Prediabetes and Fracture Risk Among Midlife Women in the Study of Women’s Health Across the Nation

Key Points Question Is prediabetes associated with increased fracture risk among midlife women? Findings In this cohort study of 1690 midlife women without diabetes from the Study of Women’s Health Across the Nation cohort, relative to not having prediabetes at any visit before the menopause transition (MT), having prediabetes at every pre-MT visit was associated with a statistically significant 120% greater hazard for fracture during the MT and after menopause. This association was independent of bone mineral density at the start of the MT. Meaning This study suggests that, for midlife women, prediabetes may be a risk factor for future fractures.


Introduction
2][3][4][5] At the present time, whether prediabetes is also a risk factor for fractures is uncertain.7][8] Moreover, at levels observed in individuals without type 2 diabetes, greater insulin resistance is associated with a lower bone mineral density (BMD), 9 a lower trabecular bone score, 8 lower indices of hip strength, 10,11 and faster bone loss, 8 all risk factors for fractures.
Elucidating whether prediabetes increases fracture risk is a step in understanding its clinical relevance.Although prediabetes is a risk factor for developing diabetes, clinicians disagree on how aggressively to treat it. 12,13Not every person with prediabetes develops type 2 diabetes, and prediabetes itself has not been definitively associated with end-organ complications.With respect to bone health, whether prediabetes is associated with fractures in the absence of prior or future progression to type 2 diabetes is unknown.
The first objective of this study was to examine whether prediabetes among midlife women is associated with subsequent fracture in the absence of type 2 diabetes.We focused on midlife, when women undergo the menopause transition (MT) and fracture risk accelerates. 14,15Because prediabetes may affect BMD, 16 our second objective was to assess whether the potential association of prediabetes with fracture was independent of BMD.

Methods
This cohort study used data collected between January 6, 1996, and February 28, 2018, in the Study of Women's Health Across the Nation (SWAN), a multicenter cohort study of 3302 diverse, community-dwelling women.At the SWAN baseline visit, participants were 42 to 52 years of age and in premenopause (no change from usual menstrual bleeding) or early perimenopause (less predictable menstrual bleeding at least once every 3 months).Potential participants were excluded if they did not have an intact uterus and 1 or more ovaries or were using hormonal therapy or hormonal contraception.Women were recruited at 7 clinical sites: Boston, Massachusetts; Chicago, Illinois; Detroit, Michigan; Pittsburgh, Pennsylvania; Los Angeles, California; Newark, New Jersey; and

Samples
To be included in this analysis, SWAN Bone Cohort participants needed 1 or more study visits before the MT, could not be taking a bone-beneficial medication before the MT, could not have type 2

JAMA Network Open | Diabetes and Endocrinology
Prediabetes and Fracture Risk Among Midlife Women

Outcome
The outcome was time to first fracture after the start of the MT.Fracture occurrence and location were ascertained using standardized questionnaires at all study visits.Pre-SWAN fractures were recorded at the SWAN baseline visit.SWAN initiated fracture adjudication at follow-up visit 7; for the first 6 follow-up visits, fracture date was imputed as the midpoint between the participant's previous and index visits.Since adjudication began, 95% of reported fractures were confirmed.Fractures were atraumatic if they occurred after a fall from a height of less than 15.2 cm and if they did not occur during a motor vehicle accident, rapid movement, playing sports, or from impact with heavy or fastmoving projectiles.We excluded craniofacial and digital fractures but included atraumatic and traumatic fractures because both fracture types are risk factors for subsequent fractures. 17,18imary Exposure The primary exposure was prediabetes before the MT.We modeled our prediabetes exposure as the proportion of visits from the SWAN baseline visit through the last visit before the MT at which prediabetes was present.Criteria for prediabetes were a fasting blood glucose level between 100 and 125 mg/dL and not taking a diabetes medication.The prediabetes exposure was a continuous variable with values ranging from 0 to 1, which allowed us to capture how consistently a participant had prediabetes.[21] Fasting blood glucose was measured at 2 different central laboratories in SWAN.Through follow-up visit 7, glucose was measured at Medical Research Laboratories (Lexington, Kentucky) using a hexokinase-coupled reaction assay (Roche; intra-assay coefficient of variability, 1.6%).
Subsequent glucose measurements were performed at the University of Michigan, Ann Arbor, using the ADVIA Chemistry Glucose Hexokinase assay (intra-assay coefficient of variability, 0.7%-0.9%).
A between-laboratory calibration equation was developed using 565 randomly selected values across the range of glucose results.This equation was applied to convert Medical Research Laboratories results to corresponding University of Michigan values.

