Teriparatide (TPTD) is indicated for the treatment of postmenopausal women with osteoporosis (OP) at high risk for fracture.1 High risk for fracture is defined as having a history of osteoporotic fractures, multiple fracture risks, and failed or nontolerated OP therapies. A better understanding of treatment selection and proper case-mix adjustment for studies conducting comparative analyses can be achieved by examining for whom TPTD is prescribed and the course of care leading to TPTD initiation in a real-world setting.

Previous research established that patients who initiate TPTD are different from those who initiate bisphosphonates (BPs), in terms of their age and baseline health status.2 However, no previous studies have assessed the sequence of care and healthcare resource utilization leading up to TPTD initiation, or have attempted to find an appropriate match for TPTD patients in a real-world setting. This study addresses that gap and defines TPTD patients by providing real-world evidence of patient characteristics, sequence of OP care, and healthcare resource utilization, as well as by comparing patients initiating TPTD treatment versus non-TPTD treated patients.

METHODS
The study was a retrospective observational analysis of administrative medical and pharmaceutical insurance claims data available in the Truven Health MarketScan Commercial Claims and Encounters and the Medicare Supplemental and Coordination of Benefits databases (Truven Health Analytics, USA). The MarketScan databases provide de-identified patient-level data. The databases have a large sample size (170 million covered lives, 300 contributing employers, 25 contributing health plans)3 within the United States that enables investigation of rare events, such as fragility fractures. The databases contain a longitudinal history of patient-specific data, including demographic information (sex, age, dates of plan enrollment), clinical encounters (inpatient and outpatient services), and outpatient pharmaceutical prescription dispensing (retail and mail order). The data available for this study were for the period from January 1, 2004, to December 31, 2011.

Women 50 years or older initiating TPTD treatment with a requirement of continuous enrollment throughout a study period of 36 months prior to treatment initiation were eligible for study inclusion to the TPTD cohort. Given the objective of the study, 36 months of history was selected as the time frame for the evaluation, although historical data were available from 12 months to 60 months. Thirty-six months provided more complete information on patients’ clinical characteristics and sequence of care with an acceptable sample size versus 12 months or 24 months of history. Even though the patient sample size was larger for both 12 months and 24 months of continuous enrollment prior to the index date (Figure 1), the 36-month history provided a higher rate of relevant clinical event information for patients. Several examples include: an observed bone mineral density (BMD) test (16% more at 36 months than at 12 months), use of BP therapy (14% more at 36 months than at 12 months), and hospitalizations (16% more at 36 months than at 12 months). Longer history times of 48 months or 60 months included substantially fewer total patients and did not provide significantly more event information on the BMD test, use of BP therapy, or hospitalizations.

To avoid including a fracture risk not related to OP, patients were excluded who had a record of either Paget’s disease diagnosis (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 731.0), Paget’s disease treatment (risedronate 30 mg, alendronate 40 mg, etidronate 200/400 mg), malignant neoplasm (ICD-9-CM codes 140-208), radiation oncology (Current Procedural Terminology [CPT] codes 77261- 77799), or chemotherapy (CPT codes 96400-96549, or V-codes: 58.0, 58.1, 07.3x) at any time during the study period. The TPTD patient selection is presented in Figure 1.

Patient demographic and clinical characteristics, healthcare resource utilization (HRU), and timing of healthcare-related events—such as OP diagnosis, BMD test, OP therapy, and comorbidities—were assessed and described for the TPTD cohort overall, and stratified by previous OP treatment exposure during the 36 months prior to the index date. The stratification by prior use of OP therapies was in a hierarchical order as follows: 1) oral BP: any prior oral BP use with switching to other OP therapies allowed; 2) intravenous (IV) BP: no oral BP use with other OP therapies use allowed; 3) other OP treatments (raloxifene, nasal calcitonin, or oral hormone replacement therapy [HRT]) that were not oral or IV BP; or 4) no prior OP treatment. All of these treatment groups were mutually exclusive.

The non-TPTD cohort was identified by selecting women 50 years or older in the database who had either a record of a prescription claim for BP, raloxifene, nasal calcitonin, or HRT, or a record of an OP diagnosis (ICD-9-CM 733). The initial prescription or OP diagnosis was set as the index date for this cohort. Patients in the non-TPTD cohort were also required to have at least 36 months of continuous enrollment after their index date (Figure 1). The same TPTD patient exclusion criteria were applied to non-TPTD patients.

