Introduction

The commencement of antiretroviral therapy (ART) in human immunodeficiency virus (HIV)-infected individuals has significantly reduced their morbidity and mortality [1]. However, HIV infection and the use of ART have been associated with a higher incidence of aging complications, such as osteoporosis [2]. According to the World Health Organization, osteoporosis is diagnosed when T-score is ≤ − 2.5; and osteopenia is diagnosed when T-score is between − 1 and − 2.5. The causes of osteopenia/osteoporosis in HIV-infected individuals are multifactorial, and may be due to a complex interaction between HIV infection, traditional osteoporosis risk factors, and ART factors [3]. If osteoporosis is left untreated, it can lead to fragility fractures.

A meta-analysis by Brown and Qadish, 2006, found that the odds ratio (OR) for HIV-infected individuals to develop osteopenia/osteoporosis was 6.4 times greater than HIV-uninfected individuals, whilst the OR for ART-treated individuals was 2.5 times when compared to ART-naïve individuals [4]. In protease inhibitor (PI)-treated individuals, the OR to develop osteopenia/osteoporosis was 1.5 times greater than PI-untreated individuals [4]. Since then, an additional 13 cross-sectional and 6 longitudinal studies have been published from 2005 till 2015. Therefore, the aim of our meta-analysis was to review published literature on the prevalence of osteopenia/osteoporosis and percent change of bone mineral density (BMD) in HIV-infected, ART-treated, PI-treated, and tenofovir-treated individuals. Factors associated with osteopenia/osteoporosis were also examined.

Materials and methods

Study selection and search strategy

This meta-analysis was registered with PROSPERO (CRD: 42016047294) and conducted according to PRISMA Guidelines (Supplementary material 1: PRISMA checklist). A search was conducted using six databases: MEDLINE, CINAHL, EMBASE, Science Direct, Cochrane, and Web of Science from 1989 till May 2015. A combination of Medical Subject Heading (MESH) and free text terms were used to define the search (“HIV”[MeSH Terms] OR “HIV”[Text Word] OR “Human Immunodeficiency Virus”[Text Word]) AND (“Antiretroviral Therapy, Highly Active”[MeSH Terms] OR “HAART therapy”[Text Word] OR “HIV Protease Inhibitors”[Pharmacological Action] OR ‘tenofovir’) AND (“Osteoporosis”[MeSH Terms] OR “Osteoporosis”[Text Word] OR “Bone Diseases, Metabolic”[MeSH Terms] OR “Bone Density”[MeSH Terms] OR “Bone Density”[Text Word] OR “osteopenia*”[Text Word]) AND (epidemiologic studies”[MeSH Terms] OR “epidemiology”[MeSH Terms] OR “epidemiology”[Text Word]) OR “prevalence”[MeSH Terms] OR “prevalence”[Text Word] OR “incidence”[MeSH Terms] OR “incidence”[Text Word]). The results of the above search strategies were combined to yield a pool of preliminary studies. Reference mining and related citations of potential references were also examined. Duplicated studies with the same title and author were excluded.

Cross-sectional or longitudinal studies published in English, original research articles that used dual-energy X-ray absorptiometry (DXA) to measure BMD on the lumbar spine, femoral neck, or total hip, and compared at least two groups (e.g., HIV-infected versus HIV-uninfected, ART-treated versus ART-naive, PI-treated versus non PI-treated, tenofovir-treated versus non tenofovir-treated individuals), aged ≥ 18 years old, and used a validated conversion equation if BMD were measured using different DXA machines, were included. Longitudinal studies were only included if change in BMD were reported > 12 months from baseline as change in BMD has to be more than the least significant change of the DXA machine (This can only occur 1–2 years from the previous DXA scan) [5]. Studies were excluded if the outcomes of interest (BMD or T-score) were not reported and if the study was only published as an editorial, commentary, brief report, expert opinion, case study, or conference abstracts. Articles that studied HIV and chronic viral hepatitis (Hepatitis B or Hepatitis C co-infection) were also excluded, as a review was recently published in 2014 [6].

Quality assessment and data extraction

The quality of each study was assessed independently by two teams of researchers using the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies 2014, developed by the National Institutes of Health, USA [7]. Data were extracted and recorded in a standardized extraction form. When disagreements occurred between the two teams, both teams met and referred back to the original text to clarify the issue that was raised. Once the evidence was highlighted, the two teams then reached a consensus via discussion.

If authors only reported osteopenia/osteoporosis as continuous variables, they were contacted and asked to provide information on the proportion of participants with osteopenia/osteoporosis. Studies were excluded if authors did not provide the required data. The prevalence of osteopenia/osteoporosis included information from cross-sectional and longitudinal studies at baseline. In longitudinal studies, the percent change of BMD from baseline to follow-up was examined.

