Fracture Healing in Patients With HIV in South Africa: A Prospective Cohort Study

Supplemental Digital Content is Available in the Text.

immunosuppression resulting from HIV infection on the fracture-repair process following a musculoskeletal injury are not well understood.
A number of factors could theoretically affect fracture healing. HIV principally affects immunological status by exhausting the host CD4 T cells, resulting in an increase in the risk of opportunistic infections. HIV also affects other cellular chemical mediators, including interleukins 1 and 6 (IL1, IL6), and tumour necrosis factor (TNF), which have been shown to play a role in the fracture-repair process. 6,7 HIV-and ART-associated interruption in osseous blood supply can also lead to osteonecrosis, 8,9 and this microvascular effect may lead to higher rates of delayed fracture healing and non-union. 10 Reduced bone mineral density, bone mineralisation, and bone turnover have also been shown to occur in HIV-positive individuals and those on ART. 11,12 This not only increases the risk of fragility fractures but could potentially influence fracture healing. 13 A small number of clinical studies have investigated the role of HIV in the fracture-healing process. [14][15][16] These suggest that HIV and/or ART are associated with delayed fracture healing and may result in non-union, although patient numbers were small, and no underlying mechanisms were identified. If this hypothesis were shown to be true, the surgical management of fractures could be tailored to optimise bone union during the fracture-healing phase in HIVpositive patients, improving outcomes and reducing the substantial physical and social burden that occurs in these patients as a result of traumatic injuries.
Overall, the true effect of HIV and ART on bone healing is very poorly understood. This paper reports the findings of the HIV in Orthopaedic Skeletal Trauma (HOST) study, which aimed to investigate whether HIV infection is a risk factor for the development of delayed union or non-union following a fracture.

Study design and participants
The HOST study was a multi-centre prospective study of patients undergoing fracture surgery at two tertiary referral hospitals -Groote Schuur Hospital and Tygerberg Hospital -in Cape Town, South Africa. Recruitment was undertaken over a 14-month period, between September 2017 and December 2018.

Definitions and outcomes
The primary study outcome was the proportion of participants with delayed bone union at 6 months, compared between HIV-positive and HIV-negative participants. The secondary study outcomes were non-union at 9 months and infection.
Bone healing was assessed using the validated radiological union scoring system for the tibia (RUST scoring system). [18][19][20] Fracture union was defined as: radiological union on RUST score (score of three on at least three of the four cortices (anterior, lateral, medial, or posterior cortex) -a total RUST score of nine or more) within 6 months of surgery. [18][19][20] Delayed bone union was defined as impaired bone healing at 6 months on RUST score (RUST score <9). [18][19][20] Nonunion was defined as either impaired bone healing at 9 months on RUST score (RUST score <9), [18][19][20] or the need for further surgery to achieve union (RUST score <9) before 9 months (decision made by two orthopaedic surgeons).
Two reviewers (both orthopaedic surgeons), blinded to HIV status, independently assessed radiological fracture union on radiographs. In case of discrepancies in RUST scoring between reviewers, a third reviewer (orthopaedic surgeon) independently undertook a review of the radiograph to determine the final outcome.
Infection was diagnosed using the United States Centers for Disease Control and Prevention (CDC) criteria for "superficial surgical site infection (SSI)" and "deep surgical site infection (DSI)". SSI was defined as wound infection involving the skin and subcutaneous tissue that occurred within 30 days of surgery, 21 and DSI as a wound infection involving the tissues deep to the skin that occurred within 30 days of injury (closed reduction of fracture) or 90 days (open reduction of fracture). 22 Late infection was diagnosed as any late wound breakdown (>30 days for closed reduction of fractures or >90 days for openly reduced fractures) or sinus formation, or unexplained late pain with associated radiological changes consistent with peri-implant infection. 23

Statistical methods
A previously established orthopaedic surgical register suggested that 400 participants were likely to undergo IM nailing of the tibia and femur at the two centres over the 14-month study period and 80% (n = 320) were assumed to be able to complete follow-up to 9 months. On the basis of previous research, [24][25][26] it was estimated that 85% (n = 272) of the 320 participants would have fracture union at 6 months (control), and 15% (n = 48) would have delayed bone union (cases).
Assuming that 20% of participants without delayed union would be HIV positive, a sample size of 400 participants would give 82·8% power to detect at least a two-times relative difference in HIV prevalence between cases and controls at the p=0·05 threshold.

A C C E P T E D
Baseline characteristics were summarised using means (with standard deviations), medians (with interquartile ranges) and percentages, and compared between HIV-positive and HIV-negative participants. For the primary outcome (delayed union), a multivariable logistic regression model was constructed to estimate the odds ratio (OR) and 95% confidence interval (CI) for delayed union comparing HIV-positive and HIV-negative participants and adjusting for important confounders, identified a priori through construction of putative causal diagrams. A separate model was constructed for Participants with HIV only, to estimate the associations between HIV-associated predictors (e.g. CD4 cell count, viral load, ART use) and delayed union. For the secondary outcome (non-union), a multivariable logistic regression model was constructed to estimate the OR and 95% CI for non-union comparing HIV-positive and HIV-negative participants and adjusting for important confounders. Statistical analysis was done using R statistical software.
Some of the enrolled participants had more than one tibia or femur fracture. Therefore, in the analysis, confidence intervals were adjusted for clustering by including a random effects term in regression models.

Role of funding source
The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication.

