Are serious falls associated with subsequent fragility fractures among Veterans living with HIV?

Abstract

Background:

The extensive research on falls and fragility fractures among persons living with HIV (PWH) has not explored the association between serious falls and subsequent fragility fracture. We explored this association.

Setting:

Veterans Aging Cohort Study (VACS)

Methods:

This analysis included 304,951 six-month person-intervals over a 15-year period (2001–2015) contributed by 26,373 PWH who were 50+ years of age (mean age 55 years) and taking antiretroviral therapy (ART). Serious falls (those falls significant enough to result in a visit to a healthcare provider), were identified by external cause of injury codes and a machine-learning algorithm applied to radiology reports. Fragility fractures were identified using ICD9 codes and included hip fracture, vertebral fractures, and upper arm fracture and were modelled with multivariable logistic regression with generalized estimating equations.

Results:

After adjustment, serious falls in the previous year were associated with increased risk of fragility fracture (odds ratio (OR) 2.10; 95% confidence interval (CI) 1.83, 2.41). Use of integrase inhibitors was the only ART risk factor (OR 1.17; 95% CI 1.03, 1.33). Other risk factors included diagnosis of alcohol use disorder (OR 1.49; 95% CI 1.31, 1.70) and having a prescription for an opioid in the previous six months (OR 1.40; 95% CI 1.27, 1.53).

Conclusion:

Serious falls within the past year are strongly associated with fragility fractures among PWH on ART – largely a middle-aged population – much as they are among older adults in the general population.

INTRODUCTION

As the population of persons living with HIV (PWH) ages, interest in geriatric syndromes such as falls and fragility fractures (hip, vertebral and upper arm fractures) has increased, even though most PWH are middle-aged (50–64 years of age).¹ Among older adults (aged 65+ years) in the general population, the strong, positive association between falls and fragility fractures is well-established.^2,3 While there is an extensive literature that addresses falls^4–8 and fragility fractures^9–11 in PWH, the connection between the two has yet to be explored. There are certainly reasons to think that this association is an important concern. PWH are more likely to experience decreased bone mineral density,^12–17 fragility fractures^9–11 and falls^4–8 than uninfected comparators. However, research from the general population suggests that middle-aged individuals, even those with osteoporosis, have a relatively low probability of fracture¹⁸ after a serious fall. To address this question, we explored the adjusted association between serious falls in the past year and fragility fracture among PWH who were 50+ years of age and taking antiretroviral therapy (ART). We also wanted to understand whether serious falls in the past year were more strongly associated with fragility fractures among those 65+ years of age and among those 50–64 years.

METHODS

Sample

We used data from the Veterans Aging Cohort Study (VACS), an electronic health record (EHR)-based cohort that includes all individuals diagnosed with HIV within the Veterans Health Administration (VA).¹⁹ We included all PWH on ART who were 50+ years of age and who presented for care between 01/01/2001 and 09/30/2015.

Observation time

Observations contributed by each participant consisted of six-month person-intervals beginning on January 1, 2001 for participants enrolled prior to 2000 and at the date of cohort enrollment for all others.

Outcome

The outcome of interest was a fragility fracture that occurred within a person-interval. We explored hip fractures (ICD9 820.0X, 820.1X, 820.2X, 820.3X, 820.8, 820.9), vertebral fractures (805.2, 805.3, 805.4, 805.5, 805.6, 805.7), and upper arm fractures (812.0X, 812.1X, 812.2X, 812.3X, 812.4X, 812.5X). Our sample was predominantly male (98%) with a mean age of 55 years. Because wrist fractures among middle-aged men are thought to be related to severe trauma rather than osteoporosis,²⁰ we excluded wrist fractures from this analysis.

