients [7, 38]. Congruently, persistently activated elevated levels of platelet reactivity have been reported following 6 to 24 months of initiating treatment [7, 8, 22, 38,39,40,41, 47]. In contrast, three studies reported on the normalisation of platelet activation levels following 12 months of ART [11, 23, 26]. Moreover, ART-treated patients had higher levels of spontaneous platelet aggregation in response to submaximal concentrations of various endogenous platelet agonists [47].

Only 18 studies had adequate study-level data and were included in the meta-analysis [6,7,8, 11, 12, 22, 23, 26,27,28, 30, 32, 33, 35, 37,38,39,40,41, 43, 45]. The pooled estimates showed elevated levels of platelet activation in ART-treated patients compared to controls (Hedges’ g 2.05 [95%CI 0.58, 3.52]; z = 2.71, p = 0.0067) (Fig. 3), and these elevated levels were comparable between ART-treated patients with HIV and treatment-naïve patients (Hedges’ g 0.21 [95%CI − 0.69, 1.10], z = 0.83 p = 0.4080) (Fig. 4). However, these levels of heterogeneity were substantial (I2 = 98.68%). To explore the sources of unexplained heterogeneity, we then conducted a subgroup analysis based on the reported effect measure of platelet activation within the included studies (Fig. 3a).

Fig. 4
figure 4

The effects of antiretroviral therapy on the level of platelet activation in HIV-infected patients. The forest plot (a) shows the pooled effect estimate of platelet activation levels in HIV-infected patients on antiretroviral therapy (ART) compared to treatment-naïve patients. b The subgroup meta-analysis based on the study design, method used to determine the levels of platelet activation and the reported platelet-associated marker

Full size image

In all, the levels of both surface CD62P (Hedges’ g 5.10 [95%CI 1.45, 8.78], p = 0.006) and sCD62P (Hedges’ g 0.57 [95%CI 0.09, 1.04], p = 0.019) remained significantly elevated in patients with HIV on ART when compared to uninfected controls (Fig. 3). We further conducted a subgroup meta-analysis based on the study design, the technique used and reported marker of platelet activation (Fig. 3b). The test for subgroup differences demonstrated a statistically significant interaction effect, in the subgroup based on the technique (p < 0.01) and marker used to enumerate activated platelets (p < 0.09).

Discussion

To date, this is the first systematic review and meta-analysis of studies reporting on platelet activation in adult patients living with HIV. We retrieved and analysed 30 studies that reported on platelet activation in adult PLWHIV on ART. Several of these studies (50%, n = 15) reported on data obtained from serum or plasma samples using enzyme-linked immunosorbent assays. While the methods used to measure platelet activation included the detection of surface markers of platelet activation using flow cytometry (23%, n = 7), Luminex technology (10%, n = 3) and immunoassay-based methods (50%, n = 15). Platelet function was only evaluated in a few studies (1.66%, n = 2) using light transmission aggregometry. Notably, the subgroup meta-analysis showed that methodological variance may influence the reported levels of platelet activation, with studies employing the Luminex technology showing a larger pooled estimate when compared to ELISA-based methods.

In all, several studies reported on elevated levels of soluble, CD62P (sCD62P) and CD40L (sCD40L), in PLWHIV compared to uninfected controls (Table 3). This supports the finding of increased levels of platelet activation in adult HIV-infected patients [7, 8, 11, 12, 33, 41]. Moreover, in follow-up studies, these levels were shown to persist following 3 to 24 months of successful ART [8, 12, 22, 38, 47]. Although ART attenuated the levels of platelet activation in patients who were not receiving protease inhibitors (PIs) [11], in those patients on PI-based therapy, the levels of platelet activation persisted despite successful ART [21, 24, 28, 29, 39, 41, 43, 45,46,47]. Moreover, the levels of platelet activation were 2-fold higher in patients on PI-based therapy compared to treatment naïve patients with HIV [7, 30, 38]. Notably, different measures of platelet activation were reported in the included studies (Table 3); the vast majority of the studies report on congruent findings of elevated levels of platelet activation despite achieving viral suppression, although a few of the included observational studies reported on discordant findings, on the levels of CD62P and CD40L in ART-treated patients living with HIV [7, 8].

To our knowledge, this is the first pooled analysis of platelet activation levels reported in patients living with HIV. The primary objective of the planned meta-analysis was to determine the association between the levels of platelet activation in HIV-infected treatment-naïve patients. While the secondary objective of this analysis of pooled data was to assess whether effective ART attenuates the levels of platelet activation in patients living with HIV. In all, the levels of platelet activation were elevated in adult patients living with HIV when compared to uninfected controls (Fig. 2). Notably, this supports the previously reported findings of elevated levels of platelet activation in HIV-infected patients [6, 13, 36, 48]. The current study adds value in highlighting that sCD40L levels are increased at greater magnitude when compared to levels of sCD62P in treatment-naïve HIV-infected patients. These findings further highlight the potential benefit of using sCD40L and sCD62P in thrombotic-risk profiling of PLWHIV on ART, as sCD62P is associated with platelet and vascular activation.

The present meta-analysis of patients with HIV on ART and treatment-naïve infected patients suggests that ART had no effect on the levels of platelet activation, although the RANTES, sCD40L and surface CD40L levels were comparable following successful ART (Fig. 3b). In contrast, the sC62P and surface CD62P levels remained significantly increased following effective ART (Fig. 3b). Differences in the treatment regimens may account for the incongruent effect estimates. In fact, all studies reporting on patients receiving PI-based ART showed elevated levels of platelet activation despite successful viral suppression [6,7,8, 39, 47].

The strengths of this review include the comprehensive search and data extraction which was independently performed by two reviewers. The inter-rater agreement was high for most of the risk of bias. Although the included studies showed high levels of heterogeneity, the potential risk of bias of these studies was scored as fair. The presented cumulative evidence is limited by the lack of adequate randomised control trials and reporting of ART regimen used. Due to the lack of adequate number of RCTs addressing the changes in platelet activation following ART, we considered and included observational studies alongside RCTs in the meta-analysis [49]. The included studies were scored low in the external validity, thus limiting the generalisability of these findings. Lastly, the pooled effect estimates on the levels of platelet activation were derived from studies conducted mainly in the Americas and countries in Europe. Caution should be taken in extrapolating these findings into a different geographic setting.

Conclusion

This meta-analysis provides evidence that the levels of soluble CD62P and CD40L are elevated in HIV-infected patients and that sCD62P levels persist despite successful therapy. Overall, we report on elevated levels of platelet activation in adult PLWHIV which persist despite successful ART.