Otros genes presentes en los fósmidos.

IV methods have not yet been applied to the study of osteoporosis medications

The safety concerns of BPs3, 4 _{combined with the introduction of new OP medications}5 _{in recent} years have led to increasing interest in evaluation of the risk-benefit profile of OP medications among clinicians, patients, the pharmaceutical industry, and regulatory agencies. Few head-to-head clinical trials comparing the relative effectiveness of these drugs have been conducted6-10 _{and only a small number of} observational studies have directly compared different OP drugs in their effect on the risk of Fx or other outcomes.11-14 _{A major limitation of these studies is that important potential confounders, such as bone} mineral density, BMI, physical activity, family history of OP and frailty, were not available for analyses. IV methods, which may control for these unmeasured confounder in observational studies, have yet to be applied to the study of comparative effectiveness of OP medications. Furthermore, the physician preference IV has not been evaluated in this context.

97

Innovative model to predict prescribing physician may be useful in many research areas

The predicted probabilities from our predictive model could be applied to databases such as MarketScan, which has a provider ID variable in medical service claims but no physician ID in pharmacy claims, to predict the most likely prescribing physician. Although “crosswalks” relating one ID to another may become available for databases in the future, until they are, trying to identify the prescriber remains a challenge. Using the linkage method we have proposed may allow researchers to better track physician prescribing and to more fully utilize the administrative claims databases that are currently available.

D. Limitations

IV approach underpowered in this context

Using the two-stage IV framework, we calculated the differences in risk between IV groups and differences in risk between IV groups divided by IV strength. All 95% CIs for both difference statistics included 0, were relatively wide, and many overlapped. Because of this and if the IV assumptions hold, we are unable to rule out a clinically substantial difference between the BP and other OP groups. Also, the analysis suggests that the IV method may be underpowered for this study question and therefore not very informative.

IV may not be an improvement over standard OLS methods for this study question

For the IV to have less asymptotic bias compared to OLS regression, the absolute PDR of unmeasured confounders, defined as the prevalence difference between IV groups divided by the prevalence differences between treatment groups, should be less than the strength of the instrument (Brookhart IJB 2007). For our physician preference IVs, this threshold is 12 to 17%. Furthermore, most

98

prevalence differences were close to 0 and fluctuations in the PDR could be due to these small

prevalence differences. Besides measuring PDR, we conducted a statistical analysis to test whether the IV was superior to OLS methods (Choi B 2015 unpublished). We failed to reject the null hypothesis of the IV being equivalent or worse than OLS methods in our study. Therefore, the IV may not have been an improvement but was likely not worse than standard OLS methods.

Physician drug preference is not easily quantified

In general, the preference-based IV assumes that physicians or groups of providers differ in their preferences for medical treatments or procedures for similar patients.15 _{Furthermore, differences in} hospital capacity, drug benefit plans or formularies may lead to regional differences in medical decision making among physician groups.16 _{The IV assumptions are not violated as long as the factors}

contributing to physician prescribing decisions are not related to patients’ characteristics. For OP medications, a physician’s prescribing preferences may be influenced by patient preference which may depend on affordability,17 _{potential side effects,}18, 19 _{convenience of dosing schedule,}20 _{and route of} administration.21 _{To the extent that these physician preferences are associated with potential} confounding variables, the validity of the IV may be compromised.

E. Future directions

Future research is needed to further explore the findings of this dissertation. Osteoporotic Fx is a rare outcome and if the sample size was larger or the IV was stronger, the analysis may have resulted in more informative findings. From this research, it is apparent that comprehensive evaluation of different definitions of an IV is essential to identifying the most appropriate characterization that has the maximal

99

strength, best suits the effect estimate (e.g., risk ratio or risk difference) used in the study and satisfies other assumptions of the IV.

With respect to the current physician preference IV, it would be informative to apply it to settings where greater imbalances across treatment groups were observed to more comprehensively evaluate the performance of the current IV, for instance. Additional studies may build on this research by stratifying analyses by physician specialty (i.e., orthopedic surgeons) where preferences may vary more and by supplementing the current IV with medical claims data. Also, the study may be repeated in databases where confounders which are unmeasured in the current study are available.

