Effects of aerobic and resistance exercise alone or combined on strength and hormone outcomes for people living with HIV. A meta-analysis

Background Infection with human immunodeficiency virus (HIV) affects muscle mass, altering independent activities of people living with HIV (PLWH). Resistance training alone (RT) or combined with aerobic exercise (AE) is linked to improved muscle mass and strength maintenance in PLWH. These exercise benefits have been the focus of different meta-analyses, although only a limited number of studies have been identified up to the year 2013/4. An up-to-date systematic review and meta-analysis concerning the effect of RT alone or combined with AE on strength parameters and hormones is of high value, since more and recent studies dealing with these types of exercise in PLWH have been published. Methods Randomized controlled trials evaluating the effects of RT alone, AE alone or the combination of both (AERT) on PLWH was performed through five web-databases up to December 2017. Risk of bias and study quality was attained using the PEDro scale. Weighted mean difference (WMD) from baseline to post-intervention changes was calculated. The I2 statistics for heterogeneity was calculated. Results Thirteen studies reported strength outcomes. Eight studies presented a low risk of bias. The overall change in upper body strength was 19.3 Kg (95% CI: 9.8–28.8, p< 0.001) after AERT and 17.5 Kg (95% CI: 16–19.1, p< 0.001) for RT. Lower body change was 29.4 Kg (95% CI: 18.1–40.8, p< 0.001) after RT and 10.2 Kg (95% CI: 6.7–13.8, p< 0.001) for AERT. Changes were higher after controlling for the risk of bias in upper and lower body strength and for supervised exercise in lower body strength. A significant change towards lower levels of IL-6 was found (-2.4 ng/dl (95% CI: -2.6, -2.1, p< 0.001). Conclusion Both resistance training alone and combined with aerobic exercise showed a positive change when studies with low risk of bias and professional supervision were analyzed, improving upper and, more critically, lower body muscle strength. Also, this study found that exercise had a lowering effect on IL-6 levels in PLWH.


Methods
Randomized controlled trials evaluating the effects of RT alone, AE alone or the combination of both (AERT) on PLWH was performed through five web-databases up to December 2017. Risk of bias and study quality was attained using the PEDro scale. Weighted mean difference (WMD) from baseline to post-intervention changes was calculated. The I 2 statistics for heterogeneity was calculated.

Results
Thirteen studies reported strength outcomes. Eight studies presented a low risk of bias. The overall change in upper body strength was 19.3 Kg (95% CI: 9.8-28.8, p< 0.001) after AERT and 17.5 Kg (95% CI: [16][17][18][19] meta-analyses have addressed this matter to date. Finally, subgroup analysis for each exercise intervention, upper and lower body muscles, professional supervision of exercise, performing more than 150 minutes of exercise per week, and controlling for active control groups and high-quality studies (PEDro score ! 5) were conducted in compliance with the Cochrane meta-analysis standards.

Methods
This systematic review and meta-analysis was registered in the PROSPERO international prospective register of systematic reviews (CRD42018087004) and performed following the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [25].

Eligibility criteria
Randomized controlled trials (RCTs) comparing resistance training alone, aerobic exercise alone or aerobic exercise in combination with resistance training against a non-exercising control group (CG) were considered for inclusion. The studies had to include participants with HIV at any stage of the infection process, older than 18 years, with or without co-morbidities, and investigate strength outcomes (lifted external resistance in kg or lbs) as the primary outcome and hormones (i.e. testosterone) related to the muscular system as secondary outcomes, in response to exercise. Aerobic exercise was defined according to the American College of Sports Medicine (ACSM) as "any activity that: uses large muscle groups, can be maintained continuously, and is rhythmic in nature" [26], and resistance training was outlined as "a form of physical activity that is designed to improve muscular fitness by exercising a muscle or a muscle group against external resistance" [27]. Both exercises had to be performed more than two times per week as described by Gomes Neto et al. [18] rather than three times per week as in O´Brien et al. [21], and for at least four weeks. Studies administering steroid supplementation to the IGs and/or CGs were excluded due to the possibility of an additional effect on muscle strength. Other forms of exercise (e.g. tai chi, qi gong) were not considered because tai chi interventions varied in the tai chi practiced forms [28] and homogeneity among the types of exercise performed by the IGs needed to be achieved. Studies investigating two exercising groups without any non-exercising control groups were considered to be excluded because an exercising control group could lead to a significant improvement in muscle strength, resulting in a decreased ability of the intervention group (IG) to demonstrate minimal changes. Not necessarily physical activities (placebo-treated, social contact exercise recommendations, counseling, recreational activities) and very light physical activity groups, were considered to be active CGs. Groups following their usual activity and explicitly not exercising were considered to be passive CGs.