Patient demographic characteristics included age, geographic region (northeast, north central, south, west, or unknown), and insurance type (commercial or Medicare). Patient clinical characteristics included health status measured by the Charlson Comorbidity Index4 (CCI) and the Chronic Disease Score (CDS)5; fracture risk factors history of fragility fracture, rheumatoid arthritis (RA), chronic obstructive pulmonary disease (COPD), OP, use of oral glucocorticoid (GC) medications at least 5 mg/day for at least 90 consecutive days, and BMD test (ICD-9-CM and CPT codes provided in eAppendix Table A, available at www.ajmc.com).

Incident fragility fractures were defined as nontraumatic closed hip, femur, pelvic, forearm, wrist, humerus, tibia, fibula, and spine fractures verified by either inpatient overnight stay or at least 2 outpatient visits within 90 days of each other. Incident hip fractures were included for patients with a confirmed inpatient stay only. Trauma fractures were identified by E codes (E800-E848, E916-E919, E928.8-E928.9, E929.0-E929.1, E953, E957-959, E960-979, E988-E989, E999) within 7 days before and after a fracture claim. HRU was measured by hospitalizations, outpatient visits, and concomitant medication use.

Oral BP is the first-line therapy for fracture prevention and the treatment of OP. However, the therapy is known to aggravate or cause gastrointestinal (GI) problems in patients.6,7 Patients with potential GI intolerability to oral BP were identified in treatment cohorts by a history of relevant upper- and lower-tract GI events (ICD-9-CM provided in eAppendix Table B) or use of GI medications (ie, H2 antagonists, proton pump inhibitors, and cytoprotectives) after beginning of oral BP treatment.

Statistical Analysis
Patient demographic and clinical characteristics between 2 cohorts were compared using χ² tests for categorical variables and t tests for continuous variables. For the attempt to find an appropriate match for TPTD patients, a propensity-score matching model was applied. A propensity score was assigned to each patient as the probability of being treated with TPTD given the patient’s characteristics, and was calculated using logistic regression modeling. Further, using derived propensity scores, TPTD and non-TPTD cohorts were matched 1:1.8,9 In addition to matching on the propensity score, patients were matched exactly on the following variables: treatment subgroup, age (range of 5 years), OP diagnosis, BMD test, fracture sites (hip, pelvis, femur, and spine), RA, COPD, oral GC use, previous OP therapy, back disorders, and GI event. These variables were added to the matching algorithm to further improve balance between the matched groups.

SAS version 9.2 (Cary, North Carolina) was utilized for all statistical analyses. Statistical significance was defined at P <.05.

RESULTS
Patient characteristics, HRU and sequence of OP-related care among TPTD patients

The TPTD cohort consisted of 5882 women who initiated TPTD treatment between January 1, 2007, and December 31, 2011. These patients were a mean age of 69.1 years (SD = 10.9), with a mean CCI of 1.19 (SD = 1.5) (Table 1). Of these patients, 91.1% (n = 5356) had a record of OP diagnosis, 83.7% (n = 4921) had a record of a BMD test, and 33.6% (n = 1977) were observed with at least 1 previous fragility fracture. The most frequently observed fracture site was spine (n = 1145, 19.5%), followed by wrist (n = 323, 5.5%), hip (n = 256, 4.4%), pelvis (n = 228, 3.9%), femur (n = 195, 3.3%), tibia/fibula (n = 192, 3.3%), humerus (n = 190, 3.2%), forearm (n = 147, 2.5%), and clavicle (n = 27, 0.5%). Among patients observed with at least 1 fragility fracture (n = 1977), 25.7% (n = 507) had a record of 2 or more, and 8.09% (n = 160) had a record of 3 or more fractures at unique fracture sites during the study period. A diagnosis of back disorder, including back pain, was observed in 37.8% (n = 2226) of TPTD patients. Overall, 38.9% (n = 2289) of TPTD patients had an all-cause overnight hospital stay and a mean of 61.9 (SD = 47.3) outpatient visits per patient during the study period (36 months prior to TPTD treatment).

A majority of TPTD patients had previous use of either oral BP (n = 3837, 65.2%), IV BP (n = 88, 1.5%), or other OP treatments (n = 856, 14.6%). However, 18.7% of all TPTD patients did not have any OP treatment during the 36 months prior to TPTD therapy. There were statistically significant differences in patients characteristics observed between these treatment subgroups (ie, oral BP, other OP treatments, and no OP treatment prior to TPTD initiation) (Table 1). Patients with prior exclusive use of IV BP were not included in the comparison with other subgroups due to the small sample size (n = 88).