JAMA Network Open | Diabetes and Endocrinology
Prediabetes and Fracture Risk Among Midlife Women

Covariates
Analyses controlled for variables that were potentially associated with fracture.These variables included age at the start of the MT, body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) at the start of the MT, cigarette use (yes or no) at the start of the MT, fracture before the MT (yes or no), race and ethnicity (Black, Chinese, Japanese, or White) by selfreport, study site, exposure to bone-detrimental medications before the MT, and exposure to bonedetrimental medications during fracture observation.Bone-detrimental medications were oral or injectable glucocorticoids, aromatase inhibitors, gonadotropin-releasing hormone agonists, or antiepileptic medications.Exposure to these medications before the MT or during fracture observation was estimated as the proportion of visits during the respective time period at which use was reported.We adjusted for exposure to bone-detrimental medications, instead of excluding users, because very few SWAN participants took these drugs consistently over prolonged intervals. 15 address the second study objective, we adjusted for lumbar spine (LS) or femoral neck (FN) BMD at the start of the MT.Lumbar spine and FN areal BMD were measured by dual x-ray absorptiometry using Hologic Inc instruments.SWAN's protocols for cross-site calibration, crosscalibration after machine upgrades, and quality control have been described previously. 22

Statistical Analysis
Statistical analysis was conducted from January to May 2022.We assessed all continuous variables for normality.Using the Spearman rank correlation, we examined the correlations between the proportion of visits before the MT with prediabetes and the mean glucose level across pre-MT visits with prediabetes.
We conducted 2 sets of primary analyses.The first (model 1) examined whether prediabetes before the MT was associated with subsequent fracture.We used Cox proportional hazards regression with the proportion of visits before the MT with prediabetes as the primary exposure and the time to first fracture after the start of the MT as the outcome.Covariates were age, BMI, and cigarette use at the start of the MT; exposure to bone-detrimental medication before the MT or after the start of the MT; fracture before the MT; race and ethnicity; and study site.Our second set of analyses assessed whether the hypothesized association of prediabetes with fracture was independent of BMD by adding to model 1 controls for LS (model 2A) or FN (model 2B) BMD.In all models, we added quadratic terms for the continuous prediabetes exposure and BMI to test for nonlinearity in their association with fracture.Neither quadratic term made statistically significant associations and were dropped from the final models.
Because the primary exposure is a continuous variable with values ranging from 0 to 1, the hazard ratio (HR) for prediabetes generated by the Cox proportional hazards regression model is the HR for prediabetes at all pre-MT visits (exposure = 1) relative to prediabetes at no pre-MT visit (exposure = 0).Because the primary exposure is continuous and had a linear association with fracture hazard, the estimated HR (for exposure = 1) can be converted to HRs for different exposure values between 0 and 1, corresponding to an individual woman's proportion of pre-MT visits with prediabetes.
We conducted 2 sets of sensitivity analyses.First, we reran the models but used a binary exposure: any prediabetes before the MT (yes or no).Second, we reran models 1, 2A, and 2B using the original continuous prediabetes exposure but included only major nonvertebral fractures (defined in some osteoporosis therapy trials as fractures of the pelvis, distal femur, proximal tibia, ribs, proximal humerus, forearm, and hip) as outcomes. 23,24l analyses were performed using Stata, version 17 (StataCorp LLC).We used a 2-tailed significance level of .05. proportion of pre-MT visits with prediabetes was 0.070 (0.208).Fifty-six women (3.3%) had a fracture before the MT.Mean (SD) follow-up from the start of the MT to fracture or censoring was 12 (6) years.One-hundred thirty-six (8.0%) women sustained a fracture during the MT or in postmenopause.Of the 225 women with prediabetes, 25 (11.1%)sustained a fracture, while 111 of the 1465 women without prediabetes (7.6%) sustained a fracture.Thirty-three women were censored for initiating a bone-beneficial medication, and 94 for incident diabetes.

