[Skip to Content]
[Skip to Content Landing]
Figure 1.  Hip Fracture Hazard Ratios
Hip Fracture Hazard Ratios

Cox regression analysis results are presented as hazard ratios, with 95% CIs presented as whiskers. Fully adjusted models controlled for age, sociodemographic characteristics, receipt of fracture-protective drugs, and comorbidities. The unit of analysis was person-day. The models revealed risk associated with current receipt of 1 of 21 individual fracture-associated drug (FAD) groups, current receipt of 1 of 210 possible FAD pairs, or any concurrent receipt of 3 or more FADs (eTable 6 in the Supplement). Non-FAD indicates other systemic drugs.

Figure 2.  Hip Fracture Risk Associated With Select Fracture-Associated Drugs (FADs) and 2-Way Combinations of FADs
Hip Fracture Risk Associated With Select Fracture-Associated Drugs (FADs) and 2-Way Combinations of FADs

Fully adjusted Cox regression analysis results are presented as hazard ratios and 95% CIs. Models were stratified by sex and adjusted for age, sociodemographic characteristics, receipt of fracture-protective drugs, comorbidities, and non-FAD drug receipt. Results displayed were selected for apparent population-level impact; specifically, combinations met all 4 of the following criteria: (1) common (ie, top quintile of person-year exposure); (2) associated with a 2-fold or greater increase in sex-specific, crude, absolute risk; (3) associated with a relative risk of 3.00 or more; and (4) associated with 50 or more fractures in our cohort. Only results that achieved a false-discovery rate P < .05 were considered significant. The unit of analysis was person-day. Full models provided in eTable 7 and eTable 8 in the Supplement. PPI indicates proton pump inhibitor; SSRI/SNRI, selective serotonin reuptake inhibitor and selective noradrenergic reuptake inhibitor; SGAP, second-generation antipsychotic.