TPTD patients with prior use of an oral BP had the highest proportion of recorded OP diagnosis (92.5%) or BMD testing (85.9%), the lowest average number of comorbidities (CCI mean, 1.13, SD = 1.43), and the lowest rate of all-cause hospitalization (36.9%) compared with the other 2 groups. TPTD patients with prior use of other OP treatments (ie, raloxifene, nasal calcitonin, or oral HRT) had the highest rates of cardiovascular (CV) disease (78.2%), back disorder (44.2%), and GI (79.4%) diagnoses. In addition, patients with prior use of other OP treatments had the greatest number of all-cause outpatient visits per patient (mean visits = 68.4, SD = 51.8) and the highest concomitant medication use per patient (mean prescriptions = 18.4, SD = 8.4) of the 3 groups during the 36-month study period. TPTD patients with no prior OP treatment exposure had the highest proportion of fragility fractures (36.1%) and fracture-related hospitalizations (9.8%), as well as the lowest rates of concomitant medication use (mean = 15.0, SD = 8.3), GC use (4.5%), and BMD testing (76.8% ) during the 36 months prior to TPTD initiation.

Among TPTD patients with prior use of oral BP, 58.7% (n = 2254) were exclusive users of the therapy; the rest alternated among other OP treatments. The first observed oral BP therapy prescription occurred a median of 34 months prior to TPTD initiation (Figure 2a). A median of 14 days after the first observed oral BP prescription, 67.3% (n = 2584) experienced a diagnosed GI event (Table 1). During the study period, 33.2% (n = 1274) of TPTD patients with prior use of oral BP experienced at least 1 fragility fracture, and the most recent fracture occurred a median of 7 months prior to TPTD initiation. Among patients with a fragility fracture, 76.2% (n = 971) had a prior GI condition. The median time on OP therapy after starting oral BP treatment and before initiating TPTD was 23 months, with a median gap between therapies of 24 days.

The majority of TPTD patients with prior other OP treatments, 79.4% (n = 680) had a record of a preexisting GI condition. During the study period, 31.3% (n = 268) of patients with other OP treatments experienced at least 1 fragility fracture, and the most recent fracture occurred a median of 7 months prior to TPTD initiation (Figure 2b). The median time on OP therapy after starting other OP treatments and before initiating TPTD was 13 months, with a median gap between therapies of 26 days. For TPTD patients with no history of OP treatments prior to TPTD initiation, 67.5% (n = 743) had a preexisting GI condition and 36.1% (n = 398) experienced at least 1 fragility fracture, and the most recent fracture occurred a median of 5 months prior to TPTD therapy initiation (Figure 2c).

Comparison With the Non-TPTD Cohort
There were 284,701 non-TPTD patients identified in the database with either a prescription claim for a BP, raloxifene, nasal calcitonin, or HRT, or a recorded OP diagnosis. Compared with the non-TPTD cohort, TPTD patients were statistically significantly different on most observed patient characteristics during the 36-month study period (Table 2): they were significantly older (69.1 years vs 61.9 years) and had more comorbidities (CCI, 1.2 vs 0.8) when compared with non-TPTD patients. A higher proportion of TPTD patients experienced a fragility fracture (33.6% vs 6.7%), back disorders (37.8% vs 15.6%), RA (9.7% vs 3.7%), COPD (24.1% vs 13.6%), CV disease (75.1% vs 68.5%), thyroid disease (28.2% vs 24.2%), a recorded OP diagnosis (91.1% vs 63.9%), BMD test (83.7% vs 57.4%), all-cause and fracture-related hospitalizations (38.9% vs 21%, and 7.6% vs 0.9%, respectively), and oral GC use (6.7% vs 1.7%).

OR and 95% CIs from a logistic regression are presented in Table 2. Patients who initiated TPTD treatment were more likely to have an OP diagnosis (OR, 5.763; 95% CI, 5.267-6.304), a BMD test (OR, 3.793; 95% CI, 3.539-4.066), prior oral BP therapy (OR, 3.395; 95% CI, 3.173-3.632), and prior fragility fracture (OR, 6.995; 95% CI, 6.614-7.398) than non-TPTD patients. Patients with RA, COPD, back disorders, oral GC medication, fracture-related hospitalization, and GI events during oral BP therapy were also more likely to be TPTD initiators (Table 2).