Literature search strategy for study identification
A literature search was performed using five databases (clinicaltrials.gov, PEDro physiotherapy evidence database, PubMed, the Cochrane Central Register of Controlled Trials (CENTRAL) and Web of Science), restricted to English-language studies published up to the end of December 2017. Two reviewers (CP and PZ) individually screened and recorded the relevant citations following the above eligibility criteria and recorded them in a standard data format. After selecting the relevant citations by title, the abstracts were screened. After fulfilling both previous steps, full texts were obtained and evaluated. In case of disagreement, both authors discussed their differences until reaching an agreement. If this was not possible, a third author (AH) was consulted to determine the final decision.
Search parameters and syntax were adapted to each database's requirements. Text words and Combined Medical Subject Headings (MeSH) terms were related to exercise and physiological parameters. The search strategy is presented in Table 1.

Data collection
Data was extracted by both reviewers (CP and PZ) independently, using a standard digital sheet form.
Measuring units were independently converted by the two reviewers (CP and PZ), pounds (lbs) to kilograms (kg) and mmol/l to pg/dl. Outcomes reported as Mean ± standard error or Mean change (post minus pre) ± standard error were converted into mean ± standard deviation.
In case of missing relevant data in the selected studies, the original authors were contacted via email asking for the required missing information. If two weeks passed without an answer from the author, the author was kindly reminded and the co-authors were contacted via email. If the author did not answer our emails, then the study was left out of the quantitative synthesis.

Risk of bias and quality of included studies
CP and PZ individually assessed the risk of bias and the quality of the included studies using the PEDro scale [25,29]. Every PEDro criterion had to be clearly met and described in the selected study. The PEDro scale consists of eleven criteria in which the first "criterion of eligibility" is marked with a "yes" or "no." If the study had no eligibility criteria, the study was excluded. The rest of the criteria were marked with a checkmark or a "0." The possible PEDro score range is from 0 to 10. Discrepancies on the studies' PEDro score between the two reviewers were resolved by a third author (AH). The results of the quality and risk of bias assessment of the included articles are shown in Table 2 in the Results section.
Studies with a PEDro score ! 5 where categorized as high-quality studies, because blinding might be difficult to achieve and maintain for various reasons and is less frequently reported in non-pharmacological treatment RCTs [30,31]. Moreover, Moseley et al. [32] investigated the number of RCTs available in the PEDro database that satisfied the blinding criteria (subject blinding, therapist blinding or assessor blinding) and found a low prevalence of blinding, with 5% using blinded therapists, 9% blinded subjects and only 34% blinded assessors. For these Cochrane library ("HIV" OR "human immunodeficiency virus") AND ("exercise" OR "physical activity" OR "aerobic exercise" OR "resistance training" OR "exercise therapy") AND ("hormone" OR "testosterone" OR "cardiovascular" OR "strength" OR "fitness" OR "physiological" AND "randomized controlled trial") reasons, the total PEDro score for RCTs involving exercise interventions can be affected and thus a PEDro score lower than six can be attained even if the other criteria are satisfied.