Outcome and analysis

The primary outcome of this meta-analysis was the prevalence of osteopenia/osteoporosis and percent change of BMD. A fixed effects model was implemented to examine the heterogeneity between studies. Funnel plots were used to investigate for potential publication bias along with Begg’s test. Analyses were conducted using Review Manager (RevMan) Version 5 (Copenhagen, Denmark).

The quality of each study was assessed using the United States Preventative Services Task Force (USPSTF) guideline. A “good” study meets all criteria for that study design; a “fair” study does not meet all criteria but is judged to have no fatal flaw that invalidates its results; and a “poor” study contains a fatal flaw [8]. Strength of evidence regarding the association between risk factors and BMD was analyzed according to Supplementary material 2.

Results

A total of 21 cross-sectional and 8 longitudinal studies met our inclusion criteria (Fig. 1). No publication bias was observed in the funnel plots (Supplementary material 3).

Fig. 1
figure 1

Flow chart of studies included in meta-analysis. *Authors were contacted for additional information regarding the proportion of reduced BMD, but none responded. Hence, these studies were excluded

Full size image

HIV-infected versus HIV-uninfected individuals

Seventeen studies [cross-sectional (n = 14); longitudinal (n = 3)] compared HIV-infected versus HIV-uninfected individuals (Tables 1 and 2) [1, 9,10,11,12,13,14,15,16,17,18,19,20,21,22, 30, 31]. Seven studies only included men [1, 9, 11, 15, 20, 22, 30], four studies only included women [10, 14, 21, 31], while six studies included both men and women [12, 13, 16,17,18,19]. The majority of participants were male (range: 30–86%) [12, 16]. All studies were matched for gender except for three studies [16, 18, 19]. Age and BMI were well matched between HIV-infected versus HIV-uninfected individuals. Three studies recruited only Caucasians [9, 11, 17], nine studies recruited participants of mixed ethnicity [1, 10, 12, 14,15,16, 19, 30, 31], while five studies did not specify the ethnicity group [13, 18, 20,21,22].

Table 1 Baseline characteristics of cross sectional studies (n = 21) and longitudinal study (n = 1) and proportion of osteopenia or osteoporosis in HIV-infected individual and HIV-uninfected individuals, antiretroviral-treated, and antiretroviral-naive individuals, in protease inhibitor-treated and non protease inhibitor-treated individuals, and in tenofovir-treated and non tenofovir-treated individuals
Full size table
Table 2 Baseline characteristics of longitudinal studies (n = 8) and percent change in bone mineral density from baseline to follow-up in HIV-infected individual and HIV-uninfected individuals, antiretroviral-treated and antiretroviral-naïve individuals, in protease inhibitor-treated and non protease inhibitor individuals, and in tenofovir-treated and non tenofovir-treated individuals
Full size table

Odds ratio of osteopenia/osteoporosis in HIV-infected individuals versus HIV-uninfected individuals

Fourteen cross-sectional studies and one observational longitudinal study were included in this analysis (Table 1a) [1, 9,10,11,12,13,14,15,16,17,18,19,20,21,22]. Two longitudinal studies were excluded as the proportion of osteopenia/osteoporosis at baseline was not reported [30, 31]. The odds of developing osteopenia/osteoporosis at the lumbar spine and hip was OR = 2.4 (95%Cl 2.0, 2.8), p < 0.001 and OR = 2.6 (95%Cl 2.2, 3.0), p < 0.001, respectively (Fig. 2). The overall assessment of heterogeneity between studies for osteopenia/osteoporosis at lumbar spine and hip was I 2 = 83% (Q = 80.7, p < 0.001), and I 2 = 85% (Q = 92.7, p < 0.001), respectively.

Fig. 2
figure 2

Odds ratio of osteopenia/osteoporosis in HIV-infected versus HIV-uninfected individuals at a lumbar spine and b hip

Full size image

Percent change in BMD from baseline to follow-up in HIV-infected individuals versus HIV-uninfected individuals

Two observational longitudinal studies were included in this analysis (Table 2a) [30, 31]. Duration of follow-up ranged from 24 to 72 months [30, 31]. Bone loss only occurred at the total hip from baseline to 12 months [31] and 72 months [30]. However, when a meta-analysis was performed from baseline to 24–72 months, no significant difference was seen at the lumbar spine [OR = 4.4 (95%Cl 0.8, 26.2), p = 0.22] and total hip [OR = 0.6 (95%Cl 0.1, 4.6), p = 0.61] in HIV-infected and HIV-uninfected individuals (Supplementary material 4). The overall assessment of heterogeneity between studies for percent change in BMD analysis at lumbar spine and hip was I 2 = 34% (Q = 1.5, p = 0.16,) and I 2 = 0% (Q = 0.3, p = 0.62,), respectively.