Table 1.  Characteristics of Cohort, Composed of a 20% Random Sample of Fee-for-Service Medicare Beneficiaries Meeting Inclusion Criteria, 2006-2014
Characteristics of Cohort, Composed of a 20% Random Sample of Fee-for-Service Medicare Beneficiaries Meeting Inclusion Criteria, 2006-2014
Table 2.  Fracture Count, Distribution of FAD and Non-FAD Exposure Intensity, and Crude Hip Fracture Incidence Rate
Fracture Count, Distribution of FAD and Non-FAD Exposure Intensity, and Crude Hip Fracture Incidence Rate
1.
Rosen  CJ.  Serotonin rising: the bone, brain, bowel connection.  N Engl J Med. 2009;360(10):957-959. doi:10.1056/NEJMp0810058PubMedGoogle ScholarCrossref
2.
Eom  CS, Lee  HK, Ye  S, Park  SM, Cho  KH.  Use of selective serotonin reuptake inhibitors and risk of fracture: a systematic review and meta-analysis.  J Bone Miner Res. 2012;27(5):1186-1195. doi:10.1002/jbmr.1554PubMedGoogle ScholarCrossref
3.
Eom  C-S, Park  SM, Myung  S-K, Yun  JM, Ahn  J-S.  Use of acid-suppressive drugs and risk of fracture: a meta-analysis of observational studies.  Ann Fam Med. 2011;9(3):257-267. doi:10.1370/afm.1243PubMedGoogle ScholarCrossref
4.
Xiao  X, Xu  Y, Wu  Q.  Thiazide diuretic usage and risk of fracture: a meta-analysis of cohort studies.  Osteoporos Int. 2018;29(7):1515-1524. doi:10.1007/s00198-018-4486-9PubMedGoogle ScholarCrossref
5.
Tseng  PT, Chen  YW, Yeh  PY, Tu  KY, Cheng  YS, Wu  CK.  Bone mineral density in schizophrenia: an update of current meta-analysis and literature review under guideline of PRISMA.  Medicine (Baltimore). 2015;94(47):e1967. doi:10.1097/MD.0000000000001967PubMedGoogle Scholar
6.
Woolcott  JC, Richardson  KJ, Wiens  MO,  et al.  Meta-analysis of the impact of 9 medication classes on falls in elderly persons.  Arch Intern Med. 2009;169(21):1952-1960. doi:10.1001/archinternmed.2009.357PubMedGoogle ScholarCrossref
7.
Ghosh  M, Majumdar  SR.  Antihypertensive medications, bone mineral density, and fractures: a review of old cardiac drugs that provides new insights into osteoporosis.  Endocrine. 2014;46(3):397-405. doi:10.1007/s12020-014-0167-4PubMedGoogle ScholarCrossref
8.
Aung  K, Htay  T.  Thiazide diuretics and the risk of hip fracture.  Cochrane Database Syst Rev. 2011;(10):CD005185. doi:10.1002/14651858.CD005185.pub2PubMedGoogle Scholar
9.
Hartikainen  S, Lönnroos  E, Louhivuori  K.  Medication as a risk factor for falls: critical systematic review.  J Gerontol A Biol Sci Med Sci. 2007;62(10):1172-1181.PubMedGoogle ScholarCrossref
10.
Donnelly  K, Bracchi  R, Hewitt  J, Routledge  PA, Carter  B.  Benzodiazepines, z-drugs and the risk of hip fracture: a systematic review and meta-analysis.  PLoS One. 2017;12(4):e0174730. doi:10.1371/journal.pone.0174730PubMedGoogle Scholar
11.
Cumming  RG, Le Couteur  DG.  Benzodiazepines and risk of hip fractures in older people: a review of the evidence.  CNS Drugs. 2003;17(11):825-837. doi:10.2165/00023210-200317110-00004PubMedGoogle ScholarCrossref
12.
Munson  JC, Bynum  JP, Bell  JE,  et al.  Patterns of prescription drug use before and after fragility fracture.  JAMA Intern Med. 2016;176(10):1531-1538. doi:10.1001/jamainternmed.2016.4814PubMedGoogle ScholarCrossref
13.
Munson  J, Morden  N, Goodman  D, Valle  L, Wennberg  J. The Dartmouth atlas of Medicare prescription drug use. https://www.rwjf.org/en/library/research/2013/10/the-dartmouth-atlas-of-medicare-prescription-drug-use.html. Accessed October 2, 2019.
14.
Hahn  TJ, Hendin  BA, Scharp  CR, Haddad  JG  Jr.  Effect of chronic anticonvulsant therapy on serum 25-hydroxycalciferol levels in adults.  N Engl J Med. 1972;287(18):900-904. doi:10.1056/NEJM197211022871803PubMedGoogle ScholarCrossref
15.
Kanis  JA, Johansson  H, Oden  A,  et al.  A meta-analysis of prior corticosteroid use and fracture risk.  J Bone Miner Res. 2004;19(6):893-899. doi:10.1359/JBMR.040134PubMedGoogle ScholarCrossref
16.
O’Brien  CA, Jia  D, Plotkin  LI,  et al.  Glucocorticoids act directly on osteoblasts and osteocytes to induce their apoptosis and reduce bone formation and strength.  Endocrinology. 2004;145(4):1835-1841. doi:10.1210/en.2003-0990PubMedGoogle ScholarCrossref
17.
Weinstein  RS, Jilka  RL, Parfitt  AM, Manolagas  SC.  Inhibition of osteoblastogenesis and promotion of apoptosis of osteoblasts and osteocytes by glucocorticoids: potential mechanisms of their deleterious effects on bone.  J Clin Invest. 1998;102(2):274-282. doi:10.1172/JCI2799PubMedGoogle ScholarCrossref
18.
Rejnmark  L, Vestergaard  P, Mosekilde  L.  Reduced fracture risk in users of thiazide diuretics.  Calcif Tissue Int. 2005;76(3):167-175. doi:10.1007/s00223-004-0084-2PubMedGoogle ScholarCrossref
19.
Schwartz  AV.  TZDs and bone: a review of the recent clinical evidence.  PPAR Res. 2008;2008:297893. doi:10.1155/2008/297893PubMedGoogle Scholar
20.
Benyamin  R, Trescot  AM, Datta  S,  et al.  Opioid complications and side effects.  Pain Physician. 2008;11(2)(suppl):S105-S120.PubMedGoogle Scholar
21.
Vuong  C, Van Uum  SH, O’Dell  LE, Lutfy  K, Friedman  TC.  The effects of opioids and opioid analogs on animal and human endocrine systems.  Endocr Rev. 2010;31(1):98-132. doi:10.1210/er.2009-0009PubMedGoogle ScholarCrossref
22.
Hugenholtz  GW, Heerdink  ER, van Staa  TP, Nolen  WA, Egberts  AC.  Risk of hip/femur fractures in patients using antipsychotics.  Bone. 2005;37(6):864-870. doi:10.1016/j.bone.2005.07.005PubMedGoogle ScholarCrossref
23.
Thorell  K, Ranstad  K, Midlöv  P, Borgquist  L, Halling  A.  Is use of fall risk-increasing drugs in an elderly population associated with an increased risk of hip fracture, after adjustment for multimorbidity level: a cohort study.  BMC Geriatr. 2014;14:131. doi:10.1186/1471-2318-14-131PubMedGoogle ScholarCrossref
24.
Takkouche  B, Montes-Martínez  A, Gill  SS, Etminan  M.  Psychotropic medications and the risk of fracture: a meta-analysis.  Drug Saf. 2007;30(2):171-184. doi:10.2165/00002018-200730020-00006PubMedGoogle ScholarCrossref
25.
Haentjens  P, Lamraski  G, Boonen  S.  Costs and consequences of hip fracture occurrence in old age: an economic perspective.  Disabil Rehabil. 2005;27(18-19):1129-1141. doi:10.1080/09638280500055529PubMedGoogle ScholarCrossref
26.
Braithwaite  RS, Col  NF, Wong  JB.  Estimating hip fracture morbidity, mortality and costs.  J Am Geriatr Soc. 2003;51(3):364-370. doi:10.1046/j.1532-5415.2003.51110.xPubMedGoogle ScholarCrossref
27.
Leal  J, Gray  AM, Prieto-Alhambra  D,  et al; REFReSH study group.  Impact of hip fracture on hospital care costs: a population-based study.  Osteoporos Int. 2016;27(2):549-558. doi:10.1007/s00198-015-3277-9PubMedGoogle ScholarCrossref
28.
Hoffman  GJ, Hays  RD, Shapiro  MF, Wallace  SP, Ettner  SL.  The costs of fall-related injuries among older adults: annual per-faller, service component, and patient out-of-pocket costs.  Health Serv Res. 2017;52(5):1794-1816. doi:10.1111/1475-6773.12554PubMedGoogle ScholarCrossref
29.
Heinrich  S, Rapp  K, Rissmann  U, Becker  C, König  HH.  Cost of falls in old age: a systematic review.  Osteoporos Int. 2010;21(6):891-902. doi:10.1007/s00198-009-1100-1PubMedGoogle ScholarCrossref
30.
Davis  JC, Robertson  MC, Ashe  MC, Liu-Ambrose  T, Khan  KM, Marra  CA.  International comparison of cost of falls in older adults living in the community: a systematic review.  Osteoporos Int. 2010;21(8):1295-1306. doi:10.1007/s00198-009-1162-0PubMedGoogle ScholarCrossref
31.
Morrison  A, Fan  T, Sen  SS, Weisenfluh  L.  Epidemiology of falls and osteoporotic fractures: a systematic review.  Clinicoecon Outcomes Res. 2013;5:9-18.PubMedGoogle Scholar
32.
Van Spall  HG, Toren  A, Kiss  A, Fowler  RA.  Eligibility criteria of randomized controlled trials published in high-impact general medical journals: a systematic sampling review.  JAMA. 2007;297(11):1233-1240. doi:10.1001/jama.297.11.1233PubMedGoogle ScholarCrossref
33.
Gurwitz  JH.  The exclusion of older people from participation in cardiovascular trials.  Virtual Mentor. 2014;16(5):365-368. doi:10.1001/virtualmentor.2014.16.05.pfor1-1405PubMedGoogle ScholarCrossref
34.
Bourgeois  FT, Orenstein  L, Ballakur  S, Mandl  KD, Ioannidis  JPA.  Exclusion of elderly people from randomized clinical trials of drugs for ischemic heart disease.  J Am Geriatr Soc. 2017;65(11):2354-2361. doi:10.1111/jgs.14833PubMedGoogle ScholarCrossref
35.
Avorn  J.  In defense of pharmacoepidemiology: embracing the yin and yang of drug research.  N Engl J Med. 2007;357(22):2219-2221. doi:10.1056/NEJMp0706892PubMedGoogle ScholarCrossref
36.
Benson  K, Hartz  AJ.  A comparison of observational studies and randomized, controlled trials.  N Engl J Med. 2000;342(25):1878-1886. doi:10.1056/NEJM200006223422506PubMedGoogle ScholarCrossref
37.
Radford  MJ, Foody  JM.  How do observational studies expand the evidence base for therapy?  JAMA. 2001;286(10):1228-1230. doi:10.1001/jama.286.10.1228PubMedGoogle ScholarCrossref
38.
Vandenbroucke  JP.  When are observational studies as credible as randomised trials?  Lancet. 2004;363(9422):1728-1731. doi:10.1016/S0140-6736(04)16261-2PubMedGoogle ScholarCrossref
39.
von Elm  E, Altman  DG, Egger  M, Pocock  SJ, Gøtzsche  PC, Vandenbroucke  JP; STROBE Initiative.  The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.  Ann Intern Med. 2007;147(8):573-577. doi:10.7326/0003-4819-147-8-200710160-00010PubMedGoogle ScholarCrossref
40.
Cumming  RG.  Epidemiology of medication-related falls and fractures in the elderly.  Drugs Aging. 1998;12(1):43-53. doi:10.2165/00002512-199812010-00005PubMedGoogle ScholarCrossref
41.
Ruths  S, Bakken  MS, Ranhoff  AH, Hunskaar  S, Engesæter  LB, Engeland  A.  Risk of hip fracture among older people using antihypertensive drugs: a nationwide cohort study.  BMC Geriatr. 2015;15:153. doi:10.1186/s12877-015-0154-5PubMedGoogle ScholarCrossref
42.
Motyl  KJ, DeMambro  VE, Barlow  D,  et al.  Propranolol attenuates risperidone-induced trabecular bone loss in female mice.  Endocrinology. 2015;156(7):2374-2383. doi:10.1210/en.2015-1099PubMedGoogle ScholarCrossref
43.