Statistical analysis
The random effect model [33,34] was used to calculate the weighted mean difference (WMD) between intervention and control group changes from baseline and post-intervention. When the change was not available, the change (mean pre-intervention minus mean post-intervention) and the standard deviation according to the Cochran handbook for systematic reviews of interventions [35] were calculated. Parameters were analyzed for upper and lower body strength as well as hormones. Subgroup analyses were performed for resistance training (RT) alone, aerobic exercise combined with resistance training (AERT), upper and lower body muscles, professional supervision of exercise, performing more than 150 minutes of exercise per week, excluding active CGs, pre ART-era, and high-quality studies (PEDro score ! 5) [23]. A p value equal to or less than 0.05 was considered significant. Interpretation of the effect size was done with the most commonly used cut-off defined by Cohen [36], d = .20 small, .50 medium and .80 large.
To test for heterogeneity, the I 2 statistics and 95% CI [37] was calculated. Values between 25-50% reflect low, 50%-75% moderate, and values greater than 75% reflect high heterogeneity [37]. To explore the heterogeneity, this meta-analysis looked for differences across subgroups by calculating the Chi square (Χ 2 ). Publication bias was assessed by the egger´s test [38]. All analyses were performed using Review Manager Version 5.3 [39].

Search description of selected studies
A total of 398 citations through the search criteria from the databases described in the Methods section where retrieved. After screening the titles, 231 citations were excluded due to ineligible focus. Of the remaining 167 studies, 50 were excluded for the following reasons: exercise was EC, eligibility criteria; I: allocated randomization of subjects to groups; II: concealed allocation; III: similarities of groups at baseline; IV: blinding of subjects; V: blinding of researchers/evaluators; VI: blinding of assessors; VII: measure of at least one key outcome obtained from more than 85% of subjects initially allocated to groups; VIII: intention to treat; IX: comparison results between groups; X: measured at least one key outcome at two time points; ✓, criterion is present otherwise; 0, criterion is missing. performed by the control group (n = 9), the language was other than English (n = 4), the intervention or control groups were partially or totally integrated with HIV seronegative participants (n = 6), the searched outcomes were not addressed (n = 11), or the citations referred to non-RCTs (n = 13) or reviews (n = 7). The remaining 117 citations were screened before full texts were acquired; 66 citations were duplicates. 51 studies' full texts were read and 17 studies had to be excluded: in one study, the intervention groups performed only one bout of exercise [40], one study's control group performed exercise [41], two studies did not investigate exercise [42,43], two studies' intervention or control groups were partially or totally integrated with HIV seronegative participants [44,45], five studies did not investigate the desired outcomes [46][47][48][49][50], five were not RCTs [51][52][53][54][55] and one study administered nandrolone [56]. In total, 34 studies from the systematic search and four studies added by cross-referencing citations met the eligibility criteria and were considered relevant for inclusion in the meta-analysis. In total, 13 studies reported strength outcomes and were included in the quantitative analysis. Of these, four studies [57][58][59][60] reported strength and hormone outcomes (See Fig 1).

Characteristics of studies excluded from quantitative synthesis
Following our search strategy, five studies [61][62][63][64][65] were excluded due to incomplete data or because the requested data was not available. See excluded studies in S1

Description of the studies included in the meta-analysis
Of the 13 included studies, eight studies [57,60,[66][67][68][69][70][71] were classified as high-quality studies (see Table 2), four studies investigated RT [57,60,66,72] and eight studies AERT [58,59,[67][68][69][70][71]73]. The studies of Lox et al. [74] and [72] shared the same two intervention groups and results: one group performed AE alone and the other RT alone, with the exact same number of control group participants in both studies. Thus, for the purpose of analysis we decided to treat Lox et al. [74] as an AE-only intervention and Lox et al. [72] as an RT-only intervention. Passive CGs were identified in six studies [58,60,67,68,71,73] and one with no lifestyle modification [71]. Seven studies had active CGs, three studies CGs performed very low-intensity physical activity like walking or stretching [70,72,74], one study [66] applied protein supplementation 1 gÁkg -1 Áday -1 to the CG, and three studies [57,59,69] administered testosterone placebo injections to the CGs. Characteristics of all the included studies can be seen in S1 Table.