Participant Characteristics
A total of 225 participants (13.3%) had prediabetes at 1 or more study visits before the MT (Table 1).A greater proportion of Black, Chinese, or Japanese women were represented in this group.
Also, among these 225 participants, the mean (SD) proportion of pre-MT visits with prediabetes was 0.524 (0.293), and prediabetes criteria were met in the majority of (Ն50%) pre-MT visits for 124 women (55.1%) or all pre-MT visits for 47 women (20.9%) women.Seventy-three participants (32.4%) did not meet prediabetes criteria at any visit after their first prediabetes visit.Spearman rank correlations between the proportion of visits before the MT with prediabetes and the mean fasting glucose level at pre-MT visits with prediabetes was 0.83.At the start of the MT, the mean (SD) BMI among women with prediabetes was 31.7 (7.0), and the mean (SD) BMD was 1.093 (0.141) g/cm 2 at the LS and 0.869 (0.130) g/cm 2 at the FN.Among participants with prediabetes, the mean (SD) follow-up from the start of the MT to the time of fracture or censoring was 10 (6) years.Twenty-five women with prediabetes (11.1%) sustained a fracture, 1 was censored for starting a bone-beneficial medication, and 62 were censored for incident type 2 diabetes.

Association of Prediabetes Before the MT With Subsequent Fractures
Having prediabetes more consistently before the MT was associated with greater rates of fracture during the MT or in postmenopause.In Cox proportional hazards regression, adjusted for age, BMI, cigarette use, bone-detrimental medications, fracture before the MT, race and ethnicity, and study site (model 1), relative to prediabetes at no pre-MT visit, prediabetes at every visit before the MT was associated with a 120% greater hazard of subsequent fracture (HR, 2.20 [95% CI, 1.11-4.37])(Table 2).Compared with no prediabetes, HRs for prediabetes at 25%, 50%, and 75% of visits before the MT were 1.22 (95% CI, 1.03-1.45),1.48 (95% CI, 1.05-2.09),and 1.81 (95% CI, 1.08-3.02),respectively.The observed fracture hazard of 6.3 per 1000 person-years for women with no prediabetes before the MT translated to an absolute increase of approximately 3 fractures per 1000 person-years for women who had prediabetes at half of the pre-MT visits and 7 fractures per 1000 person-years for women who had prediabetes at all pre-MT visits.
Adding additional controls for LS (model 2A) or FN (model 2B) BMD at the start of the MT did not substantially alter the magnitude or statistical significance of the association between prediabetes and fracture.Hazard ratios for prediabetes at all pre-MT visits vs no pre-MT visits in models 2A and 2B were 2.24 (95% CI, 1.12-4.46)and 2.26 (95% CI, 1.13-4.49),respectively (Table 2).