Rejnmark  L, Vestergaard  P, Mosekilde  L.  Treatment with beta-blockers, ACE inhibitors, and calcium-channel blockers is associated with a reduced fracture risk: a nationwide case-control study.  J Hypertens. 2006;24(3):581-589. doi:10.1097/01.hjh.0000203845.26690.cbPubMedGoogle ScholarCrossref
44.
Yang  S, Nguyen  ND, Eisman  JA, Nguyen  TV.  Association between beta-blockers and fracture risk: a Bayesian meta-analysis.  Bone. 2012;51(5):969-974. doi:10.1016/j.bone.2012.07.013PubMedGoogle ScholarCrossref
45.
Yang  S, Nguyen  ND, Center  JR, Eisman  JA, Nguyen  TV.  Association between beta-blocker use and fracture risk: the Dubbo Osteoporosis Epidemiology Study.  Bone. 2011;48(3):451-455. doi:10.1016/j.bone.2010.10.170PubMedGoogle ScholarCrossref
46.
Lewiecki  EM.  Treatment of low bone density or osteoporosis to prevent fractures in men and women.  Ann Intern Med. 2017;167(12):902. doi:10.7326/L17-0484PubMedGoogle ScholarCrossref
47.
Lewiecki  EM, Bilezikian  JP, Carey  JJ,  et al.  Proceedings of the 2017 Santa Fe Bone Symposium: insights and emerging concepts in the management of osteoporosis.  J Clin Densitom. 2018;21(1):3-21.PubMedGoogle ScholarCrossref
48.
Bhutani  G, Gupta  MC.  Emerging therapies for the treatment of osteoporosis.  J Midlife Health. 2013;4(3):147-152.PubMedGoogle Scholar
49.
Bynum  JPW, Bell  JE, Cantu  RV,  et al.  Second fractures among older adults in the year following hip, shoulder, or wrist fracture.  Osteoporos Int. 2016;27(7):2207-2215. doi:10.1007/s00198-016-3542-6PubMedGoogle ScholarCrossref
50.
Wennberg  JE, Cooper  MM; Dartmouth Atlas of Health Care Working Group. The Dartmouth atlas of health care, 1998. https://www.dartmouthatlas.org/downloads/atlases/98Atlas.pdf. Accessed October 2, 2019.
51.
Charlson  M, Szatrowski  TP, Peterson  J, Gold  J.  Validation of a combined comorbidity index.  J Clin Epidemiol. 1994;47(11):1245-1251. doi:10.1016/0895-4356(94)90129-5PubMedGoogle ScholarCrossref
52.
Fried  TR, O’Leary  J, Towle  V, Goldstein  MK, Trentalange  M, Martin  DK.  Health outcomes associated with polypharmacy in community-dwelling older adults: a systematic review.  J Am Geriatr Soc. 2014;62(12):2261-2272. doi:10.1111/jgs.13153PubMedGoogle ScholarCrossref
53.
Marcum  ZA, Perera  S, Thorpe  JM,  et al; Health ABC Study, USA.  Anticholinergic use and recurrent falls in community-dwelling older adults: findings from the Health ABC Study.  Ann Pharmacother. 2015;49(11):1214-1221. doi:10.1177/1060028015596998PubMedGoogle ScholarCrossref
54.
Andersen  PK, Gill  RD.  Cox’s regression model for counting processes: a large sample study.  Ann Stat. 1982;10(4):1100-1120. doi:10.1214/aos/1176345976Google ScholarCrossref
55.
Hosmer  DW, Lemeshow  S, May  S. Appendix 2: an introduction to the counting process approach to survival analysis. In: Hosmer DW, Lemeshow S, May S.  Applied Survival Analysis: Regression Modeling of Time-to-Event Data. 2nd ed. Hoboken, NJ: Wiley and Sons; 359-363. doi:10.1002/9780470258019.app2
56.
Glickman  ME, Rao  SR, Schultz  MR.  False discovery rate control is a recommended alternative to Bonferroni-type adjustments in health studies.  J Clin Epidemiol. 2014;67(8):850-857. doi:10.1016/j.jclinepi.2014.03.012PubMedGoogle ScholarCrossref
57.
National Committee for Quality Assurance. HEDIS measures and technical resources. https://www.ncqa.org/hedis/measures. Accessed October 2, 2019.
58.
Masnoon  N, Shakib  S, Kalisch-Ellett  L, Caughey  GE.  What is polypharmacy? a systematic review of definitions.  BMC Geriatr. 2017;17(1):230. doi:10.1186/s12877-017-0621-2PubMedGoogle ScholarCrossref
59.
Carnahan  RM, Lund  BC, Perry  PJ, Pollock  BG, Culp  KR.  The Anticholinergic Drug Scale as a measure of drug-related anticholinergic burden: associations with serum anticholinergic activity.  J Clin Pharmacol. 2006;46(12):1481-1486. doi:10.1177/0091270006292126PubMedGoogle ScholarCrossref
60.
Kanis  JA, Hans  D, Cooper  C,  et al; Task Force of the FRAX Initiative.  Interpretation and use of FRAX in clinical practice.  Osteoporos Int. 2011;22(9):2395-2411. doi:10.1007/s00198-011-1713-zPubMedGoogle ScholarCrossref
61.
Fracture Risk Assessment Tool. Welcome to FRAX. https://www.sheffield.ac.uk/FRAX/. Accessed October 2, 2019.
62.
Hammond  T, Wilson  A.  Polypharmacy and falls in the elderly: a literature review.  Nurs Midwifery Stud. 2013;2(2):171-175. doi:10.5812/nms.10709PubMedGoogle ScholarCrossref
63.
de Jong  MR, Van der Elst  M, Hartholt  KA.  Drug-related falls in older patients: implicated drugs, consequences, and possible prevention strategies.  Ther Adv Drug Saf. 2013;4(4):147-154. doi:10.1177/2042098613486829PubMedGoogle ScholarCrossref
64.
Dhalwani  NN, Fahami  R, Sathanapally  H, Seidu  S, Davies  MJ, Khunti  K.  Association between polypharmacy and falls in older adults: a longitudinal study from England.  BMJ Open. 2017;7(10):e016358.doi:10.1136/bmjopen-2017-016358PubMedGoogle Scholar
65.
Machado-Duque  ME, Castaño-Montoya  JP, Medina-Morales  DA, Castro-Rodríguez  A, González-Montoya  A, Machado-Alba  JE.  Association between the use of benzodiazepines and opioids with the risk of falls and hip fractures in older adults.  Int Psychogeriatr. 2018;30(7):941-946. doi:10.1017/S1041610217002745PubMedGoogle ScholarCrossref
66.
Aspinall  SL, Springer  SP, Zhao  X,  et al.  Central nervous system medication burden and risk of recurrent serious falls and hip fractures in Veterans Affairs nursing home residents.  J Am Geriatr Soc. 2019;67(1):74-80. doi:10.1111/jgs.15603PubMedGoogle ScholarCrossref
67.
Hanlon  JT, Zhao  X, Naples  JG,  et al.  Central nervous system medication burden and serious falls in older nursing home residents.  J Am Geriatr Soc. 2017;65(6):1183-1189. doi:10.1111/jgs.14759PubMedGoogle ScholarCrossref
68.
King  T, Vardanyan  A, Majuta  L,  et al.  Morphine treatment accelerates sarcoma-induced bone pain, bone loss, and spontaneous fracture in a murine model of bone cancer.  Pain. 2007;132(1-2):154-168. doi:10.1016/j.pain.2007.06.026PubMedGoogle ScholarCrossref
69.
Daniell  HW.  Opioid osteoporosis.  Arch Intern Med. 2004;164(3):338. author reply. doi:10.1001/archinte.164.3.338-aPubMedGoogle ScholarCrossref
70.
Guo  Z, Wills  P, Viitanen  M, Fastbom  J, Winblad  B.  Cognitive impairment, drug use, and the risk of hip fracture in persons over 75 years old: a community-based prospective study.  Am J Epidemiol. 1998;148(9):887-892. doi:10.1093/oxfordjournals.aje.a009714PubMedGoogle ScholarCrossref
71.
Janas  A, Folwarczna  J.  Opioid receptor agonists may favorably affect bone mechanical properties in rats with estrogen deficiency-induced osteoporosis.  Naunyn Schmiedebergs Arch Pharmacol. 2017;390(2):175-185. doi:10.1007/s00210-016-1295-6PubMedGoogle ScholarCrossref
72.
Endo  Y, Aharonoff  GB, Zuckerman  JD, Egol  KA, Koval  KJ.  Gender differences in patients with hip fracture: a greater risk of morbidity and mortality in men.  J Orthop Trauma. 2005;19(1):29-35. doi:10.1097/00005131-200501000-00006PubMedGoogle ScholarCrossref
73.
Panula  J, Pihlajamäki  H, Mattila  VM,  et al.  Mortality and cause of death in hip fracture patients aged 65 or older: a population-based study.  BMC Musculoskelet Disord. 2011;12:105. doi:10.1186/1471-2474-12-105PubMedGoogle ScholarCrossref
74.
Haentjens  P, Magaziner  J, Colón-Emeric  CS,  et al.  Meta-analysis: excess mortality after hip fracture among older women and men.  Ann Intern Med. 2010;152(6):380-390. doi:10.7326/0003-4819-152-6-201003160-00008PubMedGoogle ScholarCrossref
75.
Berry  SD, Zhu  Y, Choi  H, Kiel  DP, Zhang  Y.  Diuretic initiation and the acute risk of hip fracture.  Osteoporos Int. 2013;24(2):689-695. doi:10.1007/s00198-012-2053-3PubMedGoogle ScholarCrossref
76.
Tudor  CG. Transition to Part D coverage of benzodiazepines and barbiturates beginning in 2013. https://www.cms.gov/Medicare/Prescription-Drug-Coverage/PrescriptionDrugCovContra/Downloads/BenzoandBarbituratesin2013.pdf. Accessed October 2, 2019.
77.
Zhu  ZN, Jiang  YF, Ding  T.  Risk of fracture with thiazolidinediones: an updated meta-analysis of randomized clinical trials.  Bone. 2014;68:115-123. doi:10.1016/j.bone.2014.08.010PubMedGoogle ScholarCrossref
78.
Shen  C, Chen  F, Zhang  Y, Guo  Y, Ding  M.  Association between use of antiepileptic drugs and fracture risk: a systematic review and meta-analysis.  Bone. 2014;64:246-253. doi:10.1016/j.bone.2014.04.018PubMedGoogle ScholarCrossref
79.
Lee  SH, Hsu  WT, Lai  CC,  et al.  Use of antipsychotics increases the risk of fracture: a systematic review and meta-analysis.  Osteoporos Int. 2017;28(4):1167-1178. doi:10.1007/s00198-016-3881-3PubMedGoogle ScholarCrossref
80.
Oderda  LH, Young  JR, Asche  CV, Pepper  GA.  Psychotropic-related hip fractures: meta-analysis of first-generation and second-generation antidepressant and antipsychotic drugs.  Ann Pharmacother. 2012;46(7-8):917-928. doi:10.1345/aph.1Q589PubMedGoogle ScholarCrossref
81.
Rabenda  V, Nicolet  D, Beaudart  C, Bruyère  O, Reginster  JY.  Relationship between use of antidepressants and risk of fractures: a meta-analysis.  Osteoporos Int. 2013;24(1):121-137. doi:10.1007/s00198-012-2015-9PubMedGoogle ScholarCrossref
82.
Torstensson  M, Hansen  AH, Leth-Møller  K,  et al.  Danish register-based study on the association between specific cardiovascular drugs and fragility fractures.  BMJ Open. 2015;5(12):e009522. doi:10.1136/bmjopen-2015-009522PubMedGoogle Scholar
83.
Li  L, Setoguchi  S, Cabral  H, Jick  S.  Opioid use for noncancer pain and risk of fracture in adults: a nested case-control study using the general practice research database.  Am J Epidemiol. 2013;178(4):559-569. doi:10.1093/aje/kwt013PubMedGoogle ScholarCrossref
84.
Yang  YX, Lewis  JD, Epstein  S, Metz  DC.  Long-term proton pump inhibitor therapy and risk of hip fracture.  JAMA. 2006;296(24):2947-2953. doi:10.1001/jama.296.24.2947PubMedGoogle ScholarCrossref
85.
Ryu  JY, Kim  HU, Lee  SY.  Deep learning improves prediction of drug-drug and drug-food interactions.  Proc Natl Acad Sci U S A. 2018;115(18):E4304-E4311. doi:10.1073/pnas.1803294115PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    Geriatrics
    November 13, 2019