Characteristics of studies included in the meta-analysis
The number of participants included in the 13 studies were for the intervention group n = 249 at baseline and n = 246 post-intervention, whereas for the control group n = 1216 at baseline and n = 210 at the end of the study. Eight studies had a mean dropout rate of 10.7 ± 12.1%. Two studies reported no dropouts [60,68] and three studies [72][73][74] only reported participants who completed the study. The mean age for the control group was 42 ± 5.7 years and for the intervention group 42.9 ± 5.3 years. Five studies [60,66,67,69,71] included 42 women in the control group and 41 women in the intervention group. Seven studies recruited only male subjects [57][58][59]68,70,72,74]. Only the study by Mendes et al. [73] did not report the age and gender of the participants. The average baseline CD4 cell count was reported in 12 studies [57][58][59][60][66][67][68][69][70][71][72]74]. For the CGs was 432.2 ± 147.9 cells μl -1 and for the IGs 431.5 ± 167 cells μl -1 .
The 1-RM test used to measure muscle strength was reported in six studies [57,60,66,67,69,73] and the peak isometric force test was used in three studies [59,72,74]. Characteristics of all the included studies can be seen in S1 Table.

Risk of bias
According to the risk of bias analysis, eight studies scored ! 5 in the PEDro scale, specifying a low risk of bias. Five studies presented a high risk of bias with a PEDro score < 5. See Table 2.

Changes on body strength in PLWH
All types of repetition maximum (1-RM, 3-RM, 6-RM and 12-RM) were included. No details on the type of machine used to perform the RM test was mention in the included studies. See S1 Table. Two overall meta-analyses (see Figs 2 and 3) and 34 subgroup analyses were performed. The overall change after intervention on upper body strength in PLWH from baseline was 18 kg (95% CI: 11.2-24.8, p< 0.001) favoring the IG. Lower body strength also increased by 16.8 kg (95% CI: 13-20.6, p< 0.001) favoring the IG. Sub-analysis revealed a significant increase on lifted weight for each muscle group, favoring the IG. After long-term exercise, IG upper body strength showed a significant change 13.7 kg (95% CI: 6-21.5, p < 0.001). This was also true for IG lower-body strength with a mean change of 16 kg (95% CI: 11.6-20.4, p < 0.001), but significant changes were only for leg flexion and extension long-term exercise muscle groups (See Table 3). Professional supervision was reported for upper body strength in seven studies. Only the chest press muscle group was reported in these studies, with a significant change in the IG of 17.9 kg (95% CI: 10.5-25.3, p < 0.001).
Eight studies reported a change in lower body strength under professional supervision in the IG of 19.5 kg (95% CI: 13.4-25.7, p< 0.001), which was higher than the main meta-analysis. See Table 3.
After excluding studies pre-ART era, performed or published before or in the year 1996 [ < 0.001), although the results from the main meta-analyses were higher for lower body strength. Upper and lower body strength heterogeneity was high (> 75%) for the two main meta-analyses and all sub-analyses. See Table 3.

