Advances in the treatment of acquired immune deficiency syndrome (AIDS) impacts directly on the demographic profile of the infection. Highly active antiretroviral therapy (HAART) is increasing gradually life expectancy of people living with human immunodeficiency virus (HIV), allowing a better understanding of associations between the virus and the changes inherent to the aging process [1-4].

Bone mineral density (BMD) reduction is a known finding related to aging that has been discussed for decades as the main pathophysiological drive of osteoporosis - a disease in which bone quality, density, and resistance are reduced, resulting in an increased risk of fractures [5].

Osteoporosis is defined by a BMD reduction until 2.5 (or more) standard deviation (SD) below the average for the Caucasian population using dual energy X-ray absorptiometry as a measurement tool. Likewise, the disease is also defined by the occurrence of a frailty fracture [6].

Among HIV patients, osteoporosis is more prevalent than in the general population, despite the age group and the use of HAART, which highlights the virus as an isolated risk factor for bone demineralization [2,3]. Furthermore, additional decreases in BMD were demonstrated after starting HAART, establishing pathways for bone demineralization different of those correlated to virus itself [7].

As number of people over 50 living with HIV has been growing markedly since HAART introduction, the deleterious effects of HIV and its treatment on bone metabolism is calling the attention of researches and health managers due to the greater risk of frailty fractures in a group of patients with many comorbidities and high rates of post-surgical complications were expected [8-11].

Driven by this warning, many studies have been conducted in recent decades to clarify the risk of frailty fractures, as well as to identify treatment strategies for bone demineralization in HIV patients [12-19].

As the use of bisphosphonates, calcium, and vitamin D is a widespread option against bone demineralization in the general population, and we performed a systematic review and meta-analysis about the effects of bisphosphonates, calcium and vitamin D in bone demineralization of patients with HIV/AIDS.

A systematic review and meta-analysis of the literature were performed according to the Cochrane Manual for Systematic Reviews of Interventions and Preferred Reporting Items for Systematic Review and Meta-Analysis recommendations [20,21].

An extensive literature search was conducted in electronic databases Medline, OVID, Cochrane Library, and EMBASE from January 2004 to January 2020, using the following descriptors: “HIV”, “AIDS”, “acquired immunodeficiency syndrome”, “bisphosphonate”, “vitamin D”, “vitamin D3”, “(25-hydroxyvitamin D)”, “1,25-hydroxyvitamin D3 (25-hydroxyvitamin D3)”, “cholecalciferol”, “calcium (calcium)”, “1,25-hydroxyvitamin D bone mineral density”, “osteopenia”, “osteoporosis”, “fracture”, “frailty, bone disease”.

Only trials published in English were selected. An inclusion criterion was the use of bisphosphonates, vitamin D, and calcium on HIV patients against bone demineralization, regardless the dosage used. The use of any other drug to enhance BMD was an exclusion criterion for the trials. The main outcome considered in meta-analysis was the lumbar spine (LS) BMD evaluated by bone densitometry.

Other variables related to BMD were also included in this meta-analysis: patients’ age, time of HIV infection, nadir of CD4+ T cells, CD4+ T cells count, body mass index (BMI), tobacco consumption, alcohol consumption, proportion of Caucasians, and LS T-score. Therefore, it is important to show that the intervention and control groups of the global meta-analysis sample were under the same comparison conditions, with no significant differences in baseline variables distribution. So, for BMD outcome and baselines variables, the meta-analysis groups were compared regarding effect size, homogeneity, and consistency. We used Cochrane’s Q and I² statistics to address heterogeneity and consistency of trials (significant if P≤0.05 in Q and I² greater than 50%). Effect sizes were reported as odds ratios for qualitative factors or as standardized mean difference for continuous variables, and 95% confidence intervals.

Two reviewers, an orthopedist and an infectologist specialized in HIV/AIDS evaluated the search results. Titles and abstracts of trials were submitted for evaluation. Manuscripts were provided in full whenever the information was considered insufficient to meet the eligibility criteria. Data were collected independently by reviewers following a prototype database in an electronic spreadsheet developed by the authors for this purpose. The agreement between the evaluators in the construction of the database was assessed by Cohen’s κ coefficient. Divergent choices regarding the selection of a study were solved by consensus. For each trial selected, the data collected follows in Tables 1 and 2.