Antiretroviral-treated versus non antiretroviral-treated individuals

Nine studies [cross-sectional (n = 8); longitudinal (n = 1)] compared ART-treated versus non ART-treated individuals (Table 1) [9, 13, 18, 23,24,25,26,27,28]. Five studies only included men [9, 24,25,26, 28], one study only included women [27], while two studies included both men and women [13, 18]. The majority of participants were male (range: 70–91%) [13, 18]. Two studies were not matched for gender [13, 18]. One study recruited only Caucasians [9], while another study recruited both Caucasians and Blacks [27].

Odds ratio of osteopenia/osteoporosis in antiretroviral-treated versus antiretroviral-naive individuals

Eight cross-sectional studies and one observational longitudinal study were included in this analysis (Table 1b) [9, 13, 18, 23,24,25,26,27,28]. The odds of developing osteopenia/osteoporosis at the lumbar spine and hip was OR = 2.8 (95%Cl 2.0, 3.8), p = 0.004 and OR = 3.4 (95%Cl 2.5, 4.7), p = 0.0002, respectively (Fig. 3). The overall assessment of heterogeneity between studies for osteopenia/osteoporosis at lumbar spine and hip was I 2 = 64% (Q = 22.2, p < 0.001) and I 2 = 74% (Q = 30.4, p < 0.001), respectively.

Fig. 3
figure 3

Odds ratio of osteopenia/osteoporosis in antiretroviral-treated and antiretroviral naive individuals at a lumbar spine and b hip

Full size image

A meta-analysis on the percent change in BMD from baseline to follow-up for ART-treated individuals could not be calculated as there was only one observational longitudinal study [18].

Protease inhibitor-treated versus non protease inhibitor-treated individuals

Eleven studies [cross-sectional (n = 7); longitudinal (n = 4)] compared PI-treated versus non PI-treated individuals (Tables 1 and 2) [13, 18, 20, 24, 25, 27,28,29, 32,33,34]. Of the four longitudinal studies, only one was a randomized controlled trial (RCT) [33]. Four studies only included men [20, 24, 25, 28], one study only included women [27], while six studies included both men and women [13, 18, 29, 32,33,34]. The majority of participants were male (range: 63–96%) [18, 29]. All studies were matched for gender except for five studies [13, 18, 29, 32, 33]. Age and BMI were well matched between PI-treated and non-PI-treated individuals.

Odds ratio of osteopenia/osteoporosis in protease inhibitor-treated versus non protease inhibitor-treated individuals

Seven cross-sectional studies and one observational longitudinal study were included in this analysis (Table 1c) [13, 18, 20, 24, 25, 27,28,29]. Three longitudinal studies were excluded as the proportion of osteopenia/osteoporosis at baseline was not reported [32,33,34]. The odds of developing osteopenia/osteoporosis at the lumbar spine and hip was OR = 1.3 (95%Cl 1.0, 1.8), p = 0.20 and OR = 1.3 (95%Cl 1.0, 1.7), p = 0.17, respectively (Supplementary material 5). The overall assessment of heterogeneity between studies for osteopenia/osteoporosis at lumbar spine and hip was I 2 = 29% (Q = 9.8, p = 0.05) and I 2 = 33% (Q = 10.4, p = 0.07), respectively.

Percent change in BMD from baseline to follow-up in protease inhibitor-treated versus non protease inhibitor-treated individuals

Four longitudinal studies [observational (n = 3); RCT (n = 1)] were included in this analysis (Table 2b) [18, 32,33,34]. The duration of follow-up ranged from 14 to 33 months [18, 33]. Bone loss occurred at both lumbar spine from baseline to 6 [33], 9, 21 [32], and 33 months [33] and at the femur from baseline to 11 [33], 14 [18], and 33 months [33]. However, there was no significant difference between the percent change in BMD at the lumbar spine [OR = 1.1 (95% Cl 0.5, 2.7)], p = 0.70 and femoral neck [OR = 1.2 (95% Cl 0.4, 3.8)], p = 0.53 from baseline to 14–33 months in PI-treated versus non PI-treated individuals (Supplementary material 6). The overall assessment of heterogeneity between studies for percent change in BMD analysis at lumbar spine and hip was I 2 = 0% (Q = 1.5, p = 0.82) and I 2 = 0% (Q = 0.4, p = 0.77), respectively.