Primary predictor and covariates

The primary exposure was occurrence of a serious fall in the year preceding the start of the index person-interval. Serious falls were identified by external cause of injury codes (Ecodes) and by a machine learning algorithm that identified falls in radiology reports.²¹

We controlled for risk factors associated with fragility fractures identified among older adults including demographics (age, sex, and race/ethnicity), body mass index (BMI),²² and pain. We also included a count of chronic medical conditions (identified by ICD9 codes): cognitive impairment, dementia,^23,24 stroke,²⁵ seizure disorder,²⁶ vision impairment (blindness and cataracts),²⁷ diabetes,²⁸ anemia,²⁹ HCV, cirrhosis,³⁰ renal disease,³¹ hypertension,³² abnormal gait and osteoarthritis.³³ We included a flag for participants who had at least one mental health diagnosis³⁴ (anxiety, major depressive disorder, bipolar disorder, schizophrenia, and psychosis). We used the VACS Index to adjust for disease severity.^35–37 The VACS Index includes data on age; HIV biomarkers (HIV-1 RNA [viral load]; CD4 cell count); and non-HIV biomarkers (hemoglobin, hepatitis C; FIB-4 to assess liver function; and estimated glomerular filtration rate to assess renal function). Potential scores range from 0 to 164, with higher scores being associated with a greater risk of mortality. We also assessed VACS Index score components to evaluate the contribution of CD4 count and HIV-1RNA to fracture risk. We adjusted for a count of chronic medications: antihypertensives,³⁸ hypoglycemics,²⁸ antithrombotic agents, proton pump inhibitors,³⁹ antiarrhythmics, nitrates⁴⁰ statins,⁴¹ steroids/glucocorticoids,²⁴ and drugs that are active in the central nervous system (CNS)⁴² including anticonvulsants, opioids, muscle relaxants, benzodiazepines, serotonin and norepinephrine reuptake inhibitors, selective serotonin reuptake inhibitors, tricyclic antidepressants, atypical antidepressants (e.g. bupropion, maprotiline, mirtazapine, nefazodone, trazodone), antipsychotics, atypical sleep medications (e.g. amitriptyline, doxepin, mirtazapine, trazodone), and antihistamines. We created separate flags for the use of mental health medications, opioids, muscle relaxants, anticonvulsants, and benzodiazepines. We also evaluated antiretroviral therapy (ART) both by class (PI, NRTI, NNRTI, INSTI) and as individual medications.

Demographic variables were assessed at baseline. Pain and medication use were evaluated in the 6-month interval prior to the interval of interest. Occurrence of a serious fall was assessed in the year preceding the start of the index person-interval. All other covariates were assessed at the start of each 6-month interval.

Ethics

VACS was approved by the Institutional Review Boards of VHA Connecticut Healthcare System and Yale University School of Medicine. It has been granted a waiver of informed consent and is HIPAA compliant ¹⁹.

Statistical methods

The analytic unit was a six-month person-interval.⁴³ We calculated descriptive statistics for variables at baseline (the first six-month interval) and compared them by fall status: those who had at least one serious fall during follow-up vs those who did not. For continuous variables, we used either means and standard deviations or medians and interquartile ranges. For categorical variables, we used percentages. Only BMI, pain, and the VACS index score had missing values (8.3%, 7.9%, and 9.0%, respectively). As implemented in the SAS Proc MI,⁴⁴ we used fully conditional specification for imputation of variables with missing values.⁴⁵ We multiply imputed the source data set of more than 2.7 million person-intervals a total of five times, drawing from the 22 candidate variables eligible for model selection.

The final analytic sample consisted of 304,951 person-intervals contributed by 26,373 PWH who were 50+ years of age and using ART. Using this analytic sample, we developed a multivariable model of fragility fractures. We forced five variables into the multivariable model: an indicator of any serious fall in the previous year, indicators of the two non-reference levels of BMI (25–29.9 kg/m² and ≥ 30 kg/m²), female sex, non-white race, and the VACS Index.³⁵ We then used backwards selection based on minimization of the Bayesian information criterion⁴⁶ to select a parsimonious model. We subsequently added indicators of each of the four classes of ART medications (protease inhibitors [PI], nucleoside reverse transcriptase inhibitors [NRTI], non-nucleoside reverse transcriptase inhibitors [NNRTI], and integrase inhibitors ⁴⁷). We also examined an alternative model that replaced the ART classes with indicators for 29 individual ART medications. As none of these individual medications were associated, we used the model that included ART classes. We then explored this model in those aged 50–64 years and in those 65+ years. We identified two variables that appeared to perform differently across the strata, (i.e., use of INSTI and use of mental health medications) and ran a model with the interactions of those variables with age to assess statistical significance.