Results were presented using an organization chart to studies selection and forest plots for the comprehension of the meta-analysis. We used Cochrane’s Q and I² statistics to address heterogeneity and consistency of trials (significant if P≤0.05 in Q and I² greater than 50%). The difference between the groups compared (intervention and control) was made at the significance level of 5% (α=0.05).

For bias evaluation, we used the Bias Risk Assessment (BRA) tool recommended by the Cochrane Manual [20], according to which the reviewers chose and classified the risk of bias to each source of bias and each trial in 4 categories: absent, low, medium, or high. Thus, to each trial and each source of bias was calculated the bias risk as the double of a weighted average of these categories frequencies (risk absent - weight 0; low risk - weight 1; medium risk - weight 2; high risk - weight 3). BRA is a measure that varies from 0 to 100%. The risk of publication bias was assessed using the graphical assessment of the funnel plot on BMD outcome.

Sensitivity analysis testing the influence of each study on the overall results was done by omitting one study at a time to explore potential sources of heterogeneity and to test the stability of regrouped results.

Descriptive analyzes were performed using the IBM SPSS Statistics for Windows (version 20.0; IBM Corp., Armonk, NY, USA), and the software used for the meta-analysis was the Comprehensive Meta-Analysis for Windows (version 3.0; Biostat Software Inc., Englewood, NJ, USA).

Full search results are represented in the flow-chart of Figure 1. Out of the 214 publications identified, 8 trials met the criteria selection and were selected to be critically evaluated in relation to authenticity, methodological quality, and the importance of the information (κ=0.92 of concordance between the 2 reviewers). All of them were controlled and compared the effects of biphosphonates (alendronate or zoledronate) combined with vitamin D and calcium on HIV patients using HAART and presenting osteopenia or osteoporosis. Vitamin D and calcium was used in all trials by the control groups as a comparison parameter. Table 1 gathers the characteristics of the trials selected, as the bisphosphonates used in each of them, its dosage, and adverse effects related to use.

The trial by Negredo et al. [17] made an experiment with 2 different treatments over 2 years. In this review, we considered each treatment as an isolated trial to perform the meta-analysis (CT6 and CT7). The trial conducted by Pepe et al. [22] had a methodological approach different from the other trials, separating the treatment and control groups into other subgroups according to the patients’ gonadal status. As none of the other trials had an approach like that used by Pepe et al. [22], this study was removed from the meta-analysis.

Random allocation of patients in groups was performed in all trials. The effects of intervention in bone metabolism were inferred by variables such as collagen n-tetra-peptide type 1, bone-specific alkaline phosphatase, osteoprotegerin, receptor activator of nuclear factor κ-β ligand (RANKL), femoral neck BMD, LS BMD, femoral head BMD, whole-body BMD. As the measurement of LS BMD (g/cm²) using bone densitometry was the unique variable assessed in all clinical trials, this was chosen as the main outcome of meta-analysis. The follow-up time (FUT) was quite discrepant among trials, ranging from 48 to 104 weeks (Table 1).

After gathering the participants of all trials, the global sample for meta-analysis had 394 patients, of whom 200 (50.8%) received bisphosphonate, calcium, and vitamin D (intervention group); the other 194 patients (49.2%) received only calcium and vitamin D (control group). For these patients, the average of FUT was 62.9 weeks. The proportion of men was greater than women, 82.5 and 17.5%, respectively; however, this disparity occurred in both study groups without statistical difference between the joint distribution of men and women in intervention and control groups. Characteristics of global sample, intervention and control groups of a meta-analysis are described in Table 2.

Table 3 summarizes the tests for heterogeneity and inconsistency performed with the model used in the analysis of SMD or, considering fixed or random effects for each variable. Meta-analysis for baselines variables showed no significant difference between groups before intervention (forest plots for baselines variables available as Supplementary Appendix 1); on the contrary, for the LS BMD after intervention (main outcome), a significant difference was found (P=0.027) and the overall diamond does not cross value 0 (Fig. 2).