Hormone changes in PLWH
After screening the included studies, this meta-analysis found that three studies reported inflammatory markers (IL-6,IL-1β, CRP) and given their relation to the muscular system they were included in the following section.
Out of the three studies that reported hormone outcomes in PLWH, two studies [58,59] had a high risk for bias with a PEDro score 5. Two studies reported testosterone [57,59], three studies free testosterone [57][58][59] and two studies interleukin-6 and interleukin-1β [58,60]. All studies reported that samples were taken in a fasting state at the same time of day and in a rested state. For characteristics of the studies, please refer to S1 Table. Four overall analyses and four sub-meta-analyses were performed. There were no significant differences between IG and CG from baseline to post-intervention in PLWH for total testosterone, and the range of change was widely spread across the main meta-analyses. Likewise, free testosterone and IL-1β also presented a non-significant change between IGs and CGs after intervention (see Table 4).
A significant change towards lower levels of IL-6 in the IG was found, with a WMD reduction of -2.4 ng/dl (95% CI: -2.6 to -2.1, p< 0.001) (see Table 4).
Only the study by Dudgeon et al. [58] reported changes in cortisol and insulin-like growth factor 1 (IGF-1) for PLWH. After exercise intervention, lower levels of cortisol upon waking (p<  [69], where AERT exercise significantly decreased the levels of this inflammatory marker in the IG compared to the CG (−1.6 ± 0.7 vs. 0.1 ± 0.4 mg/L, p = 0.05).

Discussion
The main results of this meta-analysis suggest that aerobic exercise combined with resistance training or resistance training alone has a positive effect on muscle strength and levels of IL-6 in PLWH.
New in this meta-analysis is the inclusion of 13 studies reporting changes in strength, testosterone and inflammatory markers in PLWH. Moreover, separate analyses of the main outcomes for RT alone and AERT, studies with a low risk of bias, professional supervision and the exclusion of active control groups were performed.

Muscle strength
As previously stressed by Gomes-Neto et al. [18], aerobic exercise and resistance training programs have limited benefit on muscle strength in healthy people. Nonetheless, two previous meta-analyses have reported positive changes after AERT intervention in muscle strength in PLWH [18,21]. However, these changes were restricted only to chest press, biceps curl, and leg extensors, with no changes to the leg flexors [21]. Our results indicated that both RT alone and AERT exercise has a significant positive effect on increasing upper and lower body muscle strength in PLWH. This study highlighted that the results were significantly higher after excluding studies with a high risk of bias over upper and lower strength.
Of importance here is that after different sub-meta-analyses were performed, lower strength increased more after RT alone or if the exercise was supervised by an exercise professional (exercise scientist, physiotherapist or certified trainer [23]). Upper strength increased more after AERT intervention, excluding active CGs and in studies carried out during the ART era.
Of special importance is the significant positive change in the leg press, leg flexors and leg extensors (see Table 3). The data contrasts with that of O´Brien et al. [21] who reported only a trend toward increasing leg flexor muscle strength and a non-significant improvement on other lower extremity muscle groups in LPWH. This difference in results might be due to the increased number of included studies in the meta-analyses for lower body strength in comparison to O´Brien et al. [21]. In detail, for leg press six further studies [57,60,68,[72][73][74], for leg flexion six further studies [57,58,60,66,71,73] and for leg extension five further studies [58,66,70,71,73] were included. It is important to remark that in the study of Dudgeon et al. [58], the intervention group exercised less than three times per week, an exclusion criterion in the study by O'Brien et al. [21]. See S1 Table. This meta-analysis tried to have a more homogeneous IG by grouping the IGs according to the amount of time (more or less than 150 minutes per week) dedicated to exercise. Due to complex differences in the types of training (progressive resistance training, continuous resistance training, AERT), the use of free weights, color-coded therapeutic bands or multi-training exercise machines, alongside great differences in exercise intensity (refer to S1 Table, Information on excluded and included studies), no differences in strength compared to the main results were found.
The clinical relevance of muscle strength increases in PLWH has not yet been established. Gomes-Neto et al. [18] discussed how an increase of 40% in muscle strength likely represents a clinically meaningful strength gain. On the other hand, for O´Brien et al. [21] a clinically important change was an improvement of 2 kg for upper body and 5 kg for lower body strength. Nevertheless, no statement can be made, about the clinically relevant change due to the different possible biases in our study (i.e. types of training, exercise intensity, strength testing) and because our focus was not the standard error of measurement (SEM) and the small real difference (SRD) values of the upper and lower body strength tests.  The high degree of heterogeneity (I 2 > 75%) found among the meta-analyses might be the result of differences in the associated comorbidities (i.e. insulin resistance, HCV and altered testosterone levels), gender (only women were evaluated), and strength testing procedures (1-RM test, 3RM, 6-RM, 12-RM or peak isometric force). Furthermore, the number of participants in the included studies as well as the number of studies included in each of the meta-analyses can contribute to this high heterogeneity. Thus, caution is advised when interpreting the results.