The associations reported for two of the continuous variables were scaled such that incremental change represented an increase in the median number of units. For example, the median VACS Index score was 29, so the reported OR corresponds to an incremental change of 29 units. Similarly, the median count of chronic medications was 2, so the reported OR corresponds to an incremental change of 2. With median values close to one, the scaling of the other two continuous variables (count of physical comorbidities and pain score) was not modified.

The variables retained in the parsimonious model were used to fit a multivariable logistic regression model to each of the five imputations. We used generalized estimating equations with an autoregressive correlation structure to adjust for the clustering of repeated intervals within patients. The coefficients from the separate imputations were combined using Rubin’s rules to yield final model coefficients.⁴⁴ Statistical significance was defined as a two-sided p-value < 0.05 with all analyses performed in SAS Version 9.4. with SAS/STAT 14.3.⁴⁷

RESULTS

Our analysis included 26,373 PWH who were 50+ years of age and were taking ART. Of these, 22% experienced a serious fall at least once during follow-up. Of those who fell, 81% fell two or more times. The mean duration of follow-up was 3.5±3.0 years and 4.0±3.5 years respectively among those who did and did not fall. The annual rate of fracture was 5.76% among those who experienced at least one fall in the past year, and 0.86% among those who did not. Most of the sample (90%) started follow-up before 65 years of age. Mean (±SD) age at time of fall was 58±7 years. Non-fallers were more likely than fallers to be of non-white race (61% vs 58% respectively; p<0.001), have a BMI of 25–29.9 (36% vs 34% respectively) or ≥ 30 kg/m² (18% vs 17% respectively, p=0.003), and to fill prescriptions for NNRTIs (48% vs 47% respectively; p=0.03) or INSTIs (8% vs 4% respectively; p<0.001). Neither age at baseline (mean 55±6 years) nor NRTI use (98%) differed by fall status. All other covariates of interest were more prevalent among those who fell (Table).

Table.

Baseline data from VACS participants living with HIV who were 50+ years of age and on ART between 01/2001 and 09/2015 (N=26,373)

Variables	Never fell N=20,700	Ever fell N=5,673	P-value
Number of person-intervals	216,471	88,480	<0.001
Annual rate of fragility fractures	0.86%	5.76%	<0.001
Age in years at baseline (mean±SD)	55±6	55±6	0.82
Non-White race	61%	58%	<0.001
Female sex	2%	3%	0.001
Body mass index   <25 kg/m2   25–29.9 kg/m2   ≥30 kg/m2	46% 36% 18%	49% 34% 17%	0.003
Pain (mean±SD)	1.8±2.9	2.2±3.2	<0.001
Alcohol use/abuse	19%	25%	<0.001
Illicit substance use/abuse	22%	28%	<0.001
Count of physical comorbidities [median (IQR)]	1 (0, 2)	1 (1, 3)	<0.001
Diagnosis of one or more mental health comorbidities	37%	44%	<0.001
Vision impairment	4%	5%	0.03
Count of chronic medications [median (IQR)]	2 (1, 4)	3 (1, 5)	<0.001
Prescription opioids*	22%	32%	<0.001
Prescription for at least one mental health medication*	34%	46%	<0.001
Proton pump inhibitor* use	16%	20%	<0.001
Glucocorticoid* use	14%	18%	<0.001
Anticonvulsant* use	13%	19%	<0.001
Benzodiazepine* use	10%	16%	<0.001
Muscle relaxant* use	7%	10%	<0.001
VACS Index (mean±SD)	30±20	35±20	<0.001
CD4 count (cells/mm3)	449±273	427±262	<0.001
HIV-1RNA (percent undetectable [<400 copies/mm3])	61%	55%	<0.001
Antiretroviral therapy drug classes
NRTI	98%	98%	0.12
NNRTI	48%	47%	0.03
PI	51%	56%	<0.001
INSTI	8%	4%	<0.001

^*Medications identified in the six-month interval prior to the interval of interest to avoid identifying medications prescribed as a result of the fracture. Medications identified using fill/refill data.