The database linkage method we have proposed may allow researchers to better track physician prescribing and to more fully utilize the administrative claims databases that are currently available. However, the accuracy of this technique needs further evaluation. The regression models may be refined by adding additional predictors, for example. Also, the method could be evaluated in a setting where the equivalence of two different physician IDs in different parts of a database is known.

F. Conclusions

IV methods may adjust for unmeasured confounding in observational studies however the IV in our current study was underpowered. If our sample size was larger and the IV was stronger, our analysis may have generated more informative results. The database linkage method we have developed is innovative in that it may help predict the prescribing physician in databases where physician IDs may be unavailable or where there is no “crosswalk” between different physician IDs. Additional work (e.g.,

100

evaluate methods in data sources where the equivalence of different physician identifiers is known) is needed to further test and develop this linkage method.

101 REFERENCES

1. Brookhart MA, Schneeweiss S. Preference-based instrumental variable methods for the estimation of treatment effects: assessing validity and interpreting results. The International Journal of

Biostatistics. 2007; 3(1): Art. 14.

2. Angrist JD, Krueger AB. Instrumental variables and the search for identification: from supply and demand to natural experiments. The Journal of Economic Perspectives. 2001; 1(4):69-85. 3. Fellows JL, Rindal DB, Barasch A, et al. ONJ in two dental practice-based research network

regions. J Dent Res. 2011;90:433-438.

4. Dell RM, Adams AL, Greene DF, Funahashi TT, Silverman SL, Eisemon EO, Zhou H, Burchette RJ, Ott SM. Incidence of atypical nontraumatic diaphyseal fractures of the femur. J Bone Miner Res. 2012;27:2544-50.

5. Rachner TD, Khosla S, Hofbauer L. Osteoporosis: now and the future. Lancet. 2011;377(9773):1276-1287.

6. Bonnick S, Saag KG, Kiel DP, McClung M, Hochberg M, Burnett SM, et al. Comparison of weekly treatment of postmenopausal osteoporosis with alendronate versus risedronate over two years. J Clin Endocrinol Metab. 2006; 91:2631-7.

7. Luckey M, Kagan R, Greenspan S, Bone H, Kiel RD, Simon J, et al. Once-weekly alendronate 70 mg and raloxifene 60 mg daily in the treatment of postmenopausal osteoporosis. Menopause. 2004; 11:405-15

8. Recker RR, Kendler D, Recknor CP, Rooney TW, Lewiecki EM, Utian WH, et al. Comparative effects of raloxifene and alendronate on fracture outcomes in postmenopausal women with low bone mass. Bone. 2007; 40: 843-51.

9. Rosen CJ, Hochberg MC, Bonnick SL, McClung M, Miller P, Broy S, et al. Fosamax Actonel Comparison Trial Investigators. Treatment with once-weekly alendronate 70 mg compared with once-weekly risedronate 35 mg in women with postmenopausal osteoporosis: a randomized double- blind study. J Bone Miner Res. 2005; 20:141-51.

10. Sambrook PN, Geusens P, Ribot C, Solimano JA, Ferrer-Barriendos J, Gaines K, et al. Alendronate produces greater effects than raloxifene on bone density and bone turnover in postmenopausal women with low bone density: results of EFFECT (Efficacy of FOSAMAX versus EVISTA Comparison Trial). International J Intern Med. 2004; 255:503-11.

11. Cadarette SM, Katz JN, Brookhart MA, Sturmer T, Stedman MR, Solomon DH. Relative

Effectiveness of Osteoporosis Drugs for Preventing Nonvertebral Fracture. Ann Intern Med. 2008; 148: 637-646.

102

12. MacLean C, Newberry S, Maglione M, McMahon M, Ranganath V, Suttorp M, Mojica W, Timmer M, Alexander A, McNamara M, Desai S, Zhou A, Chen S, Carter J, Tringale C, Valentine D, Johnsen B, Grossman J. Systematic Review: Comparative Effectiveness of Treatments to Prevent Fractures in Men and Women with Low Bone Density or Osteoporosis. Annals of Internal Medicine. 2008; 148: 197-213.