Most trials of meta-analysis have demonstrated some efficiency of bisphosphonates in reducing bone resorption and BMD increasing, however, some of them could not found statistical significance between the interventions and control groups. Table 4 shows the averages and SDs for LS BMDs after intervention.

The reviewers used and judged the possible sources of bias in the selected clinical trials: no random selection, no blinding, atypical FUT, absence of media, and SD of the variables in the papers, which were estimated by other biased statistics. Regarding the occurrence of bias, Table 5 shows the BRA for each trial and a global sample of meta-analysis. The Global BRA is 11.1%, considering, in the scale from 0% to 100%, refer to low bias. The possible impact of publication bias on BMD outcomes was explored by the funnel plot (Fig. 3). The studies were not evenly distributed across both sides of the funnel plot, reveling soft asymmetry. So, the visual inspection of funnel plots suggests that studies publishing negative effects may be missing. However, the Egger and Begg tests results suggest the absence of this source of bias (P=0.23 for the Egger test and P=0.33 for the Begg test). Sensitivity analysis demonstrated that results remained statistically significant despite all deletions.

In the present meta-analysis, we found that the use of alendronate or zoledronate with calcium and vitamin D could significantly improve the BMD in patients with HIV/AIDS taking HAART (up to 0.227 g/cm²). Considering the normal values of LS BMD (1.237 g/cm²), the gain obtained with the intervention is sufficient to bring the average of LS BMD of patients with HIV/AIDS (1.176 g/cm²) to the levels of not infected population, suggesting the existence of clinical benefits [23].

The decrease in the risk of fractures due to bone frailty is certainly the main clinical benefit expected; however, establishing this correlation is a complex task that has been tried by other authors with minor success [10,24-29]. The hardest part of this correlation stems from the number of accumulated variables associated with the risk of fractures in HIV patients [26,29]. Obviously, since the reduction of BMD is an isolated risk factor for fractures in the non-infected population, we infer that the same should occur among HIV patients, although with a reduction in relative risk different from that observed in the general population [30]. Believing in this theoretical reduction in the risk of fractures when an increase in BMD is obtained, trials with HIV patients have been conducted investigating the role of vitamin D, calcium, and bisphosphonates for this purpose. As noted in previous reviews, the selected clinical trials did not allow us to establish a correlation between improved BMD and decreased risk of fracture [27,31-34]. The short FUT of patients was an important obstacle in this regard. As frailty fractures can occur at any time after the onset of osteoporosis, short-term trials do not allow us to establish this relationship, despite patients are using bisphosphonates or not.

Equally important regarding the use of bisphosphonates is the duration of its effects when its use is stopped. Studies with bisphosphonates in postmenopausal women suggest that patients with a t-score below -2.5 in the femoral neck still maintain a high risk of vertebral fractures even after 3 to 5 years of treatment, suggesting that prolonged use may be beneficial. Up to the time of writing this manuscript, the only study evaluating the prolonged use of a bisphosphonate (zoledronate) in patients with HIV/AIDS demonstrated that the effects of 2 annual doses lasted for five years after the second medication [18]. However, this study was relatively small and did not have enough strength to detect significant differences in fracture risk; the authors also recruited relatively young male patients with a t-score below -0.5 and with well-controlled disease, which makes it impossible to generalize their results to populations of patients with HIV/AIDS, low BMD and infection uncontrolled. In this meta-analysis, none of trials allowed us to assess adequately the effects of treatment interruption, nor the minimal time of use for an ideal outcome, either in terms of improving BMD or in the duration of the results obtained.