Sensitivity Analyses
Our first set of sensitivity analyses used a binary prediabetes exposure (any prediabetes before the MT) instead of the proportion of visits before the MT with prediabetes.The HR for any prediabetes before the MT were b Prediabetes defined as a fasting blood glucose level between 100 and 125 mg/dL (to convert to millimoles per liter, multiply by 0.0555), without use of diabetes medication.
c Proportion of visits at which prediabetes criteria were met or bone-detrimental medication use was reported.This was a continuous variable with values ranging from 0 to 1.For example, women without prediabetes had values of 0. Women with prediabetes at all visits before the MT had values of 1. Women with prediabetes at 1 or more, but not all, pre-MT visits had values between 0 and 1.
d Bone-detrimental medications were oral or injectable glucocorticoids, aromatase inhibitors, gonadotropin-releasing hormone agonists, or antiepileptic medications.Participants were considered users if they reported any use (regardless of duration).

JAMA Network Open | Diabetes and Endocrinology
Prediabetes and Fracture Risk Among Midlife Women In our second set of sensitivity analyses, we retained the continuous prediabetes exposure from our primary models but considered only major nonvertebral fractures (n = 63) as outcomes.The HR for prediabetes at all vs no pre-MT visits was 2.58 (95% CI, 0.99-6.83)adjusted for BMD, 2.65 (95% CI, 1.00-7.03)adjusted for LS BMD, and 2.71 (95% CI, 1.02-7.17)adjusted for FN BMD.

Discussion
In this longitudinal cohort study of the association between prediabetes in midlife women and fractures, we found that prediabetes before the MT was associated with more subsequent fractures; having prediabetes at 50% or 100% of pre-MT visits was associated with 49% and 120% greater hazard, respectively, for fracture during the MT and in postmenopause.The observed fracture hazard of 6.3 per 1000 person-years for women with no prediabetes before the MT translated to an absolute increase of approximately 3 and 7 fractures per 1000 person-years for women who had prediabetes at half of the pre-MT visits and all pre-MT visits, respectively.This increase in fracture risk was specific to prediabetes and was not associated with overt type 2 diabetes.Similar to type 2 diabetes, 1-5 however, the prediabetes-fracture association was independent of BMD, suggesting that pathways other than BMD are associated with this risk.
[27] Among ). 27r analysis differed from previously published studies in 3 major ways.First was our prediabetes exposure, which models the consistency of prediabetes over time.This contrasts with prior studies that assessed glycemic status at 1 time point.Our finding that the proportion of visits with prediabetes had a positive, linear association with fracture suggests that more consistent prediabetes could be detrimental to fracture risk.Greater prediabetes consistency may indicate more severe prediabetes (inferred by the positive correlation between prediabetes consistency and Abbreviations: BMD, bone mineral density; HR, hazard ratio; MT, menopause transition. a Hazard ratios estimated using Cox proportional hazards regression with prediabetes before the MT as the primary exposure and time to fracture after start of the MT as the outcome.Start of the MT was defined as first visit in late perimenopause (less predictable menses once every 3-12 months), or first visit in postmenopause if participant transitioned directly from premenopause (no change in menstrual bleeding patterns) or early perimenopause (less predictable menstrual bleeding once every 1-3 months) to late perimenopause.Prediabetes was modeled as proportion of visits from the Study of Women's Health Across the Nation baseline visit until last visit before the MT start that participants met prediabetes criteria.This was a continuous exposure with values ranging from 0 (prediabetes at 0 pre-MT visits) to 1 (prediabetes at all pre-MT visits).Women who had prediabetes at 1 or more, but not all, pre-MT visits had exposure values between 0 and 1. Model 1 was adjusted for age at start of the MT, body mass index at start of the MT, cigarette use at start of the MT, fracture before the MT, use of bone-detrimental medications before the MT, use of bone-detrimental medications during fracture observation, race and ethnicity, and study site.Models 2A and 2B additionally controlled for lumbar spine or femoral neck BMD, respectively.Because the primary exposure was continuous and had a linear association with fracture hazard, the estimated HR (for exposure = 1) can be converted to HRs for different exposure values between 0 and 1, corresponding to an individual woman's proportion of pre-MT visits with prediabetes.To calculate the HR for a given proportion of visits before the MT with prediabetes, raise the HR presented to the desired proportion.Thus, the unadjusted HRs for prediabetes at 25%, 50%, or 75% of visits before the MT were 1.21 (2.20 0.25 ), 1.49 (2.20 0.5 ), and 1.81 (2.2 0.75 ), respectively.