Tenofovir-treated versus non tenofovir-treated individuals

Three studies [cross-sectional (n = 1); longitudinal (n = 2] compared tenofovir-treated versus non tenofovir-treated individuals (Tables 1 and 2) [29, 35, 36]. Only one longitudinal study was a RCT [36]. All three studies included both men and women [29, 35, 36]. The majority of the participants were male (ranged 74–99%) [29, 35], and all studies were not matched for gender [29, 35, 36].

Proportion of osteopenia/osteoporosis in tenofovir-treated versus non tenofovir-treated individuals

Only one cross-sectional study was included in this analysis (Table 1d) [29]. Hence, it was not possible to calculate the OR. A higher proportion of tenofovir-treated individuals 30 (52.6%) had osteopenia/osteoporosis compared to non tenofovir-treated individuals 35 (42.7%), but this result was not statistically significant (p = 0.248).

Percent change in BMD from baseline to follow-up at in tenofovir-treated versus non tenofovir-treated individuals

Two longitudinal studies [observational (n = 1); RCT (n = 1)] were included in this analysis (Table 2c) [35, 36]. The duration of follow-up ranged from 22 to 33 months [35, 36]. Bone loss occurred at the lumbar spine and total hip from baseline to 11, 22 [36], and 33 months [35]. However, when a meta-analysis was performed, no significant difference was found between the percent change in BMD at the lumbar spine [OR = 1.0 (95% Cl, 0.2, 5.8)], p = 0.38 and total hip [OR = 1.8 (95% Cl, 0.4, 8.7)], p = 0.71 from baseline to follow-up at 22–33 months in tenofovir-treated versus non tenofovir-treated individuals (Supplementary material 7). The overall assessment of heterogeneity between studies for percent change in BMD analysis at lumbar spine and hip was I 2 = 0% (Q = 0.8, p = 1.00) and I 2 = 0% (Q = 0.14, p = 0.4), respectively.

Risk factors for low BMD

Fifteen studies assessed the risk factors associated with low BMD (Table 3) [1, 9,10,11, 14, 16, 17, 23,24,25,26,27, 29, 31, 33]. Thirteen studies were rated as fair quality as they did not perform a sample size calculation [1, 9,10,11, 14, 16, 23,24,25, 27, 29, 31, 33]. Two studies were rated as poor quality as they did not perform any multivariate analysis [17, 26]. Ten risk factors (age, history of bone fracture, BMI, body weight, ethnicity, testosterone level, smoking, lipodystrophy, CD4 cell count, fat mass, and lean body mass) had fair evidence of an association with low BMD, two risk factors (HIV viral load and lactic acid level) had insufficient evidence, while three risk factors (steroid use, opiate use, and vitamin D level) had inconsistent evidence.

Table 3 Risk factors of osteopenia or osteoporosis in HIV infected individuals
Full size table

Discussion

Our findings from the 21 cross-sectional and 8 longitudinal studies found that odds of developing osteopenia/osteoporosis in HIV-infected and ART-treated individuals were approximately two times more when compared to their respective controls. No significant difference was found in the odds of developing osteopenia/osteoporosis in PI-treated and tenofovir-treated when compared to their respective controls. Older age, history of bone fracture, low BMI, low body weight, being Hispanic or Caucasian, low testosterone level, smoking, low CD4 cell count, lipodystrophy, low fat mass, and low lean body mass were associated with low BMD.

The pathogenesis of bone loss in HIV-infected individual is a complex and multifactorial process. This is probably due to an interaction within the T cells, osteoclast, and osteoblast, which is more apparent in HIV-infected individuals, as well as those on ART [3]. HIV infection increases inflammatory cytokines, which has been thought to increase bone turnover through osteoclast stimulation and bone resorption [37]. Our study found that the odds of developing osteopenia/osteoporosis in HIV-infected individuals were two times lower than a previous meta-analysis [4]. This difference could be due to the additional seven cross-sectional studies [1, 10, 11, 15, 16, 19, 22]. Additionally, we excluded three studies as two studies were brief reports [38, 54] and the other included participants with hepatitis C co-infection [39].

The introduction of ART has increased the life expectancy of HIV-infected individuals [40], but the mechanism of action of ART on bone loss remains controversial [4, 24]. ART-treated individuals were found to have a two times increased risk of developing osteopenia/osteoporosis, which was similar to previous findings [4]. ART has been reported to affect osteoclast and osteoblast activity in vitro [41] and in animals [42]. However, this finding could not be replicated in humans [43]. NRTIs (like abacavir and zidovudine) were found to suppress osteoblast activity [41] and promote osteoclastogenesis/osteoclast activity in animal studies [42].