Multivariable analyses

The explanatory variables in the final model included a serious fall within the past year; alcohol use disorder; at least one mental health diagnosis; vision impairment; illicit substance use; count of physical comorbidities; prescriptions for at least one mental health medication, opioid, anticonvulsant, proton pump inhibitor, or steroid; VACS Index score; count of chronic medications; pain; body mass index with non-reference levels of 25 −29.9 kg/m² and ≥ 30 kg/m²; non-White race; and ART medication classes. Because no statistically significant interactions were identified in models stratified by age, we report results from the full cohort. Serious falls within the past year were strongly associated with an increased risk of fragility fracture (odds ratio (OR) 2.10; 95% confidence interval (CI) 1.82, 2.41). Other risk factors demonstrated expected associations with fragility fractures (Figure). Those with the strongest associations with fragility fracture (ordered by Z score) included the count of physical comorbidities (OR 1.13; 95% CI 1.10, 1.16), opioid prescription (OR 1.40; 95% CI 1.27, 1.53), alcohol use disorder (OR 1.49; 95% CI 1.31, 1.70), VACS Index score (units of 29) (OR 1.15; 95% CI 1.09, 1.22) and count of chronic medications (units of 2) (OR 1.07; 95% CI 1.04, 1.10). Elevated BMI (25–29.9 kg/m² and ≥ 30 kg/m²) (OR 0.73; 95% CI 0.66, 0.81; and OR 0.59; 95% CI 0.52, 0.68, respectively) and non-white race (OR 0.67; 95% CI 0.60, 0.74) were associated with lower odds of fragility fracture. When we included the VACS Index components rather than the composite score, HIV-1RNA (OR 1.00; 95% CI 1.00, 1.00; p=0.037) was modestly associated with increased risk of fragility fracture, but CD4 count (OR 1.00; 95% CI 1.00, 1.00; p=0.146) was not. Of the ART classes, only use of INSTIs was associated with increased risk of fracture (OR 1.18; 95% CI 1.04, 1.34).

Figure. Associations between predictors and fragility fractures.

Legend: These are multivariable results of the model using serious falls as the primary predictor and fragility fractures as the outcome. Other variables included in the model are those retained after using backwards selection based on minimization of the Bayesian information criterion to select a parsimonious model.

DISCUSSION

Our results suggest that much like older adults in the general population, serious falls within the past year among PWH on ART who are 50+ years of age are independently associated with fragility fracture. Notably, the association between a prior serious fall and fragility fracture was not stronger in PWH who were 65+ years than among those 50–64 years. A prescription for opioids and a diagnosis of alcohol use disorder were two of the strongest predictors of fragility fracture in our study. While these findings are consistent with the literature on older adults in the general population, ^48–52 rates of alcohol and prescription opioid use were higher among PWH.^48,52 High usage rates of these substances may contribute to the early onset of fragility fracture among older PWH.

This study has many strengths. VACS is a large, EHR-based cohort with substantial, longitudinal follow-up. This cohort provided access to a greater range of variables than is available in many cohorts, including pharmacy fill-refill and free text data (clinical notes) that facilitated expanded and improved variable capture. We were able to follow many PWH over a long period of time, which enabled us to reliably assess our low incidence outcome of fragility fracture.

Our study also has limitations. While our approach does not identify all falls that occurred in this population, it likely identifies those falls that are most concerning to patients and providers. Although the percentages of women and individuals taking INSTIs were quite low, the numbers of women (388 who never fell, 148 who fell) and those taking INSTIs (1634 who never fell and 233 who fell) were nonetheless meaningful.

Our study demonstrates a consistently strong association between serious falls in the past year and subsequent fragility fractures among both middle-aged (50–64 years of age) and older (65+ years of age) PWH. Serious falls and subsequent fragility fractures in the general population place an individual at increased risk for hospitalization, nursing home placement, and impaired function. These outcomes in middle age presage poor quality of life and increased hospitalization for PWH as they age and highlight the need for effective interventions to prevent serious falls and fragility fractures in this population.