JAMA Network Open | Diabetes and Endocrinology
Prediabetes and Fracture Risk Among Midlife Women fasting blood glucose) and longer duration of prediabetes Second, whereas previous studies examined older adults, we analyzed midlife women undergoing the MT, a period when insulin resistance and fracture risk accelerate. 14,15It may be easier to discern a prediabetes-fracture association during the MT because physiological changes are larger.Third, we designed our analysis such that the observed association of prediabetes with fracture could not be attributed to future progression to type 2 diabetes; prior studies did not censor participants at first type 2 diabetes diagnosis.
Our results add to a body of data suggesting that prediabetes is detrimental to bone health.
Several recent SWAN analyses showed that prediabetes was associated with worse trabecular bone microarchitecture 8 and that greater insulin resistance was associated with diminished bone strength 10 and faster BMD loss. 22[30][31] Establishing whether prediabetes is a risk factor for fractures has potential public health implications.Nearly 1 in 3 US adults has prediabetes, 32 but clinicians debate the need to treat it, in part because prediabetes has not been directly linked to end-organ complications. 13Our results suggest that more consistently having a fasting blood glucose level in the prediabetes range before the MT may be an independent risk factor for fracture.4][35][36] Thus, for this population, prediabetes before the MT may be an early, modifiable risk factor for fracture.

Limitations
This study has several limitations.First, to maximize our fracture outcomes, we included both minimal trauma and traumatic fracture in a composite outcome.We justify this approach given the increasing recognition that both fracture types are associated with future fractures. 17,18Some randomized clinical trials assessing the antifracture efficacy of osteoporosis treatments now include minimal trauma and traumatic fractures as outcomes. 37Second, the rates of prediabetes and fractures differed by race and ethnicity.However, we did not have sufficient power to test whether the association between prediabetes and fractures differed by race and ethnicity.

Conclusions
In this cohort study of midlife women, prediabetes before the MT was associated with greater risk of subsequent fractures during the MT and after menopause, independent of BMD.Because midlife fractures are associated with subsequent fractures in older age, future research could examine whether treating prediabetes before the MT reduces the risk of different fracture outcomes (eg, hip, vertebral, nonvertebral, major osteoporotic) in later life.
Oakland, California.The SWAN Bone Cohort included 2365 women from 5 sites (excluding Chicago and Newark).Since study inception in 1996, 1 baseline visit and 16 follow-up visits have occurred at a median of 1.1 years (IQR, 1.0-1.4years) between consecutive visits.All study volunteers provided written informed consent, and each site obtained institutional review board approval (University of Michigan; Massachusetts General Hospital; Rush University Medical Center; University of California, Davis, and Kaiser Permanente; University of California, Los Angeles; Albert Einstein College of Medicine; and University of Pittsburgh).This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for observational studies.
diabetes before the MT, and needed at least 1 study visit after the start of the MT (to permit observation for fractures).Bone-beneficial medications included hormone therapy, calcitonin, calcitriol, bisphosphonates, denosumab, and parathyroid hormone.We defined the start of the MT as the first visit in late perimenopause (less predictable menstrual bleeding at least once every 3-12 months).For women who transitioned directly from premenopause or early perimenopause to postmenopause ("skipped" late perimenopause), we defined the start of the MT as the first JAMA Network Open.2023;6(5):e2314835.doi:10.1001/jamanetworkopen.2023.14835(Reprinted) May 23, 2023 2/11 Downloaded From: https://jamanetwork.com/ on 10/23/2023

Table 2 .
Association of Prediabetes Before the MT With Subsequent Fracture During the MT and in Postmenopause a,b