    Association of Receiving Multiple, Concurrent Fracture-Associated Drugs With Hip Fracture Risk

    Author Affiliations
    • 1The Dartmouth Institute for Health Policy and Clinical Practice, Geisel School of Medicine, Dartmouth College, Lebanon, New Hampshire
    • 2Division of Geriatric and Palliative Medicine, Internal Medicine, Institute of Healthcare Policy and Innovation, University of Michigan, Ann Arbor
    • 3Center for Clinical and Translational Research, Maine Medical Center Research Institute, Scarborough
    • 4now with Microsoft Artificial Intelligence and Research, Healthcare NeXT, Redmond, Washington
    JAMA Netw Open. 2019;2(11):e1915348. doi:10.1001/jamanetworkopen.2019.15348
    Key Points español 中文 (chinese)

    Question  Is the concurrent receipt of 2 fracture-associated drugs associated with increased risk of hip fracture?

    Findings  In this cohort study, 11 million person-years of Medicare data revealed that the receipt of multiple fracture-associated drugs was common among older US residents. Concurrent receipt of 2 or more such drugs was associated with a more than 2-fold increase in hip fracture risk, with some specific combinations appearing especially hazardous, including commonly prescribed drugs, such as opioids, antidepressants, and sedatives.

    Meaning  If confirmed, results of this cohort study suggest caution when combining fracture-associated medications, especially when use is discretionary, alternatives exist, or baseline fracture risk is high.

    Abstract

    Importance  Many prescription drugs increase fracture risk, which raises concern for patients receiving 2 or more such drugs concurrently. Logic suggests that risk will increase with each additional drug, but the risk of taking multiple fracture-associated drugs (FADs) is unknown.

    Objective  To estimate hip fracture risk associated with concurrent exposure to multiple FADs.

    Design, Setting, and Participants  This cohort study used a 20% random sample of Medicare fee-for-service administrative data for age-eligible Medicare beneficiaries from 2004 to 2014. Sex-stratified Cox regression models estimated hip fracture risk associated with current receipt of 1, 2, or 3 or more of 21 FADs and, separately, risk associated with each FAD and 2-way FAD combination vs no FADs. Models included sociodemographic characteristics, comorbidities, and use of non-FAD medications. Analyses began in November 2018 and were completed April 2019.

    Exposure  Receipt of prescription FADs.

    Main Outcomes and Measures  Hip fracture hospitalization.

    Results  A total of 11.3 million person-years were observed, reflecting 2 646 255 individuals (mean [SD] age, 77.2 [7.3] years, 1 615 613 [61.1%] women, 2 136 585 [80.7%] white, and 219 579 [8.3%] black). Overall, 2 827 284 person-years (25.1%) involved receipt of 1 FAD; 1 322 296 (11.7%), 2 FADs; and 954 506 (8.5%), 3 or more FADs. In fully adjusted, sex-stratified models, an increase in hip fracture risk among women was associated with the receipt of 1, 2, or 3 or more FADs (1 FAD: hazard ratio [HR], 2.04; 95% CI, 1.99-2.11; P < .001; 2 FADs: HR, 2.86; 95% CI, 2.77-2.95; P < .001; ≥3 FADs: HR, 4.50; 95% CI, 4.36-4.65; P < .001). Relative risks for men were slightly higher (1 FAD: HR, 2.23; 95% CI, 2.11-2.36; P < .001; 2 FADs: HR, 3.40; 95% CI, 3.20-3.61; P < .001; ≥3 FADs: HR, 5.18; 95% CI, 4.87-5.52; P < .001). Among women, 2 individual FADs were associated with HRs greater than 3.00; 80 pairs of FADs exceeded this threshold. Common, risky pairs among women included sedative hypnotics plus opioids (HR, 4.90; 95% CI, 3.98-6.02; P < .001), serotonin reuptake inhibitors plus benzodiazepines (HR, 4.50; 95% CI, 3.76-5.38; P < .001), and proton pump inhibitors plus opioids (HR, 4.00; 95% CI, 3.56-4.49; P < .001). Receipt of 1, 2, or 3 or more non-FADs was associated with a small, significant reduction in fracture risk compared with receipt of no non-FADs among women (1 non-FAD: HR, 0.93; 95% CI, 0.90-0.96; P < .001; 2 non-FADs: HR, 0.84; 95% CI, 0.81-0.87; P < .001; ≥3 non-FADs: HR, 0.74; 95% CI, 0.72-0.77; P < .001).

