CROI 2015: FRAX Fracture Prediction Tool Underestimates Fracture Risk in Men with HIV


The Fracture Risk Assessment Tool (FRAX), an online tool developed by the World Health Organization and used to help guide decisions about who to screen or treat in order to prevent bone fractures, underestimates overall risk of fracture in people living with HIV -- even with an adjustment experts have recommended to improve its accuracy for people with HIV -- according to an analysis of the Veterans Aging Cohort Study Virtual Cohort (VACS-VC) reported at the recent 2015 Conference on Retroviruses and Opportunistic Infections (CROI) in Seattle.

Worse, using the thresholds currently recommended by U.S. guidelines for screening and treatment did not discriminate well between those individuals actually at risk and those not at risk for future fractures, according to Michael Yin from Columbia University, who presented the study findings.

Fractures in People with HIV

The incidence of bone fractures -- whether associated with trauma or fragility -- is increased in people living with HIV, especially with advancing age. According to a recent meta-analysis of the VACS, the unadjusted risk of fragility fractures was about 38% higher for HIV-positive than for HIV-negative individuals.

This understanding has led to changes in HIV management guidelines. For instance, the most recent HIV Medicine Association guidelines in the U.S. recommend dual energy x-ray absorptiometry (DEXA) bone density screening for men and women over age 50. The European AIDS Clinical Society (EACS) also recommends DEXA screening for men over 50 and postmenopausal women, but in addition recommends fracture risk assessment using the FRAX tool, without bone mineral density (BMD) assessment, for HIV-positive people over the age of 40.

As noted, FRAX is a web-based fracture risk calculator that uses easily obtainable clinical risk factors (used with or without BMD) to estimate absolute 10-year risk of having a major osteoporotic (spine, upper arm, wrist, or hip) fracture or of having a hip fracture alone. FRAX has been well validated in the general population across many different countries and populations, but there has been a question whether FRAX works equally well at predicting fractures in HIV-positive and HIV-negative individuals.

In several small studies, FRAX scores were not predictive of osteoporosis (bone loss) by DEXA score or prevalent fractures in people living with HIV. Since FRAX can underestimate fracture risk in people with HIV, some experts have advised entering "yes" for the secondary osteoporosis question, to serve as a surrogate for HIV infection -- even though HIV is not listed among the known causes of secondary osteoporosis -- in order to increase the accuracy of FRAX scores for HIV-positive people. 

The Study

There have been no studies of people with HIV looking at how well FRAX (unadjusted or with the expert-recommended tweak for HIV) predicts incident, or newly diagnosed fractures, which is what the tool was designed to do. So Yin and colleagues designed a study to compare the accuracy of FRAX in HIV-positive versus HIV-negative men aged 50-70 years. They assumed that FRAX would underestimate risk, but that the recommended adjustment would improve accuracy.

For the study they selected a sample of men in the Veterans Aging Cohort Study between age 50 and 70 for whom complete data were available from the year 2000, in order to approximate all but 2 factors used in FRAX score calculation, since they did not have complete data on secondary osteoporosis causes and parental hip fracture. For this reason, this represented a "modified FRAX" calculation. (By convention, "no" was entered for the categories of parental hip fracture and secondary osteoporosis for all participants because of the incomplete data.)

The researchers also had 10-year incident fracture data from 2001 through 2010 for new fractures of the spine, upper arm, wrist, and hip to correspond with those fractures that FRAX predicts. The resulting analytic sample included over 24,400 men, with nearly one-third being HIV-positive.

To determine how accurate FRAX estimates were in both the HIV-positive and HIV-negative populations, they looked at the agreement between observed fractures and estimated fracture predictions using the modified FRAX. They assessed the accuracy of the adjusted modified FRAX in people living with HIV, when HIV infection was treated as cause of secondary osteoporosis. They also looked at discrimination, or the ability of the FRAX estimate to discriminate between individuals who would have and who would not have a fracture.

At baseline, the participants' median age was 56 and was similar in both the HIV-positive and HIV-negative groups. However, the men with HIV were more likely to be black, weighed less, drank more alcohol, and had a higher prevalence of previous fractures, but reported less current smoking and were less likely to have rheumatoid arthritis than the HIV-negative men.

Major Findings

Observed and estimated rates of incident fracture: The modified FRAX estimates suggested that over the 10-year period, about 3.0% of the participants would have a fracture at a major osteoporotic site and about 0.3% would have a hip fracture. The actual observed rate was significantly higher: a 4.6% fracture risk at major osteoporotic sites among the HIV-positive men (vs 3.5% among HIV-negative men), and 1.3% fracture risk at the hip (vs .09%). Both the estimated fracture and the observed fracture rates were higher among HIV-positive than among HIV-negative men.

Modified FRAX with HIV as a cause of secondary osteoporosis: When HIV was treated as a cause of secondary osteoporosis, the estimates of fracture risk went up by 31% for a major osteoporotic fracture and by about 67% for hip fractures.

Accuracy of modified FRAX worse in HIV-positive men: The observed fracture versus estimated fracture ratios were 1.6 for HIV-positive men and 1.3 for HIV-negative men (the higher the ratio, the greater the under-estimation). However, the adjustment (treating HIV as a cause of secondary osteoporosis) improved the observed versus estimated fracture ratio to levels similar to those for FRAX in the HIV-negative population. The accuracy was also worse among those individuals with a higher fracture risk. 

Pharmacologic Treatment Thresholds for FRAX

In the U.S., the first step for the evaluation of osteoporosis is a DEXA screening. FRAX is used for individuals who do not meet the DEXA criteria for osteoporosis but who have either low bone density or osteopenia (less severe bone loss, with a T-score between -1.0 and -2.5). FRAX calculations are then recommended to see which individuals are at a high enough risk for fractures to justify pharmacologic treatment. The thresholds are set at  >20% formajor osteoporotic fractures or >3% for hip fractures.

In this study population, none of the participants met the criteria for pharmacologic therapy using the threshold of >20% formajor osteoporotic fracture sites. Using the threshold of >3% for hip fractures, 1% of the population -- or 74 men -- met the criteria.

However, using this threshold would have identified only 3 out of the 93 individuals who actually had a hip fracture during this period, yielding a specificity of 99% but a sensitivity of only 3%.

"So using this threshold for the modified-FRAX, 71 men would be treated unnecessarily to prevent 1 fracture -- and 97% of men will get a fracture because they remain untreated," Yin said.

There were several limitations to the study, the first being that this was a modified FRAX, as the researchers were missing data on 2 variables used in the tool. In addition, the cohort only included men, which limits the generalizability of the findings to women. Also, although this was a group of men over the age of 50, only 10% of study participants were over the age of 65, and fractures risk rises with increasing age. The researchers also did not explore the impact of hepatitis C in FRAX, though they are doing this currently.

In conclusion, Yin said, "The modified FRAX accuracy is limited for HIV-positive men and the exact role of FRAX for risk stratification in HIV-infected individuals -- both for DEXA screening and for pharmacologic therapy -- requires additional study. "



M Yin, S Shiau, K Harwood, et al. Fracture Prediction with Modified FRAX in Older HIV+ and HIV- Men.  2015 Conference on Retroviruses and Opportunistic Infections. Seattle, February 23-24, 2015. Abstract 141.