In this meta-analysis the biphosphonates used were alendronate orally, and zoledronate intravenously. Both of them were, in general, well-tolerated. No effect related to its withdrawal was reported. The most common adverse effects related to alendronate use were gastrointestinal complaints, with similar frequencies in their occurrence in trials. Possibly, due to the longer use of alendronate in protocols for the treatment of osteoporosis, care in administering the drug is already widespread (avoiding the decubitus position after taking it and ingesting it with a large amount of liquid), reducing the occurrence of adverse effects. It was related to taking the medication. Considerations regarding the prolonged use of bisphosphonates, such as osteonecrosis of the jaw and atypical femoral fractures due to suppression of bone turnover, were not reported by the authors of the analyzed trials, possibly due to the rarity of these adverse events and the short duration of the trials analyzed. There is a greater hypothetical risk of adverse gastrointestinal events with the use of alendronate in relation to the use of zoledronate, especially with regard to the direct action of alendronate on the gastroesophageal mucosa, however, this risk cannot be assessed in the meta-analysis as it has not been properly presented for studies. Zoledronate, on the other hand, is used intravenously, being less frequently related to adverse gastrointestinal events, however, the occurrence of events related to the venous infusion is expected and was presented by trials that tested this drug. Regarding the route of administration, alendronate has the advantage of being able to be taken by patients themselves at home, while zoledronate needs hospitalization and a health team for its administration. In contrast, zoledronate has the dosage convenience when administered every six months, while alendronate is used weekly, which could be advantageous in patients taking several drugs, such as those used in HAART. Respecting the annual cost of treatment, zoledronate (2 annual doses) is more expensive than alendronate, hampering to use it on a large scale both for public health programs and for studies unlinked to the pharmaceutical industry that aim to measure its effectiveness and adverse effects.

The use of calcium and vitamin D was not homogeneous between studies, with calcium doses ranging from 400 to 1,500 mg/day, and vitamin D doses ranging from 400 to 800 IU/day. Bolland et al. [18] used a monthly dose of 50,000 IU of vitamin D (1.25 mg of cholecalciferol). An adequate intake of vitamin D calcium is important for preventing bone demineralization and reducing the risk of fractures. The physiological needs of these nutrients vary throughout life, increasing with aging [32,35,36]. As vitamin D deficiency is prevalent in HIV patients, supplementation of this vitamin was used in association with calcium in both the treatment and control groups in all trials [32]. With regard to the doses of vitamin D used, previous studies carried out in patients who do not have HIV suggest a dose-dependent effect when demonstrating that supplementation with 400 IU/day has less effect on the risk of fracture than its use in doses equal to or greater than 600 IU/day [37-39]. Among HIV patients there is still no consensus in the literature regarding the dose required for supplementation and ideal serum levels of 1,25-hydroxy-vitamin D3 (25[OH]D3) for improving BMD and preventing frailty fractures [40]. Some authors consider that when the food intake of these elements is insufficient in HIV patients, Screening for hypovitaminosis should be done before starting the replacement of vitamin D and calcium [40]. In our review, only trials testing the effects of zoledronate considered serum levels of 1,25(OH)D3 as baseline characteristics among the study groups [17-19,41].

Some risk factors for bone demineralization analyzed by the trials were included in the meta-analysis, such as age, white race, smoking, drinking, HIV infection time, indicators of the immune profile (nadir and CD4+ T cells), BMI and the t-score of the LS before treatment. For these factors, there was no significant difference between the treatment and control groups in the meta-analysis sample. However, other risk factors for bone demineralization, such as physical activity, daily calcium intake, drugs used in HAART and serum calcium and vitamin D levels before interventions were not adequately presented by the authors and did not could be included in the meta-analysis, figuring as a significant limitation of this review. Other important limitations were the incomplete assessment of risk factors for bone loss among trials, the small number of trials, the short FUTs of the participants, and the methodological differences of study designs.

As some trials could not find by itself statistical differences between groups tested, despite all limitations above mentioned, the meta-analysis performed was relevant insofar as revealed no substantial heterogeneity, with high consistency and statistical significance to the main outcome considered. Consequently, as well as in the general population, the combination of Alendronate (orally) or zoledronate (intravenously), with calcium and vitamin D (orally) is effective to increase LS BMD in HIV patients. Further randomized clinical trials controlling both the variables related to bone demineralization and adverse effects can be useful to improve therapeutic protocols against bone loss in HIV.

This study provides enough evidence that BMD post-treatment is higher in HIV patients who used bisphosphonates combined with calcium and vitamin D.