The proportion of PI-treated and tenofovir-treated individuals with osteopenia/osteoporosis were higher when compare to their respective controls. However, these results did not reach statistical significance, which was similar to previous findings [4]. PIs (like ritonavir) were reported to suppress osteoclastogenesis/osteoclast function in vitro and in vivo studies [44]. Tenofovir was reported to cause proximal renal tubular dysfunction that could result in excessive renal phosphate loss, which then impairs bone mineralization and increased bone turnover leading to bone loss [45].

Our study found that there was no significant difference in the percent change of BMD at the lumbar spine, femoral neck or total hip from baseline to follow-up at 14–33 months between HIV-infected, PI-treated, and tenofovir-treated versus their controls, respectively. Our meta-analysis suggests that bone loss may occur within the first year of HIV infection or ART initiation. Our results were similar to previous studies which reported that 2–6% of bone loss occurred over the first 2 years of ART initiation [3]. However, evidence concerning short-term bone loss may not be as strong as the range of bone loss reported was small (0.1–5.0%) as our findings were based on a small number of studies [46]. This shows that the pattern of bone loss in HIV-infected individuals may be acute and accelerated, while in older people, their bone loss occurs gradually over time as they age [47].

Ideally, results from RCT and observational studies should be analyzed separately as observational studies which recruited ART-treated individuals are more likely to have active disease or complications that may impact on outcome, whereas RCTs match for baseline conditions for both treated individuals and controls. However, it was not possible for us to perform any sub-group analysis based on study design as there were only two RCTs included.

We recommend that HIV-infected and ART–treated individuals should be screened for osteopenia/osteoporosis during the first year of HIV-infection or ART initiation, regardless of age or gender. In HIV-infected individuals aged < 40 years old, DXA scan is recommended as a screening tool for osteopenia/osteoporosis since the Fracture Risk Assessment Tool (FRAX®) can only be used in those aged ≥ 40 years. FRAX should be used to screen for osteopenia/osteoporosis in those ≥ 40 years, as it is the most cost effective method. DXA scan is recommended in HIV-infected: individuals aged ≥ 40 years old who have a FRAX score ≥ 10%; men aged ≥ 50 years, post-menopausal women, individuals with a history of fragility fracture, individuals receiving chronic glucocorticoid treatment, and individuals who are at higher risk of falls as previous recommended [48].

Traditional osteoporosis risk factors play an important role in HIV-infected individuals [49]. However, HIV-infected individuals have additional risk factors such as low fat mass, low lean body mass, lipodystrophy, and low CD4 cells count which contributes to bone loss. Studies found that changes of body composition from HIV infection and complications of ART mimic the normal aging process of older individuals [50]. As a result, HIV-infected individuals are more susceptible to have lower lean body mass and lower fat mass/abnormal body fat redistribution which can lead to bone loss [50]. Both lean body mass and fat mass are independent and significant predictors of BMD [51]. Low lean body mass leads to reduce muscle strength, which can affect bone mass and structure [51]. Lipodystrophy is a medical condition that is characterized by fat loss and/or redistribution of body fat [52]. This disorder is frequently observed in HIV-infected individuals on long term ART [52]. To date, the underlying etiology of low CD4 cell count and bone loss remains unclear. However, it has been suggested that the immune system may play a potential role in skeletal maintenance [53]. Evidence shows HIV-infected individuals with low baseline CD4+ cell count (< 50 cells/mm3) have approximately 3% greater bone loss than HIV-infected individuals with > 500 cells/mm3 [53].

One of the limitations of our study was that we were not able to determine the effect of the individual types of antiretroviral medications on osteopenia/osteoporosis. Secondly, we excluded studies that were published as an editorial, commentary, brief report, expert opinion, case study, or conference abstracts as these reports were not sufficiently detailed enough for us to assess quality and data extract. Lastly, we were not able to utilize Z-score to diagnose osteoporosis in individuals < 50 years, as the majority of the studies included did not present their results as Z-score. The strength of our study was that we performed our search on six databases.

Conclusions

The prevalence of osteopenia/osteoporosis in HIV-infected and ART-treated individuals was approximately two times more compared to controls. Evidence concerning bone loss within the first year of HIV infection and ART initiation was preliminary. Older age, history of bone fracture, low BMI, low body weight, being Hispanic or Caucasian, low testosterone level, smoking, low CD4 cell count, lipodystrophy, low fat mass, and low lean body mass were associated with low BMD.