13. Silverman SL, Watts NB, Delmas PD, Lang JL, Lindsay R. Effectiveness of bisphosphonates on nonvertebral and hip fractures in the first year of therapy: the risedronate and alendronate (REAL) cohort study. Osteoporosis Int. 2007; 18:25-34.

14. Watts NB, Worley K, Solis A, Doyle J, Sheer R. Comparison of risedronate to alendronate and calcitonin for early reduction of nonvertebral fracture risk: results from a managed care administrative claims database. J Manag Care Pharm. 2004; 10:142-51.

15. Brookhart MJ, Rassen JA, Schneeweiss S. Instrumental variable methods in comparative safety and effectiveness research. Pharmacoepidemiology and Drug Safety. 2010; 19:537-54.

16. Cole JA, Norman H, Weatherby LB, Walker AM Drug copayment and adherence in chronic heart failure: effect on cost and outcomes. Pharmacotherapy. 2006; 26: 1157-1164.

17. McHorney CA, Schousboe JT, Cline RR, Weiss TW. The Impact of osteoporosis medication beliefs and side-effect experiences on non-adherence to oral bisphosphonates. Current Medical Research and Opinion. 2007; 23(12):3137-3152.

18. Papaioannou A, Kennedy CC, Dolovich L, Lau E, Adachi JD. Patient adherence to osteoporosis medications - problems, consequences and management strategies. Drugs Aging. 2007; 24(1): 37- 55.

19. Unson CG, Siccion E, Gaztambide J, Gaztambide S, Mahoney Trella P, Prestwood K.

Nonadherence and Osteoporosis Treatment Preferences of Older Women: A Qualitative Study. Journal of Women’s Health. 2003; 12(10):1037-45.

20. Reginster J, Rabenda V. Patient preference in the management of postmenopausal osteoporosis with bisphosphonates. Clinical Interventions in Aging. 2006; 1(4):4415-423.

21. Fraenkel L, Gulanski B, Wittink D. Patient treatment preference for osteoporosis. Arthritis Rheum. 2006 Oct 1; 55(5):729-735.

103 APPENDIX

Table A. Osteoporosis Diagnosis and Medication Codes and Case Algorithm

Osteoporosis Diagnosis ICD-9-CM Diagnosis Code

Osteoporosis 733.0

Unspecified osteoporosis 733.00

Senile osteoporosis (postmenopausal) 733.01

Idiopathic osteoporosis 733.02

Disuse osteoporosis 733.03

Other osteoporosis (drug-induced) 733.09

Case Algorithm for Osteoporosis Diagnosis:

Any of the above codes as a primary or secondary discharge diagnosis in an INPATIENT claim record, or one outpatient diagnosis associated with physician evaluation or management.

Osteoporosis Medication†

HCPCS J-code

(time period restriction, if any) Zoledronic acid infusion (Reclast) J3488 (from 01/01/2008)

Zoledronic acid infusion Q4095 (from 07/01/2007) Zoledronic Acid Infusion (Reclast)* J3490 (from 04/17/2007)*

104

Ibandronate injection J3490 (from 04/01/2006)* Ibandronate injection J1740 (from 01/01/2007) Ibandronate injection C9229 (from 07/01/2006)

Calcitonin injection J0630 (from 01/01/1982)

Teriparatide J3110 (from 01/01/2005)

Other Osteoporosis Medications† _Dosage

alendronate sodium 70 mg tablet

70 mg oral solution 10 mg tablet 5 mg tablet 35 mg tablet alendronate sodium/vitamin D3 tablet

(Fosamax plus D) 70 mg alendronate /2800 IR vitamin D3 70 mg alendronate/5600 vitamin D3 ibandronate sodium (Boniva) 2.5 mg tablet

150 mg tablet

risedronate 5 mg tablet

35 mg tablet 75 mg tablet 150 mg tablet

105

* An algorithm to distinguish between ibandronate and zoledronic acid, when coded as J3490 is needed based on the date of claimed code and other data system-specific data availability.