    Conclusions and Relevance  Among older adults, FADs are commonly used and commonly combined. In this cohort study, the addition of a second and third FAD was associated with a steep increase in fracture risk. Many risky pairs of FADs included potentially avoidable drugs (eg, sedatives and opioids). If confirmed, these findings suggest that fracture risk could be reduced through tighter adherence to long-established prescribing guidelines and recommendations.

    Introduction

    At least 21 prescription drug groups have been associated with increased fracture risk.1-12 The list includes some of the most commonly used products in the United States, such as opioids, proton pump inhibitors (PPIs), serotonin reuptake inhibitors (SSRIs), and sedatives.13 The wide use of many fracture-associated drugs (FADs) suggests that concurrent receipt of 2 or more may be prevalent. This in turn raises concern for fracture risk associated with such exposure. Logic suggests that risk will increase with the addition of each additional FAD. However, the fracture risk of such combined drug use has not been quantified. Understanding the prevalence of concurrent FAD exposure and its risk is especially important to older people, who have the highest baseline rate of fragility fracture and who are most likely to receive multiple FADs as diagnoses and treatments accumulate with age.

    Research suggests individual FADs increase fracture risk by weakening bones, increasing falls, or both. Pathways of bone weakening may involve direct skeletal cell effects, disruption of vitamin D or calcium management, modulation of parathyroid hormone effects, and hypogonadism.14-21 The mechanisms by which medications increase falls include sedation, loss of balance, dizziness, visual disturbance, muscle weakening, and hypotension.6,9,22-24 A total of 8 drug groups have been associated with both bone weakening and falls (eTable 1 in the Supplement). The diversity of mechanisms makes the net effect of exposure to multiple FADs difficult to predict. Risk associated with exposure to 2 FADs acting through the same mechanism could prove additive if together the FADs intensely activate the pathway of harm. Conversely, 2 such drugs could prove redundant, resulting in no additional risk, if 1 drug alone saturates the pathway. Illuminating and quantifying this potential iatrogenic source of fracture risk could inform changes in prescribing practices aimed at reducing risk and the consequences associated with fractures.25-31

    Ideally, the study of risks associated with FAD combinations would occur through randomized clinical trials (RCTs). However, RCTs addressing this question are not readily feasible because of the large number of potential drug combinations and the need for enormous sample sizes to detect rare outcomes. Ethical concerns would be raised as well by trials randomizing people not to receive drugs known to be effective for specific conditions. Additionally, RCTs often exclude multimorbid, older individuals and thus may not reveal the risk of such drug regimens in the populations most likely to receive them.32-34 Well-conducted, population-representative, observational studies thus represent a crucial opportunity to identify and estimate signals of harm from drug combinations. Such signals, if strong and reproducible, could inform basic science research exploring biological pathways of risk and perhaps effective harm reduction strategies; they could also help justify and prioritize carefully conducted, narrowly focused RCTs.35-38

    We used Medicare administrative data to examine hip fracture risk associated with prescription receipt. We first modeled hip fracture risk associated with overall drug count, then with FAD count and count of other drugs (non-FADS) to examine these broad categories separately. We then modeled risk of hip fracture associated with 21 individual FADs and their 210 possible 2-way combinations vs no FADs.

    Methods
    Data and Cohort

    We used Medicare Parts A (inpatient) and B (outpatient) administrative data from 2004 to 2014 and Medicare Part D (prescription) records from 2006 to 2014 for a 20% random sample of fee-for-service enrollees. For cohort inclusion, we identified age-eligible beneficiaries with a minimum of 24 months of Parts A and B coverage from 2004 to 2014 plus at least 12 additional months of A, B, and D coverage from 2006 to 2014 (eFigure 1 in the Supplement). Thus, age-eligible individuals were 67 years or older at the time of first observation. The 24 months before the observation period were required to ascertain past morbidities and past fractures. We excluded patients with a fragility fracture (ie, hip, wrist, humerus, and vertebra) apparent in the 24-month preobservation period. This exclusion conservatively created a relatively homogeneous, lower-risk cohort for exploration of FAD risks in the general, geriatric population. Additionally, patients with any of the following diagnoses or services during the study period were excluded: cancer (except nonmelanoma skin cancer), advanced renal disease, vertebral fracture, or hospice. Fracture and disease definitions appear in eTable 2 in the Supplement, and exclusion justification appears in the eAppendix in the Supplement. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.39 The Dartmouth Protection of Human Subjects Committee approved this study and granted a waiver of written informed consent because risk was deemed minimal and research involved materials collected for nonresearch purposes.

    Prescription Receipt

    From published studies and meta-analyses we identified 21 distinct FAD groups including, antipsychotics, opioids, thiazide diuretics, loop diuretics, glucocorticoids, PPIs, sedative hypnotics, and SSRIs.1-12,40 The full drug list is provided in eTable 3 in the Supplement. We identified prescription receipt from the Medicare Prescription Drug Event file. For analysis, the data were structured such that, for each drug group, each person-day was assigned as currently exposed or unexposed. Current exposure was assumed to begin on the date of prescription dispensing. We used days supply to set exposure duration. This same approach was applied to other systemic drugs (non-FADs) which were aggregated into 4 current exposure variables (ie, 0, 1, 2, and ≥3 non-FADs) for models. The eAppendix in the Supplement explains FAD and non-FAD ascertainment via National Drug Codes identified by Lexicomp Basic database (Wolters Kluwer) and First Databank (First Databank), respectively. Three types of fracture-protective drugs were also recorded as time varying by person-day and included in each model as currently exposed or not exposed, as follows: (1) estrogen (eg, systemic estrogen, selective estrogen receptor modulators), (2) osteoporosis medications (eg, bisphosphonates, calcitonin, parathyroid hormone, denosumab), and (3) β-blockers.41-48 Each intravenous zoledronic acid dose (retrieved from the Medicare Part B Carrier File) was assumed to confer 12 months of bisphosphonate exposure (eAppendix and eTable 3 in the Supplement).

    Outcome

    We identified hip fracture diagnosis codes from inpatient claims.12,49 Outcome date was hip fracture admission date. We focused on hip fractures because inpatient-treated hip fractures are temporally discrete events with a precise event date.50 This permits accurate sequencing of exposure and outcome, assuring we don’t attribute fracture risk to drugs received after the fracture. This is especially important in the study of opioids, which are often used to treat fracture pain (eTable 2 in the Supplement).

    Covariates

    We ascertained the following covariates for model inclusion. From the Medicare enrollment files, we obtained age (classified as aged 67-69 years; thereafter in 5-year groups), sex, race/ethnicity, original Medicare entitlement reason, Medicaid eligibility, and Part D Low-Income Subsidy status. We dichotomized the latter 2 variables for each calendar year, assigning the status if present 1 or more months. Long-term care (LTC) status was assigned if 50% or more of prescriptions in a year were dispensed by an LTC pharmacy (retrieved from Part D Pharmacy Characteristics file). From inpatient and outpatient claims, we identified 22 morbidities and assigned patients to each (dichotomous indicator variable) on the first observed diagnosis code at any point in the study. In selecting morbidities for models, we began with the Charlson Comorbidity Index list.51 From this list we dropped cohort exclusion conditions (ie, renal failure and cancer) as well as HIV/AIDS because of extremely low prevalence; we then added conditions associated with falls or fractures, eg, seizure disorder, osteoporosis, tobacco use, depression, rheumatologic disease, and alcohol use (eTable 2 in the Supplement). Morbidities and LTC were considered absorptive states, ie, once present, always present.

    Statistical Analysis

    Our principal aim was to estimate hip fracture risk associated with current exposure to FAD combinations. Preliminary models estimated risk associated with overall number of drugs.52,53 Main models estimated risk associated with FAD count and non-FAD count and, separately, risk associated with individual FADs and 210 possible 2-way FAD combinations. All models were stratified by sex, an anticipated effect modifier.

    For time-varying exposure analysis, each person-day was assigned 1 of 233 mutually exclusive FAD exposure categories: 0 FADs, 21 single FADs, 210 possible 2-way FAD combinations, or an aggregate of 3 or more FADs. For statistical reliability, 2-way exposures with less than 100 person-years (PYs) observation in each sex-specific cohort were grouped into a category called FAD pairs with fewer than 100 PYs. We calculated crude fractures per 1000 PYs and percentage of total PYs of observation for each exposure category by sex. Age group–specific crude fracture rates were also calculated.

    We applied Cox regression models to counting-process data54,55 and estimated drug-associated fracture risk in the 3 following ways: (1) risk associated with current total drug count (FADs and non-FADs), (2) risk associated with current FAD count category and non-FAD count category (1, 2, and ≥3) vs no drugs, and (3) risk associated with 231 single and 2-way exposure indicators plus an indicator of 3 or more concurrent FADs vs no FADs. These models of specific single and 2-way FAD pairs included indicators of non-FAD exposure (ie, 0, 1, 2, or ≥3). All models included the covariates described earlier, ie, age, sociodemographic factors, comorbidities, and 3 indicators for fracture-protective drugs.

    In these models, individuals may have contributed to a variety of exposure groups (including no exposure) as the number and type of drugs they received changed over time. The regression coefficient for each exposure indicator was the fracture risk associated with being in that receipt status vs receiving no FADs. Patients were censored at death; disenrollment from fee-for-service Parts A, B, or D; or first hip fracture.

    To explore residual confounding, we repeated sex-specific, categorical exposure (ie, 0, 1, 2, or ≥3) models, stratified by age group. We reasoned that drug count may be a proxy for unrecorded disease and disease severity (potential confounders); both commonly increase with age. We expected fracture risk associated with FADs and non-FADs to increase across progressively older age groups if residual confounding were substantial.

    Sensitivity analyses included models with no drug exposure indicators; these were intended to explore estimates associated with known risk factors, such as chronic conditions, when detailed drug exposure was not considered (eTable 6 in the Supplement). Separate sensitivity analyses repeated main models but included consideration of competing risk of death using both subdistribution and cause-specific hazards (eAppendix in the Supplement).

    Analyses were initiated in November 2018 and completed April 2019 and conducted with SAS version 4 (SAS Institute). Statistical tests were 2-sided. After false discovery rate control using the Benjamini-Hochberg procedures,56 risk estimates with P < .05 were considered significant.