Hormones
Testosterone levels are associated with muscle strength and physical performance in healthy men [76]. This meta-analysis found no significant changes in testosterone and free testosterone levels after the exercise interventions (see Table 4), perhaps due to differences in the testing methods, the number of included studies and, more importantly, because this secondary outcome was extracted from studies mainly reporting changes in muscle strength.
High levels of IL-6 are related to T-cell failure [77], a greater risk of losing 40% of muscle strength [76] in the aging population, and a 1.3 times higher chance of presenting with a CVD [78]. Thus, strategies directed towards controlling levels of IL-6 may be helpful in PLWH. This meta-analysis, found a significant change of lower levels of IL-6 after exercise intervention in the studies of Dudgeon et al. [58], where only males participated in an AERT intervention, and Zanetti et al. [60], where a mixed-gender intervention group performed only RT (see Table 4), supporting exercise as an alternative to reduce levels of IL-6 in PLWH.
No changes in IL-1β levels were found. This data is in agreement with Peake et al. [79], who reported that even though AE and RT increase muscle and leucocyte cells' gene expression of IL-1β, the release of this interleukin is highly regulated and thus changes in plasma may not be observed.

Limitations
The high degree of heterogeneity found in the meta-analyses was considered to be the result of a combination of various differences among the studies included in the quantitative analysis. Associated comorbidities in PLWH included coinfection of the hepatitis C virus [70,71], unhealthy habits like smoking, drinking or the use of hallucinogenic drugs [67,69,70] and other diagnosed chronic pathologies [57,67,69,71]. Gender, type of training, strength performance testing, active or passive control groups and the number of participants in each analysis was also uneven among the studies. After exploring some potential sources of heterogeneity, such as the exercise characteristics (exercise intensity, volume and weekly frequency). We were able to reduce the heterogeneity from high to low or from high to moderate in some cases. However, a high heterogeneity was still observed for most analyses. This heterogeneity can be attributed to the lack of power of the I 2 statistic when looking to small number of studies. [80][81][82]. For these reasons, the results of this meta-analysis warrant attention when interpreted.
Regarding the clinical relevance of muscle strength in PLWH, the question remains, whether a value lower or higher than 40% could reflect a better cut-off point.

Implications for research
After the above studies were evaluated in this meta-analysis, some concerns, pertinent for further RCTs involving exercise interventions in PLWH were raised. From a methodological standpoint, when possible, equally balanced groups in terms of gender, use of a unique type of training and blinding of evaluators could all increase the quality of future RCTs and decrease the risk of bias. Additionally, more studies investigating changes in testosterone and inflammatory markers (IL-6 and IL-1β) to establish the effect of RT alone or AERT on hormones in PLWH are needed.

Conclusion
Resistance training alone or combined with aerobic exercise showed a positive change after studies with a low risk of bias and professional supervision were analyzed, improving upper and, more critically, lower body muscle strength. This meta-analysis also reported that exercise had a lowering effect on IL-6 levels in PLWH.

Key considerations
• Both RT alone and AERT have a significant positive effect on increasing upper and lower body muscle strength in PLWH.
• Lower body strength increases more if exercise is supervised by a professional.
• Exercise is an alternative to reduce levels of IL-6 in PLWH.
• Taken together, exercise plays a fundamental role in HIV treatment to improve upper and lower body muscle strength in PLWH.
Supporting information S1