† The identification of injectable osteoporosis medication may be supplemented by NDC or other drug codes as appropriate based on the data availability in the data system.

risedronate sodium /calcium carbonate 35 mg risedronate /500 mg calcium

raloxifene (Evista) 60 mg tablet

calcitonin 200 IU/SPRAY

Case Algorithm:

106

Table B. Diagnosis Codes and Algorithm for Osteoporotic Fractures

Fracture Site ICD-9-CM Code*

Hip, closed Inpatient primary or secondary diagnosis codes (820.0, 820.2, 820.8, 733.14)

OR

Carrier line or outpatient claim with CPT in (27230-27248) and diagnosis code (820.0, 820.2, 820.8, 733.14)

Distal radius/ulna Inpatient primary or secondary diagnosis code in (813.4, 813.5, 733.12)

OR

Carrier line or outpatient claim with CPT in (25600, 25605, 25611, 25620, 25650, 25651, 25652 (includes ulnar styloid)) and diagnosis code in (813.4, 813.5, 733.12)

Spine, closed or pathologic Inpatient primary or secondary diagnosis code in (805.0, 805.2, 805.4, 805.8, 733.13)

OR

Carrier line or outpatient claim with CPT in (22520, 22521, 22522, 22523, 22524, 22525, 76012, 76013, 22305, 22310, 22315, 22318, 22319, 22325, 22326, 22327, 22328) and diagnosis code in (805.0, 805.2, 805.4, 805.8, 733.13) OR

107

Carrier line or outpatient claim with CPT in (physician

evaluation/management) and diagnosis in (805.0, 805.2, 805.4, 805.8, 733.13)

Pelvis-closed Inpatient primary or secondary diagnosis code (808.0, 808.2, 808.4, 808.8)

OR

Carrier line or outpatient claim with CPT in (27193-27194, 27215-27218, 27220, 27222, 27226-27228) and diagnosis code (808.0, 808.2, 808.4, 808.8)

Other femur-closed Inpatient primary or secondary diagnosis code (821.0, 821.2, 733.15)

OR

Carrier line or outpatient claim with CPT in (27500-27514) and diagnosis code (821.0, 821.2, 733.15)

Fracture Site ICD-9-CM Code*

Radius/ulna-other-closed Inpatient primary or secondary diagnosis code (813.0, 813.2, 813.8)

108

Carrier line or outpatient claim with CPT in (24650, 24655, 24665, 24666, 24670, 24675, 24685, 25500, 25505, 25515, 25520, 25525, 25526, 25530, 25535, 25545, 25560, 25565, 25574, 25575) and diagnosis code (813.0, 813.2, 813.8) Humerus-closed Inpatient primary or secondary diagnosis code (812.0, 812.2,

812.4, 733.11) OR

Carrier line or outpatient claim with CPT in (23600, 23605, 23615, 23616, 23620, 23625, 23630, 23665, 24500, 24505, 24515, 24516, 24530, 24535, 24538, 24545, 24546, 24560, 24565, 24566, 24575, 24576, 24577, 24579, 24582) and diagnosis code (812.0, 812.2, 812.4, 733.11)

Case Algorithm: One of the above codes (AND no concurrent major trauma ECODES) as a primary or secondary discharge diagnosis in an INPATIENT claim record, or one outpatient diagnosis associated with physician evaluation or management :

E800-E848 Railway, motor vehicle, other road vehicle accidents; water transport accidents; air and space transport accidents; other vehicle accidents E881-E884 Falls, other than falls on stairs and falls on same level

E908-E909 Accidents due to cataclysmic storms and earth surface movements E916-E928 Other accidents (struck by falling object, striking against or struck

accidentally by objects or persons, caught accidentally in or between objects, accidents caused by machinery, explosion, firearm, etc.)

* 4-digit substring unless otherwise specified† Drug codes (NDCs or HICL codes) for osteoporosis medications will be identified based on the corresponding generic or brand name as well as dosage form as indicated in the table.