    Results
    Cohort Characteristics

    Our cohort included 2 646 255 beneficiaries with 11 286 768 PYs of observation. Overall, 1 615 613 beneficiaries (61.1%) were women, 2 136 585 (80.7%) were white individuals, and 219 579 (8.3%) were black individuals. Mean (SD) age was 77.2 (7.3) years at study conclusion. Mean (SD) observation time was 4.3 (2.8) years. Over the 9-year study, 726 704 cohort members (27.5%) received Part D LIS, 611 006 (23.1%) were Medicaid eligible, and 211 128 (8.0%) received an LTC assignment (Table 1).

    Exposure and Outcomes

    Overall, 2 827 284 PYs (25.1%) involved receipt of 1 FAD; 1 322 296 (11.7%), 2 FADs; and 954 505 (8.5%), 3 or more FADs. Among women, 1 827 305 PYs (25.6%) were exposed to 1 FAD, 912 855 (12.8%) to 2 FADs, and 705 500 (9.9%) to 3 or more FADs; exposure in men was slightly lower (1 FAD, 999 979 PYs [24.0%]; 2 FADs, 409 441 PYs [9.8%]; ≥3 FADs, 249 006 PYs [6.0%]) (Table 2). Over the 9-year period, 1 000 118 women (61.9%) and 533 604 men (51.8%) received 2 FADs on 1 or more days; 640 178 women (39.6%) and 294 181 men (28.5%) received 3 or more concurrent FADs on 1 or more days (eTable 4 in the Supplement). Women received more fracture-protective drugs than men. The most common FAD exposures (alone or in combination with other FADs) were opioids (463 419 PYs of exposure), thiazides (1 447 094 PYs of exposure), prescription PPIs (941 159 PYs of exposure), and SSRIs (763 953 PYs of exposure) (eTable 4 in the Supplement). After uncommon 2-way exposures were grouped as FAD pairs with fewer than 100 PYs, the model for women included 204 exposures (27 pairs grouped), and the model for men included 186 exposures (45 pairs grouped) (eTable 5 in the Supplement).

    We observed 59 704 hip fractures, 47 386 (79.4%) among women and 12 317 (20.6%) among men. Crude hip fracture rate was 6.65 per 1000 PYs among women and 2.96 per 1000 PYs among men (Table 2), replicating published rates.50

    Fracture Risk Associated With Drug Count

    For women, crude fracture rate associated with 0 FADs was 3.34 per 1000 PYs; 1 FAD, 6.47 per 1000 PYs; 2 concurrent FADs, 10.38 per 1000 PYs; and 3 or more FADs, 19.55 per 1000 PYs. Among men, corresponding rates were 1.45 per 1000 PYs, 3.15 per 1000 PYs, 5.95 per 1000 PYs, and 12.51 per 1000 PYs, respectively. Crude fracture rates associated with non-FADs increased in parallel with drug count but less steeply (eg, for women receiving 0 vs ≥3 non-FADs, 4.44/1000 PYs vs 8.36/1000 PYs; for men receiving 0 vs ≥3 non-FADs, 2.00/1000 PYs vs 3.71/1000 PYs) (Table 2). Age-specific, crude fracture rates for FADs and non-FADs are presented in the eAppendix and eFigure 2 in the Supplement.

    Fully adjusted models exploring total overall drug count and hip fracture risk revealed a hazard ratio (HR) of 1.10 (95% CI, 1.09-1.10; P < .001) among women and an HR of 1.10 (95% CI, 1.09-1.11; P < .001) among men for each additional drug compared with none (eTable 6 in the Supplement). Fully adjusted models including FAD and non-FAD counts exhibited diverging risks for these broad drug categories. For women, compared with receiving 0 FADs, receiving 1 FAD was associated with an HR of 2.04 (95% CI, 1.99-2.11; P < .001); receiving 2 FADs was associated with an HR of 2.86 (95% CI, 2.77-2.95; P < .001); and receiving 3 FADs or more was associated with an HR of 4.50 (95% CI, 4.36-4.65; P < .001). Comparable non-FAD exposure was associated with an HR of 0.93 (95% CI, 0.90-0.96; P < .001), an HR of 0.84 (95% CI, 0.81-0.87; P < .001), and an HR of 0.74 (95% CI, 0.72-0.77; P < .001), respectively (Figure 1A; eTable 6 in the Supplement). Among men, compared with receiving 0 FADs, receiving 1 FAD was associated with an HR of 2.23 (95% CI, 2.11- 2.36; P < .001); receiving 2 FADs was associated with an HR of 3.40 (95% CI, 3.20-3.61; P < .001); and receiving 3 or more FADs was associated with an HR of 5.18 (95% CI, 4.87-5.52; P < .001). Comparable non-FAD exposure was associated with an HR of 0.90 (95% CI 0.84-0.96; P < .001), an HR of 0.84 (95% CI, 0.79-0.89; P < .001), and an HR of 0.71 (95% CI, 0.67-0.75; P < .001) respectively (Figure 1A; eTable 6 in the Supplement). Age-stratified analyses revealed no consistent association between increasing age group and fracture risk associated with FAD count (Figure 1B and Figure 1C; eTable 6 in the Supplement).

    Fracture Risk Associated With Specific Drug Pairs

    In presenting results of models exploring over 200 distinct exposures, we provided all estimates in eTable 7 and eTable 8 in the Supplement but focused our report on exposures contributing relatively substantial population-level risk. We defined substantial population effect combinations as those meeting the 4 following criteria: (1) common (ie, top quintile of exposure PYs), (2) associated with a 2-fold or higher increase in sex-specific, crude, absolute risk, (3) associated with an HR of 3.00 or more, and (4) associated with 50 or more fractures in our cohort.

    Among women, 1 FAD group (ie, muscle relaxers) was associated with no increased fracture risk when received alone. Significant risks associated with individual FADs ranged from an HR of 1.54 (95% CI, 1.05-2.24; P = .03) for first-generation antipsychotics to an HR of 3.26 (95% CI, 3.04-3.49; P < .001) for opioids and an HR of 3.29 (95% CI, 2.89-3.76; P < .001) for anti-Parkinson drugs. Among women, only opioids and anti-Parkinson medications were associated with a relative risk above 3.00, while 80 combinations exceeded that threshold; 7 of these 80 met our population-level impact criteria. These included opioids plus sedative hypnotics (HR, 4.90; 95% CI, 3.98-6.02; P < .001), opioids plus loop diuretics (HR, 4.48; 95% CI, 3.96-5.07; P < .001), opioids plus PPIs (HR 4.00; 95% CI, 3.56-4.49; P < .001), SSRIs plus opioids (HR, 3.91; 95% CI, 3.46-4.43; P < .001), SSRIs plus benzodiazepines (HR, 4.50; 95% CI, 3.76-5.38; P < .001), SSRIs plus loop diuretics (HR 3.05; 95% CI, 2.75-3.37; P < .001), and nitrates plus loop diuretics (HR, 3.25; 95% CI, 2.84-3.72; P < .001) (Figure 2A; eTable 7 in the Supplement).

    Among men, significant risks of single FADs ranged from an HR of 1.51 (95% CI, 1.17-1.95; P = .002) for sedative hypnotics to an HR of 3.83 (95% CI, 3.36-4.36; P < .001) for opioids and an HR of 4.23 (95% CI 3.57-5.01; P < .001) for anti-Parkinson drugs. A total of 9 pairs met our population-level impact criteria; 5 paralleled women’s high-risk pairs (opioids plus loop diuretics, opioids plus PPIs, SSRIs plus opioids, SSRIs plus loop diuretics, and nitrates plus loop diuretics). Among men, the riskiest drug pairs were opioids plus loop diuretics (HR, 6.93; 95% CI, 5.52-8.70; P < .001) and opioids plus SSRIs (HR, 6.26; 95% CI, 4.83-8.12; P < .001) (Figure 2B; eTable 8 in the Supplement).

    For a small number of FAD pairs, receipt of 2 drugs appeared less risky than receipt of a single drug. For example, among women, loop diuretics alone were associated an HR of 2.45 (95% CI, 2.32-2.59; P < .001), but loop diuretics plus centrally acting antihypertensives were associated with an HR of 1.82 (95% CI, 1.33-2.49; P < .001) (eTable 7 in the Supplement).

    Discussion

    Among older Medicare beneficiaries, receipt of FAD combinations is common. Our findings indicated that fracture risk associated with some combinations exceeded individual drug risks. In our study, 20 of 21 examined drug groups were associated with, on average, an approximate 2-fold higher relative risk of hip fracture. Risk increased steeply with the addition of a second FAD (approximately 3-fold higher relative risk) and further with a third (more than 4-fold higher relative risk). Many of the most common, risky pairs included drugs that are relatively contraindicated in our older cohort (eg, sedative hypnotics and benzodiazepines)57 or potentially discretionary in some patients (eg, PPIs and opioids). If confirmed, our findings suggest that more cautious prescribing could reduce fracture risk with few or no negative consequences. The clinical implications of risk associated with essential medications (eg, loop diuretics and nitrates) will depend on anticipated net effect (ie, risk plus benefit). That opioids were included in many high-risk pairs reflected their frequent use and associated risk; this finding might reinforce ongoing scrutiny of opioid-prescribing practices.

    Our study built on past efforts to estimate risk associated with complex prescription drug combinations. Our results did not support a common approach of defining problematic drug regimens (ie, polypharmacy) based on numeric cutoffs. We found substantial risk associated with just 2 or 3 FADs, counts well below a popular polypharmacy definition of five or more.58 Equally important, we found receipt of 3 or more non-FADs associated with a slightly reduced risk of fracture, indicating that more may be better in some cases. These results suggested that a nuanced approach is needed to assess medication count, combinations, and risk. One such nuanced approach is exemplified by the anticholinergic burden framework that explicitly addresses the additive nature of adverse effects resulting from concurrent use of multiple anticholinergic drugs.59 Our study similarly considered combinations of drugs with an overlapping adverse effect risk. A total FAD burden conceptual framework may be valuable. Verified estimates could inform risk reduction efforts and risk calculators, such as the Fracture Risk Assessment Tool, which currently considers just 1 prescription group (ie, glucocorticoids).60,61

    While many studies demonstrate that both type and number of medications are associated with fall and fracture risk, our work expanded understanding by examining drugs associated with fracture through a fall-inducing mechanism and those associated with fracture because of bone weakening.62-67 We examined these drugs in detail through identification of time-varying exposure to specific FADs and specific FAD pairs, controlling for non-FAD drug receipt. We cannot tell which mechanisms contributed to observed excess risks. We can imagine that combining an opioid with a sedative hypnotic could result in extreme sedation or loss of balance, but the mechanism may be more complex. Like many FADs, opioids have been shown to increase falls and weaken bones.68-70 Disease-disease interactions or disease-drug interactions could contribute to observed effects. How a drug might mitigate the fracture-promoting effect of another is equally elusive. This has been suggested by others, but such research is nascent.42,71 Understanding the mechanisms of harm and protection could inform effective risk reduction strategies.

