Sensitivity analyses

One-way sensitivity analyses were performed to identify the influence of input parameters on the outcomes. Sensitivity and specificity of DEXA for diagnosing osteoporosis, osteoporosis prevalence rate, annual fracture risks, treatment efficacy and HSUVs were varied by ±20% of the values used in the base-case analysis. Medication persistence and probability of being highly adherent, probability of individuals residing in a nursing home after a hip fracture, inpatient costs for hip, vertebral and wrist fractures, annual medication costs and costs of screening were varied by ±50% of the values used in the base-case analysis [54]. Additionally, we performed one-way sensitivity analyses by assuming different discount rates on cost and

effectiveness, treatment duration, no offset time effect after medication discontinuation and different screening initiation age. As some variables were defined by a mean and standard error, distributions were applied in the model. Probabilistic sensitivity analysis was performed, in which sampling of distributions of input parameters was performed to address the uncertainties around multiple parameters simultaneously [54]. Cost-effectiveness acceptability curves were generated to visually illustrate the probabilities of screening and appropriate treatment being cost-effective.

4.5 Results

4.5.1 Face validity and internal validation

The model was designed and constructed by both experienced clinicians (Andrew Palmer, Tania Winzenberg) and health economics experts (Lei Si, Andrew Palmer). From a clinical perspective, the model structure was determined to correctly represent all important clinical facets of osteoporosis screening and fractures [10].

We performed a total of 27 internal validations by comparing model predictions of age-specific hip, clinical vertebral and wrist fracture incidence rates against those data used in creating our model. The results generated by the model closely match the published data from which the input probabilities were derived: the regression line slope was 0.996, which was close to 1.00, and the R2 was 0.99 which indicated that the model faithfully reproduced the published data. The collective results for the internal validation are shown in Appendix 4B.2.

Table 4.3 summarizes the costs, effectiveness and the incremental cost-effectiveness ratio (ICER) of screening with DEXA versus no screening. The mean costs for screening and no screening were $1,939 and $1,619 respectively for the base-case analysis, the respective mean QALYs were 9.9442 and 9.722. The cost per QALY gained for screening versus no screening was $1,440 in the base-case analysis.

4.5.2 Sensitivity analyses

Without discounting for costs and effectiveness, the cost per QALY gained decreased to $931. The ICER increased to $1,844 when costs and effectiveness were discounted by 5% annually. The accuracy of the screening test also impacted on outcomes: lower test sensitivity and specificity yielded higher costs but lower effectiveness for the screening strategy while the costs and effectiveness of no screening remained unchanged.

Medication persistence and adherence both impacted on the ICER: costs per QALY gained decreased with lower and increased with higher medication persistence and adherence. Costs and effectiveness did not change in the no screening group because no alendronate treatment was assumed for individuals in the no screening group. Effectiveness in the screening arm changed only slightly compared to that in the base-case analysis, whereas average costs changed more substantially especially with varied medication persistence.

Costs of fracture and screening did not impact on ICER significantly: costs per QALY gained were all higher than the WTP threshold and close to that in the base-case analysis. However, annual medication cost had a dramatic impact on the cost-effectiveness of DEXA screening: with a 50% decrease of annual medication cost, the DEXA screening was cost-saving compared with no screening.

Costs per QALY gained were $3,347, if screening was initiated at the age of 60 DEXA screening was cost-saving if the screening was initiated at the age of 70 years. The cost- effectiveness acceptability curves (CEAC) were provided showing the probabilities of screening being cost-effective given a continuous WTP threshold (Figure 4.3). Given the WTP of $20,000 per QALY gained, screening initiated from age 65years had a probability of 99% of being cost-effective.

Table 4.3 Summary of costs, effectiveness, ICER of DEXA screening versus no screening strategy: base-case and one-way sensitivity analyses

Parameters Costs a _{Effectiveness} b ICER DEXA No screening DEXA No screening Base-case 1,939 1,619 9.944 9.722 1,440

One-way sensitivity analyses

Discount rates: 0% 2,922 2,545 12.972 12.567 931 Discount rates: 5% 1,542 1,252 8.584 8.427 1,844 1.2 times base case annual fracture risks 2,136 1,905 9.881 9.615 870 0.8 times base case annual fracture risks 1,724 1,320 10.006 9.832 2,314 1.2 times base case treatment efficacy 1,915 1,619 9.950 9.722 1,290 0.8 times base case treatment efficacy 1,963 1,619 9.937 9.722 1,598 Treatment duration: 2 years 1,818 1,619 9.937 9.722 925 Treatment duration: 10 years 1,967 1,619 9.945 9.722 1,566 No treatment offset time effect c _{1,981 1,619 9.930 9.722 1,736}

0.8 times base case DEXA sensitivity 2,403 1,619 9.887 9.722 4,751 0.8 times base case DEXA specificity 4,524 1,619 9.918 9.722 14,795 1.5 times base case medication persistence d _{2,179 1,619 9.948 9.722 2,472}

0.5 times base case medication persistence d _{1,689 1,619 9.934 9.722 328}

1.5 times base case medication adherence e _{1,961 1,619 9.947 9.722 1,514}

0.5 times base case medication adherence e _{1,915 1,619 9.939 9.722 1,357}

1.5 times base case probability of nursing home 1,972 1,670 9.942 9.719 1,355 0.5 times base case probability of nursing home 1,906 1,570 9.946 9.724 1,522 1.5 times base case fracture inpatient costs 2,401 2,377 9.944 9.722 112 0.5 times base case fracture inpatient costs 1,476 862 9.944 9.722 2,768 1.5 times base case annual medication costs 2,314 1,619 9.944 9.722 3,133 0.5 times base case annual medication costs 1,563 1,619 9.944 9.722 cost-saving 1.5 times base case screening cost 2,039 1,619 9.944 9.722 1,892 0.5 times base case screening cost 1,839 1,619 9.944 9.722 988 1.5 times base case nursing home annual cost 1,969 1,672 9.944 9.722 1,343 0.5 times base case nursing home annual cost 1,908 1,567 9.944 9.722 1,537 1.2 times base case HSUVs 1,939 1,619 11.954 11.720 1,365 0.8 times base case HSUVs 1,939 1,619 7.941 7.744 1,625 Screening population aged 60 years 2,245 1,590 11.706 11.510 3,347 Screening population aged 70 years 1,603 1,637 8.186 7.932 cost-saving

ICER = incremental cost-effectiveness ratio, DEXA = dual-energy x-ray absorptiometry, HSUV = Health-state utility value

a _{Costs are lifetime direct costs and presented in 2013 US dollars.} b _{Effectiveness is presented in quality-adjusted life year (QALY).}

c _{Medication offset time effect refers to the residual effect on fracture risks after the discontinuation of treatment.} d _{Medication adherence remains unchanged.}