Results
Between September 2017 and December 2018, 638 patients underwent 683 IM nailings of the femur and tibia at the two study sites and were screened for study eligibility; 238 participants (241 IM nailings) did not meet the study inclusion criteria and 400 participants (442 IM nailings) were enrolled in the study ( Figure 1). Baseline characteristics of all participants are presented in Table 1.
The overall prevalence of HIV in the study population was 18·8% (75/400 participants). The 75 participants with HIV underwent 87 IM nailings, giving an overall prevalence of HIV per IM nail of 19·7% (87/442). Just over half of participants with HIV (42/75, 56·0%) knew their HIV diagnosis prior to enrolment in the study. The median length of A C C E P T E D time a participant had had a diagnosis of HIV was 1397 days (inter-quartile range [IQR] 686-3565 days), or 3·8 years (IQR range 1·9-9·8 years). The remaining participants were diagnosed during their admission (25/75, 33·3%) or within 2 weeks of their discharge (8/75, 10·7%). Baseline characteristics of participants, stratified by HIV status, are shown in Table 2.
All 442 fractures underwent reamed locked (proximally and distally) IM nailings across the two study sites. The Of the 400 participants recruited to the study, 42 (10·5%, 47 IM nailings) were lost to follow-up before reaching a study outcome and were excluded from primary analysis. None of the 42 participants had developed a delayed or non-union or deep or superficial surgical site infection before being lost to follow-up. Four participants (four IM nailings) who were lost to follow-up were HIV positive.
Only 1·5% (6/395) of fractures developed an SSI (Table 4) ; owing to these low numbers, univariable and multivariable logistic regression analyses were not undertaken. The proportion of fractures that developed SSI in participants with HIV was 1·2% (1/83), compared to 1·6% (5/312) in participants without HIV. Two fractures that developed SSI subsequently went on to non-union (both HIV negative), and no cases went onto delayed union.

Discussion
To the authors' knowledge, this is the first large prospective study to assess the association between HIV infection and bone healing following a fracture. Previous small studies (fewer than seven participants) reported delayed union in people with HIV. 16,27 The present study showed that HIV-positive status was not associated with the development of delayed bone healing following an IM nailing of the tibia or femur among trauma patients in the Western Cape, South Africa. In fact, there were lower odds of fracture non-union in participants with HIV compared to HIV-negative participants. Previous studies had suggested non-union in 0-11% of fractures in HIV-positive individuals following surgical fixation of a fracture. 14 Antiretroviral therapy regimen, CD4 count, and viral load at baseline were also not associated with a significant risk of delayed union in the study population of participants with HIV. However, a greater number of ART naïve participants and more of those with a higher viral load developed delayed union and a much larger, appropriately powered, study is required to investigate this further.
HIV was not associated with the development of delayed union or non-union in open fractures. There is little evidence surrounding the risk of delayed union and non-union following an open fracture in people with HIV in the current literature but non-union rates of between 10% and 43% have been reported in a small number of studies. 14 Contrary to previous research, this study demonstrated that HIV status does not appear to affect the clinical outcome of fracture healing and may potentially lower the risk of non-union. This could be explained by a number of factors, including direct and indirect immunological effects of HIV on bone, 28,29 and/or characteristics of the study population.
However, this remains speculative as the study did not investigate mechanisms of bone healing. It is also acknowledged that this study focused solely on individuals who underwent IM nailing of a lower limb long bone fracture. Therefore, our results may not translate to fractures at different sites of the musculoskeletal system, fractures treated by other methods of fixation or managed non-operatively.

A C C E P T E D
Infection rates in the study population as a whole were similar to those published in the literature. 30 There was a low number of SSIs overall (1·5%, 6/395) in the study population, so it was difficult to make comparisons between HIVpositive and HIV-negative populations. The proportion of DSIs was higher in the participants with HIV (8·4%, 7/83) vs. 4·5%, 14/312); however, this difference was not significant on multivariable analysis (OR: 2·59, 95% CI: 0·86-7·80; p=0·090). If DSI and SSI were combined to give a rate of "early implant infection", participants with HIV had an early implant infection rate of 9·6% (8/83) compared to 6·1% (19/312) in the population of participants without HIV. Late implant infection occurred in a higher proportion of HIV-positive than HIV-negative participants (6·0%, 5/83 vs. 0·6%, 2/312). Overall, this suggests that infection was slightly more common in those who are HIV positive, but the relatively low numbers make it difficult to draw substantial conclusions When assessing the role of infection and the development of delayed union and non-union, participants with HIV who developed either an SSI, DSI, or late infection were more likely to go onto delayed union (7/13, 53·8% vs. 1/21, 4·8%) but not non-union (0/13, 0% vs. 4/21, 19·0%). However, the prevalence of infection overall was too low to make definitive conclusions and much larger studies would be needed to investigate this further.
Of the seventy-five participants with HIV enrolled in the study 56% (42/75) were taking ART on enrollment. Ideally, in order to determine the effect of HIV infection alone on fracture healing, the study should have only included those participants who were not on ART therapy. However, ethical restrictions and time limitations made this impossible.
Another limitation was the use of the RUST score to determine the primary outcome of delayed union. This scoring system has not been validated for use in the femur. However, it is the best tool available to determine bone union without the need for additional investigations.
This study demonstrates that HIV was not shown to be associated with the risk of developing delayed bone healing following an IM nailing of the tibia or femur in the study population. There was a strong trend towards lower odds of fracture non-union in participants with HIV compared to those who are HIV-negative. In conclusion, the evidence from this study suggests that fracture surgery in individuals with HIV is safe and as least as effective in HIV-positive as HIVnegative patients, with no increased risk of delayed or non-union. The results show that HIV status should not influence the decision to operate or use internal fixation in these patients.