    We found FAD-associated relative risk of hip fracture higher among men than women. This excess relative risk occurred in the context of lower absolute risk for men, but the finding may have important implications because men experience higher mortality after hip fracture than women.72-74 Sex differences in FAD risks may be an important priority for future research.

    Limitations

    Our work has limitations. First, our observational study could suffer from residual confounding if FAD use is related to unobserved behavioral risk factors, unrecorded diseases, or disease severity beyond that controlled for (eg, biological risk factors such as balance and body mass). However, age-group stratified models revealed no increasing risk associated with increasing age group as might be expected given that unobservable confounding is likely more of a problem in older populations; this reduces but does not eliminate residual confounding concerns. Some diseases are inseparable from treatments; for example, Parkinson disease and treatments are both associated with fracture. Our models revealed how the addition of other, potentially discretionary FADs to Parkinson drugs may further increase risks. The estimates of known risk factors, such as age and comorbidities, were abrogated by increasing granularity of drug exposure, as demonstrated in eTable 6 in the Supplement.

    Second, our exposure measure was based on prescription fills. The extent to which patients fill but do not take medications, take medications over periods distinct from days supply, or obtain medications through unobserved channels will bias our results. Third, we observed current exposure and classified it dichotomously within drug groups; we did not consider dose, specific drugs within a drug group, cumulative exposure, or recency of FAD initiation. We expect neglecting these factors that influence fracture risk would shift our results toward the null. Future work considering these details will provide deeper understanding of risks because, for example, initiation of drug use has been associated with greatest risk.10,75 Fourth, we studied hip fractures; how FAD combinations affect other fragility fractures should be explored. Fifth, Part D coverage for benzodiazepines was limited until 2013; benzodiazepine exposure in early years is likely underestimated.76 Sixth, we observed effect estimates for single exposures that aligned with other large-scale studies; thus, our observational design and analytic approach replicated published results.3,23,75,77-84 Nonetheless, some drugs we classified as FADs have heterogeneous associations in the literature; some studies find them fracture-associated, others fracture-protective or neutral. We included them to be thorough and to add to the knowledge base. Furthermore, our results may not be generalizable to other populations.

    Conclusions

    Our cohort study is hypothesis generating and suggests future research directions. Follow-up studies should explore factors influencing use (and combination use) of potentially avoidable drugs associated with substantial fracture risk. To support clinical decision-making, studies should explore FAD risk in the context of anticipated benefits of essential drugs. Our findings suggested a need to update and refine our conceptual approach to identifying problematic drug regimens, moving past count and appropriateness in favor of nuanced considerations of overlapping risks and benefits. Large data sets and emerging advanced analytic approaches, such as machine learning, may facilitate these efforts as that science and its application to health care data mature.85 The exploration of mechanisms through which FAD pairs confer excess risk or risk-mitigating effects could inform efforts aimed at minimizing harm. Such investigations may require basic science approaches considering drug effects at the cellular level and using rodent models involving dual drug exposure and bone-strength measurements. Studies such as ours could inform, justify, and prioritize carefully designed, narrowly targeted human RCTs that quantify comparative risk of treatment alternatives.

    Back to top
    Article Information

    Accepted for Publication: September 24, 2019.

    Published: November 13, 2019. doi:10.1001/jamanetworkopen.2019.15348

    Correction: This article was corrected on December 11, 2019, to fix an error in Figure 1A.

    Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2019 Emeny RT et al. JAMA Network Open.

    Corresponding Author: Nancy E. Morden, MD, MPH, The Dartmouth Institute for Health Policy and Clinical Practice, One Medical Center Drive, WTRB 516, Lebanon, NH 03756 (nancy.e.morden@dartmouth.edu).

    Author Contributions: Drs Emeny and Morden had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Chang, Skinner, Rosen, Morden.

    Acquisition, analysis, or interpretation of data: Emeny, Chang, Skinner, O’Malley, Smith, Chakraborti, Morden.

    Drafting of the manuscript: Emeny, O’Malley, Chakraborti, Rosen, Morden.

    Critical revision of the manuscript for important intellectual content: Emeny, Chang, Skinner, O’Malley, Smith, Morden.

    Statistical analysis: Emeny, Chang, Skinner, O’Malley, Smith, Chakraborti, Morden.

    Obtained funding: Skinner, Morden.

    Administrative, technical, or material support: Emeny, O’Malley, Smith.

    Supervision: Skinner, Rosen.

    Conflict of Interest Disclosures: Dr Skinner reported serving as a consultant for Sutter Health; serving as a program director for the National Bureau of Economic Research; and investing in Dorsata Inc outside the submitted work. Ms Chakraborti reported receiving grants from Synergy during the conduct of the study. No other disclosures were reported.

    Funding/Support: Drs Emeny, Skinner, and Morden and Ms Chakraborti were supported by The Dartmouth Clinical and Translational Science Institute, under award number UL1TR001086 from the National Center for Advancing Translational Sciences of the National Institutes of Health. Drs Chang, Skinner, O’Malley, and Morden and Mr Smith were funded by grant P01 AG019783 from the National Institute on Aging (Dr Skinner, principal investigator; Causes and Consequences of Healthcare Efficiency). Dr Morden was supported by grant K23AG035030 from the National Institute of Aging (The Risk of Polypharmacy-Related Fracture in Older Adults). Dr Rosen was funded by grant U54GM II5516-01 from the Northern New England Clinical and Translational Research Network.

    Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

    References
    1.
    Rosen  CJ.  Serotonin rising: the bone, brain, bowel connection.  N Engl J Med. 2009;360(10):957-959. doi:10.1056/NEJMp0810058PubMedGoogle ScholarCrossref
    2.
    Eom  CS, Lee  HK, Ye  S, Park  SM, Cho  KH.  Use of selective serotonin reuptake inhibitors and risk of fracture: a systematic review and meta-analysis.  J Bone Miner Res. 2012;27(5):1186-1195. doi:10.1002/jbmr.1554PubMedGoogle ScholarCrossref
    3.
    Eom  C-S, Park  SM, Myung  S-K, Yun  JM, Ahn  J-S.  Use of acid-suppressive drugs and risk of fracture: a meta-analysis of observational studies.  Ann Fam Med. 2011;9(3):257-267. doi:10.1370/afm.1243PubMedGoogle ScholarCrossref
    4.
    Xiao  X, Xu  Y, Wu  Q.  Thiazide diuretic usage and risk of fracture: a meta-analysis of cohort studies.  Osteoporos Int. 2018;29(7):1515-1524. doi:10.1007/s00198-018-4486-9PubMedGoogle ScholarCrossref
    5.
    Tseng  PT, Chen  YW, Yeh  PY, Tu  KY, Cheng  YS, Wu  CK.  Bone mineral density in schizophrenia: an update of current meta-analysis and literature review under guideline of PRISMA.  Medicine (Baltimore). 2015;94(47):e1967. doi:10.1097/MD.0000000000001967PubMedGoogle Scholar
    6.
    Woolcott  JC, Richardson  KJ, Wiens  MO,  et al.  Meta-analysis of the impact of 9 medication classes on falls in elderly persons.  Arch Intern Med. 2009;169(21):1952-1960. doi:10.1001/archinternmed.2009.357PubMedGoogle ScholarCrossref
    7.
    Ghosh  M, Majumdar  SR.  Antihypertensive medications, bone mineral density, and fractures: a review of old cardiac drugs that provides new insights into osteoporosis.  Endocrine. 2014;46(3):397-405. doi:10.1007/s12020-014-0167-4PubMedGoogle ScholarCrossref
    8.
    Aung  K, Htay  T.  Thiazide diuretics and the risk of hip fracture.  Cochrane Database Syst Rev. 2011;(10):CD005185. doi:10.1002/14651858.CD005185.pub2PubMedGoogle Scholar
    9.
    Hartikainen  S, Lönnroos  E, Louhivuori  K.  Medication as a risk factor for falls: critical systematic review.  J Gerontol A Biol Sci Med Sci. 2007;62(10):1172-1181.PubMedGoogle ScholarCrossref
    10.
    Donnelly  K, Bracchi  R, Hewitt  J, Routledge  PA, Carter  B.  Benzodiazepines, z-drugs and the risk of hip fracture: a systematic review and meta-analysis.  PLoS One. 2017;12(4):e0174730. doi:10.1371/journal.pone.0174730PubMedGoogle Scholar
    11.
    Cumming  RG, Le Couteur  DG.  Benzodiazepines and risk of hip fractures in older people: a review of the evidence.  CNS Drugs. 2003;17(11):825-837. doi:10.2165/00023210-200317110-00004PubMedGoogle ScholarCrossref
    12.
    Munson  JC, Bynum  JP, Bell  JE,  et al.  Patterns of prescription drug use before and after fragility fracture.  JAMA Intern Med. 2016;176(10):1531-1538. doi:10.1001/jamainternmed.2016.4814PubMedGoogle ScholarCrossref
    13.
    Munson  J, Morden  N, Goodman  D, Valle  L, Wennberg  J. The Dartmouth atlas of Medicare prescription drug use. https://www.rwjf.org/en/library/research/2013/10/the-dartmouth-atlas-of-medicare-prescription-drug-use.html. Accessed October 2, 2019.
    14.
    Hahn  TJ, Hendin  BA, Scharp  CR, Haddad  JG  Jr.  Effect of chronic anticonvulsant therapy on serum 25-hydroxycalciferol levels in adults.  N Engl J Med. 1972;287(18):900-904. doi:10.1056/NEJM197211022871803PubMedGoogle ScholarCrossref
    15.
    Kanis  JA, Johansson  H, Oden  A,  et al.  A meta-analysis of prior corticosteroid use and fracture risk.  J Bone Miner Res. 2004;19(6):893-899. doi:10.1359/JBMR.040134PubMedGoogle ScholarCrossref
    16.
    O’Brien  CA, Jia  D, Plotkin  LI,  et al.  Glucocorticoids act directly on osteoblasts and osteocytes to induce their apoptosis and reduce bone formation and strength.  Endocrinology. 2004;145(4):1835-1841. doi:10.1210/en.2003-0990PubMedGoogle ScholarCrossref
    17.
    Weinstein  RS, Jilka  RL, Parfitt  AM, Manolagas  SC.  Inhibition of osteoblastogenesis and promotion of apoptosis of osteoblasts and osteocytes by glucocorticoids: potential mechanisms of their deleterious effects on bone.  J Clin Invest. 1998;102(2):274-282. doi:10.1172/JCI2799PubMedGoogle ScholarCrossref
    18.
    Rejnmark  L, Vestergaard  P, Mosekilde  L.  Reduced fracture risk in users of thiazide diuretics.  Calcif Tissue Int. 2005;76(3):167-175. doi:10.1007/s00223-004-0084-2PubMedGoogle ScholarCrossref
    19.
    Schwartz  AV.  TZDs and bone: a review of the recent clinical evidence.  PPAR Res. 2008;2008:297893. doi:10.1155/2008/297893PubMedGoogle Scholar
    20.
    Benyamin  R, Trescot  AM, Datta  S,  et al.  Opioid complications and side effects.  Pain Physician. 2008;11(2)(suppl):S105-S120.PubMedGoogle Scholar
    21.
    Vuong  C, Van Uum  SH, O’Dell  LE, Lutfy  K, Friedman  TC.  The effects of opioids and opioid analogs on animal and human endocrine systems.  Endocr Rev. 2010;31(1):98-132. doi:10.1210/er.2009-0009PubMedGoogle ScholarCrossref
    22.
    Hugenholtz  GW, Heerdink  ER, van Staa  TP, Nolen  WA, Egberts  AC.  Risk of hip/femur fractures in patients using antipsychotics.  Bone. 2005;37(6):864-870. doi:10.1016/j.bone.2005.07.005PubMedGoogle ScholarCrossref
    23.
    Thorell  K, Ranstad  K, Midlöv  P, Borgquist  L, Halling  A.  Is use of fall risk-increasing drugs in an elderly population associated with an increased risk of hip fracture, after adjustment for multimorbidity level: a cohort study.  BMC Geriatr. 2014;14:131. doi:10.1186/1471-2318-14-131PubMedGoogle ScholarCrossref
    24.
    Takkouche  B, Montes-Martínez  A, Gill  SS, Etminan  M.  Psychotropic medications and the risk of fracture: a meta-analysis.  Drug Saf. 2007;30(2):171-184. doi:10.2165/00002018-200730020-00006PubMedGoogle ScholarCrossref
    25.
    Haentjens  P, Lamraski  G, Boonen  S.  Costs and consequences of hip fracture occurrence in old age: an economic perspective.  Disabil Rehabil. 2005;27(18-19):1129-1141. doi:10.1080/09638280500055529PubMedGoogle ScholarCrossref
    26.
    Braithwaite  RS, Col  NF, Wong  JB.  Estimating hip fracture morbidity, mortality and costs.  J Am Geriatr Soc. 2003;51(3):364-370. doi:10.1046/j.1532-5415.2003.51110.xPubMedGoogle ScholarCrossref
    27.
    Leal  J, Gray  AM, Prieto-Alhambra  D,  et al; REFReSH study group.  Impact of hip fracture on hospital care costs: a population-based study.  Osteoporos Int. 2016;27(2):549-558. doi:10.1007/s00198-015-3277-9PubMedGoogle ScholarCrossref
    28.
    Hoffman  GJ, Hays  RD, Shapiro  MF, Wallace  SP, Ettner  SL.  The costs of fall-related injuries among older adults: annual per-faller, service component, and patient out-of-pocket costs.  Health Serv Res. 2017;52(5):1794-1816. doi:10.1111/1475-6773.12554PubMedGoogle ScholarCrossref
    29.
    Heinrich  S, Rapp  K, Rissmann  U, Becker  C, König  HH.  Cost of falls in old age: a systematic review.  Osteoporos Int. 2010;21(6):891-902. doi:10.1007/s00198-009-1100-1PubMedGoogle ScholarCrossref
    30.
    Davis  JC, Robertson  MC, Ashe  MC, Liu-Ambrose  T, Khan  KM, Marra  CA.  International comparison of cost of falls in older adults living in the community: a systematic review.  Osteoporos Int. 2010;21(8):1295-1306. doi:10.1007/s00198-009-1162-0PubMedGoogle ScholarCrossref
    31.
    Morrison  A, Fan  T, Sen  SS, Weisenfluh  L.  Epidemiology of falls and osteoporotic fractures: a systematic review.  Clinicoecon Outcomes Res. 2013;5:9-18.PubMedGoogle Scholar
    32.
    Van Spall  HG, Toren  A, Kiss  A, Fowler  RA.  Eligibility criteria of randomized controlled trials published in high-impact general medical journals: a systematic sampling review.  JAMA. 2007;297(11):1233-1240. doi:10.1001/jama.297.11.1233PubMedGoogle ScholarCrossref
    33.
    Gurwitz  JH.  The exclusion of older people from participation in cardiovascular trials.  Virtual Mentor. 2014;16(5):365-368. doi:10.1001/virtualmentor.2014.16.05.pfor1-1405PubMedGoogle ScholarCrossref
    34.
    Bourgeois  FT, Orenstein  L, Ballakur  S, Mandl  KD, Ioannidis  JPA.  Exclusion of elderly people from randomized clinical trials of drugs for ischemic heart disease.  J Am Geriatr Soc. 2017;65(11):2354-2361. doi:10.1111/jgs.14833PubMedGoogle ScholarCrossref
    35.
    Avorn  J.  In defense of pharmacoepidemiology: embracing the yin and yang of drug research.  N Engl J Med. 2007;357(22):2219-2221. doi:10.1056/NEJMp0706892PubMedGoogle ScholarCrossref
    36.
    Benson  K, Hartz  AJ.  A comparison of observational studies and randomized, controlled trials.  N Engl J Med. 2000;342(25):1878-1886. doi:10.1056/NEJM200006223422506PubMedGoogle ScholarCrossref
    37.
    Radford  MJ, Foody  JM.  How do observational studies expand the evidence base for therapy?  JAMA. 2001;286(10):1228-1230. doi:10.1001/jama.286.10.1228PubMedGoogle ScholarCrossref
    38.
    Vandenbroucke  JP.  When are observational studies as credible as randomised trials?  Lancet. 2004;363(9422):1728-1731. doi:10.1016/S0140-6736(04)16261-2PubMedGoogle ScholarCrossref
    39.
    von Elm  E, Altman  DG, Egger  M, Pocock  SJ, Gøtzsche  PC, Vandenbroucke  JP; STROBE Initiative.  The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.  Ann Intern Med. 2007;147(8):573-577. doi:10.7326/0003-4819-147-8-200710160-00010PubMedGoogle ScholarCrossref
    40.
    Cumming  RG.  Epidemiology of medication-related falls and fractures in the elderly.  Drugs Aging. 1998;12(1):43-53. doi:10.2165/00002512-199812010-00005PubMedGoogle ScholarCrossref
    41.
    Ruths  S, Bakken  MS, Ranhoff  AH, Hunskaar  S, Engesæter  LB, Engeland  A.  Risk of hip fracture among older people using antihypertensive drugs: a nationwide cohort study.  BMC Geriatr. 2015;15:153. doi:10.1186/s12877-015-0154-5PubMedGoogle ScholarCrossref
    42.
    Motyl  KJ, DeMambro  VE, Barlow  D,  et al.  Propranolol attenuates risperidone-induced trabecular bone loss in female mice.  Endocrinology. 2015;156(7):2374-2383. doi:10.1210/en.2015-1099PubMedGoogle ScholarCrossref
    43.
    Rejnmark  L, Vestergaard  P, Mosekilde  L.  Treatment with beta-blockers, ACE inhibitors, and calcium-channel blockers is associated with a reduced fracture risk: a nationwide case-control study.  J Hypertens. 2006;24(3):581-589. doi:10.1097/01.hjh.0000203845.26690.cbPubMedGoogle ScholarCrossref
    44.
    Yang  S, Nguyen  ND, Eisman  JA, Nguyen  TV.  Association between beta-blockers and fracture risk: a Bayesian meta-analysis.  Bone. 2012;51(5):969-974. doi:10.1016/j.bone.2012.07.013PubMedGoogle ScholarCrossref
    45.
    Yang  S, Nguyen  ND, Center  JR, Eisman  JA, Nguyen  TV.  Association between beta-blocker use and fracture risk: the Dubbo Osteoporosis Epidemiology Study.  Bone. 2011;48(3):451-455. doi:10.1016/j.bone.2010.10.170PubMedGoogle ScholarCrossref
    46.
    Lewiecki  EM.  Treatment of low bone density or osteoporosis to prevent fractures in men and women.  Ann Intern Med. 2017;167(12):902. doi:10.7326/L17-0484PubMedGoogle ScholarCrossref
    47.
    Lewiecki  EM, Bilezikian  JP, Carey  JJ,  et al.  Proceedings of the 2017 Santa Fe Bone Symposium: insights and emerging concepts in the management of osteoporosis.  J Clin Densitom. 2018;21(1):3-21.PubMedGoogle ScholarCrossref
    48.
    Bhutani  G, Gupta  MC.  Emerging therapies for the treatment of osteoporosis.  J Midlife Health. 2013;4(3):147-152.PubMedGoogle Scholar
    49.
    Bynum  JPW, Bell  JE, Cantu  RV,  et al.  Second fractures among older adults in the year following hip, shoulder, or wrist fracture.  Osteoporos Int. 2016;27(7):2207-2215. doi:10.1007/s00198-016-3542-6PubMedGoogle ScholarCrossref
    50.
    Wennberg  JE, Cooper  MM; Dartmouth Atlas of Health Care Working Group. The Dartmouth atlas of health care, 1998. https://www.dartmouthatlas.org/downloads/atlases/98Atlas.pdf. Accessed October 2, 2019.
    51.
    Charlson  M, Szatrowski  TP, Peterson  J, Gold  J.  Validation of a combined comorbidity index.  J Clin Epidemiol. 1994;47(11):1245-1251. doi:10.1016/0895-4356(94)90129-5PubMedGoogle ScholarCrossref
    52.
    Fried  TR, O’Leary  J, Towle  V, Goldstein  MK, Trentalange  M, Martin  DK.  Health outcomes associated with polypharmacy in community-dwelling older adults: a systematic review.  J Am Geriatr Soc. 2014;62(12):2261-2272. doi:10.1111/jgs.13153PubMedGoogle ScholarCrossref
    53.
    Marcum  ZA, Perera  S, Thorpe  JM,  et al; Health ABC Study, USA.  Anticholinergic use and recurrent falls in community-dwelling older adults: findings from the Health ABC Study.  Ann Pharmacother. 2015;49(11):1214-1221. doi:10.1177/1060028015596998PubMedGoogle ScholarCrossref
    54.
    Andersen  PK, Gill  RD.  Cox’s regression model for counting processes: a large sample study.  Ann Stat. 1982;10(4):1100-1120. doi:10.1214/aos/1176345976Google ScholarCrossref
    55.
    Hosmer  DW, Lemeshow  S, May  S. Appendix 2: an introduction to the counting process approach to survival analysis. In: Hosmer DW, Lemeshow S, May S.  Applied Survival Analysis: Regression Modeling of Time-to-Event Data. 2nd ed. Hoboken, NJ: Wiley and Sons; 359-363. doi:10.1002/9780470258019.app2
    56.
    Glickman  ME, Rao  SR, Schultz  MR.  False discovery rate control is a recommended alternative to Bonferroni-type adjustments in health studies.  J Clin Epidemiol. 2014;67(8):850-857. doi:10.1016/j.jclinepi.2014.03.012PubMedGoogle ScholarCrossref
    57.
    National Committee for Quality Assurance. HEDIS measures and technical resources. https://www.ncqa.org/hedis/measures. Accessed October 2, 2019.
    58.
    Masnoon  N, Shakib  S, Kalisch-Ellett  L, Caughey  GE.  What is polypharmacy? a systematic review of definitions.  BMC Geriatr. 2017;17(1):230. doi:10.1186/s12877-017-0621-2PubMedGoogle ScholarCrossref
    59.
    Carnahan  RM, Lund  BC, Perry  PJ, Pollock  BG, Culp  KR.  The Anticholinergic Drug Scale as a measure of drug-related anticholinergic burden: associations with serum anticholinergic activity.  J Clin Pharmacol. 2006;46(12):1481-1486. doi:10.1177/0091270006292126PubMedGoogle ScholarCrossref
    60.
    Kanis  JA, Hans  D, Cooper  C,  et al; Task Force of the FRAX Initiative.  Interpretation and use of FRAX in clinical practice.  Osteoporos Int. 2011;22(9):2395-2411. doi:10.1007/s00198-011-1713-zPubMedGoogle ScholarCrossref
    61.
    Fracture Risk Assessment Tool. Welcome to FRAX. https://www.sheffield.ac.uk/FRAX/. Accessed October 2, 2019.
    62.
    Hammond  T, Wilson  A.  Polypharmacy and falls in the elderly: a literature review.  Nurs Midwifery Stud. 2013;2(2):171-175. doi:10.5812/nms.10709PubMedGoogle ScholarCrossref
    63.
    de Jong  MR, Van der Elst  M, Hartholt  KA.  Drug-related falls in older patients: implicated drugs, consequences, and possible prevention strategies.  Ther Adv Drug Saf. 2013;4(4):147-154. doi:10.1177/2042098613486829PubMedGoogle ScholarCrossref
    64.
    Dhalwani  NN, Fahami  R, Sathanapally  H, Seidu  S, Davies  MJ, Khunti  K.  Association between polypharmacy and falls in older adults: a longitudinal study from England.  BMJ Open. 2017;7(10):e016358.doi:10.1136/bmjopen-2017-016358PubMedGoogle Scholar
    65.
    Machado-Duque  ME, Castaño-Montoya  JP, Medina-Morales  DA, Castro-Rodríguez  A, González-Montoya  A, Machado-Alba  JE.  Association between the use of benzodiazepines and opioids with the risk of falls and hip fractures in older adults.  Int Psychogeriatr. 2018;30(7):941-946. doi:10.1017/S1041610217002745PubMedGoogle ScholarCrossref
    66.
    Aspinall  SL, Springer  SP, Zhao  X,  et al.  Central nervous system medication burden and risk of recurrent serious falls and hip fractures in Veterans Affairs nursing home residents.  J Am Geriatr Soc. 2019;67(1):74-80. doi:10.1111/jgs.15603PubMedGoogle ScholarCrossref
    67.
    Hanlon  JT, Zhao  X, Naples  JG,  et al.  Central nervous system medication burden and serious falls in older nursing home residents.  J Am Geriatr Soc. 2017;65(6):1183-1189. doi:10.1111/jgs.14759PubMedGoogle ScholarCrossref
    68.
    King  T, Vardanyan  A, Majuta  L,  et al.  Morphine treatment accelerates sarcoma-induced bone pain, bone loss, and spontaneous fracture in a murine model of bone cancer.  Pain. 2007;132(1-2):154-168. doi:10.1016/j.pain.2007.06.026PubMedGoogle ScholarCrossref
    69.
    Daniell  HW.  Opioid osteoporosis.  Arch Intern Med. 2004;164(3):338. author reply. doi:10.1001/archinte.164.3.338-aPubMedGoogle ScholarCrossref
    70.
    Guo  Z, Wills  P, Viitanen  M, Fastbom  J, Winblad  B.  Cognitive impairment, drug use, and the risk of hip fracture in persons over 75 years old: a community-based prospective study.  Am J Epidemiol. 1998;148(9):887-892. doi:10.1093/oxfordjournals.aje.a009714PubMedGoogle ScholarCrossref
    71.
    Janas  A, Folwarczna  J.  Opioid receptor agonists may favorably affect bone mechanical properties in rats with estrogen deficiency-induced osteoporosis.  Naunyn Schmiedebergs Arch Pharmacol. 2017;390(2):175-185. doi:10.1007/s00210-016-1295-6PubMedGoogle ScholarCrossref
    72.
    Endo  Y, Aharonoff  GB, Zuckerman  JD, Egol  KA, Koval  KJ.  Gender differences in patients with hip fracture: a greater risk of morbidity and mortality in men.  J Orthop Trauma. 2005;19(1):29-35. doi:10.1097/00005131-200501000-00006PubMedGoogle ScholarCrossref
    73.
    Panula  J, Pihlajamäki  H, Mattila  VM,  et al.  Mortality and cause of death in hip fracture patients aged 65 or older: a population-based study.  BMC Musculoskelet Disord. 2011;12:105. doi:10.1186/1471-2474-12-105PubMedGoogle ScholarCrossref
    74.
    Haentjens  P, Magaziner  J, Colón-Emeric  CS,  et al.  Meta-analysis: excess mortality after hip fracture among older women and men.  Ann Intern Med. 2010;152(6):380-390. doi:10.7326/0003-4819-152-6-201003160-00008PubMedGoogle ScholarCrossref
    75.
    Berry  SD, Zhu  Y, Choi  H, Kiel  DP, Zhang  Y.  Diuretic initiation and the acute risk of hip fracture.  Osteoporos Int. 2013;24(2):689-695. doi:10.1007/s00198-012-2053-3PubMedGoogle ScholarCrossref
    76.
    Tudor  CG. Transition to Part D coverage of benzodiazepines and barbiturates beginning in 2013. https://www.cms.gov/Medicare/Prescription-Drug-Coverage/PrescriptionDrugCovContra/Downloads/BenzoandBarbituratesin2013.pdf. Accessed October 2, 2019.
    77.
    Zhu  ZN, Jiang  YF, Ding  T.  Risk of fracture with thiazolidinediones: an updated meta-analysis of randomized clinical trials.  Bone. 2014;68:115-123. doi:10.1016/j.bone.2014.08.010PubMedGoogle ScholarCrossref
    78.
    Shen  C, Chen  F, Zhang  Y, Guo  Y, Ding  M.  Association between use of antiepileptic drugs and fracture risk: a systematic review and meta-analysis.  Bone. 2014;64:246-253. doi:10.1016/j.bone.2014.04.018PubMedGoogle ScholarCrossref
    79.
    Lee  SH, Hsu  WT, Lai  CC,  et al.  Use of antipsychotics increases the risk of fracture: a systematic review and meta-analysis.  Osteoporos Int. 2017;28(4):1167-1178. doi:10.1007/s00198-016-3881-3PubMedGoogle ScholarCrossref
    80.
    Oderda  LH, Young  JR, Asche  CV, Pepper  GA.  Psychotropic-related hip fractures: meta-analysis of first-generation and second-generation antidepressant and antipsychotic drugs.  Ann Pharmacother. 2012;46(7-8):917-928. doi:10.1345/aph.1Q589PubMedGoogle ScholarCrossref
    81.
    Rabenda  V, Nicolet  D, Beaudart  C, Bruyère  O, Reginster  JY.  Relationship between use of antidepressants and risk of fractures: a meta-analysis.  Osteoporos Int. 2013;24(1):121-137. doi:10.1007/s00198-012-2015-9PubMedGoogle ScholarCrossref
    82.
    Torstensson  M, Hansen  AH, Leth-Møller  K,  et al.  Danish register-based study on the association between specific cardiovascular drugs and fragility fractures.  BMJ Open. 2015;5(12):e009522. doi:10.1136/bmjopen-2015-009522PubMedGoogle Scholar
    83.
    Li  L, Setoguchi  S, Cabral  H, Jick  S.  Opioid use for noncancer pain and risk of fracture in adults: a nested case-control study using the general practice research database.  Am J Epidemiol. 2013;178(4):559-569. doi:10.1093/aje/kwt013PubMedGoogle ScholarCrossref
    84.
    Yang  YX, Lewis  JD, Epstein  S, Metz  DC.  Long-term proton pump inhibitor therapy and risk of hip fracture.  JAMA. 2006;296(24):2947-2953. doi:10.1001/jama.296.24.2947PubMedGoogle ScholarCrossref
    85.
    Ryu  JY, Kim  HU, Lee  SY.  Deep learning improves prediction of drug-drug and drug-food interactions.  Proc Natl Acad Sci U S A. 2018;115(18):E4304-E4311. doi:10.1073/pnas.1803294115PubMedGoogle ScholarCrossref
    ×