Propensity score analysis allowed for 1:10 matching of the TPTD cohort to non-TPTD patients, and identified a total of 41,759 (14.7%) non-TPTD patients who fit the TPTD patient profile. For comparative analysis and to obtain a closer match, subsequent 1:1 matching was used that resulted in 5143 (87.4% of the TPTD cohort) matched cohorts. The matched TPTD and non-TPTD cohorts were not statistically different on a vast majority of patient characteristics (data not shown); the only difference being the proportion of patients with fracture-related hospitalizations (4.5% vs 3.5%; P = .009, TPTD vs non-TPTD, respectively), fracture-related outpatient visits (1.88 vs 1.60; P = .004, TPTD vs non-TPTD, respectively), and concomitant medication use per patients (15.78 vs 15.33; P = .004, TPTD vs non-TPTD, respectively).

DISCUSSION
Using retrospective analysis of an administrative insurance claims database, this study described 36-month historical patient characteristics, sequence of OP care, and healthcare resources of patients prior to their treatment with TPTD and compared with patients who did not receive TPTD. The vast majority of TPTD patients were observed to be treated with either oral BP or other OP treatments before initiating TPTD therapy, with the first prescription of the previous therapies occurring a median of 34 months prior to TPTD initiation. Even though the TPTD cohort differed by patient characteristics when stratified by OP treatment prior to TPTD, all patients who initiated TPTD treatment were older and had more comorbidities compared with the non-TPTD cohort.

Significant differences were found between TPTD and non-TPTD cohorts on many patient characteristics and sequence of care in the 3 years prior to TPTD initiation. These differences should be taken into account in a comparative analysis as a source of bias. The study also demonstrated well-balanced matched TPTD and non-TPTD cohorts with a 1:1 propensity score matching approach. The matching included propensity scores and direct matching on additional variables, which might be the greatest influence on selection into the treatment. Nevertheless, the cohorts remained significantly different on fracture-related hospitalizations and outpatient visits, as well as concomitant medication use. These variables further indicated higher severity of OP for the TPTD patient cohort, and should be included in an adjusted model of a comparative analysis of TPTD versus non-TPTD patients.

In addition, the TPTD cohort had a high proportion of patients with GI events among previous oral BP users. A substantial number of these patients were observed with a diagnosed GI event as early as 14 days after the initiation of the oral BP. Patients with GI events might be less tolerant of oral BP therapy, which might in turn lead to treatment noncompliance or discontinuation. Previous publications have speculated that reduced treatment efficacy might be related to GI intolerability of oral BPs.6,7 On the other hand, a history of prior GI events might lead to a refusal of oral BP therapy initiation altogether. In this study, we observed a majority of patients with either other OP or no previous OP treatments had preexisting GI conditions diagnosed a median 34 and 31 months, respectively, prior to TPTD initiation. This might have precluded the use of oral BP for these patients. Meanwhile, patients either failed their other OP treatments and/or experienced a fragility fracture that eventually led to TPTD therapy.

Limitations
This study has limitations that are common to the nature of administrative claims databases, specifically the nature of the MarketScan population. These limitations include an inability to confirm diagnostic and procedural codes that might lead to a misclassification bias; however, this is not expected to vary between the study groups. Furthermore, unobserved variables (eg, BMD test results, body mass index) could introduce unmeasured confounding and treatment selection bias. Inclusion of only commercially and Medicare supplemental-insured individuals prevent generalizability of the findings to the uninsured (16% of total US population) and individually insured (5%) populations.10 Despite these limitations, MarketScan databases are among the most published proprietary databases in the United States, used by government, university, and private-sector researchers.11

CONCLUSION
TPTD patients significantly differ in their pretreatment clinical and demographic characteristics from non-TPTD patients, which should be taken into account in comparative analysis. A retrospective analysis of a national healthcare claims database demonstrates 81.3% of patients initiating TPTD treatment had prior use of either BP or other OP treatments, and a majority of these patients failed or had intolerance for their prior OP therapies. In addition, a majority of patients with no previous OP therapy prior to TPTD initiation had preexisting GI conditions that might have precluded use of an oral BP. Patients with clinical characteristics appropriate for TPTD treatment, as indicated by the TPTD prescribing information,9 should be considered for treatment with TPTD, especially patients with preexisting GI conditions who might not fully benefit from or refuse to initiate oral BP and patients who might experience an adverse GI event while on oral BP therapy.

Acknowledgments
The authors acknowledge Jennifer Meyer Harris, PharmD, an employee of Eli Lilly and Company, for editorial assistance in preparation of the manuscript.