B. Decision tree for budget scenario with drug E
8.3 Analyses for Uncertain Input Parameter Values
The second primary source of uncertainty is the uncertainty around values for parameters that are not known for certain by the modeler or the health plan. These include the efficacy and safety of current and new drugs as well as condition natural history, changing drug costs over the analysis time horizon, changing treatment shares over the analysis time horizon, and other future events over the analysis time horizon. These parameters will be uncertain to both the modeler and the health plan.
In particular, the following input parameters are estimated with uncertainty:
• Changes in treated incidence or prevalence over the analysis time horizon
• Efficacy and safety of current and new drugs in the treated population over the analysis time horizon
• Condition outcomes over the analysis time horizon with the current treatment mix and with the new treatment mix
Box 8.1 Budget-Impact Analysis for Alternative Scenarios for Expanding the Use of Bisphosphonates in Women with Low Bone Mineral Density: Cumulative Three-Year Budget Impact
(Tosteson et al. 2008)
Scenarios Current practice
All untreated receive
bisphosphonates Difference Scenario 1: All women aged
65–84 years with low bone mineral density
Total societal cost $25,957 million $31,520 million $5563 million Total number of fractures 1,118,670 728,621 −390,049 Scenario 2: Women aged
75+ years with low bone mineral density and previous fracture
Total societal cost $8154 million $8136 million −$18 million Total number of fractures 350,245 253,459 −96,786
In this analysis, two scenarios are presented: a scenario in which all women aged 65–84 years have low bone mineral density and a scenario in which women aged 75+ years have low bone mineral density and previous fracture.
Running these two scenarios helps the budget holder understand the budget implication of implementing policies using bisphosphonates to treat all eligi- ble women or just those over the age of 75 years with previous fracture.
133
• Treatment shares of the new drug over the analysis time horizon and redistribu- tion to the new drug from the current drugs
• Changes in the prices of current and new drugs over the analysis time horizon
• Impact of entry of other new branded or generic drugs during the analysis time horizon on all inputs
• Costs of providing other condition-related services and changes over the analysis time horizon
One way to present this uncertainty is to perform a one-way sensitivity analysis, varying each uncertain input parameter one at a time. This is the method recom- mended in the National Institute for Health and Care Excellence (NICE) guidance on estimating financial impact (NICE 2013) and in the updated International Society for Pharmacoeconomics and Outcomes Research (ISPOR) Budget Impact Task Force guidelines (Sullivan et al. 2014). To the extent possible, the ranges or alternative values used for these sensitivity analyses should be based on the observed variability in each data element, and the methods used to derive these ranges should be pro- vided. For example, the ranges used for efficacy could be the 95% confidence limits from the clinical trial data. However, for many of the uncertain variables, there are no observed data. Therefore, expert opinion on likely ranges may be needed for the one-way sensitivity analyses. If possible, ranges such as ± 20% for all input param- eters should not be used, because the feasible variability in input parameter values is likely to be different for different types of parameters. For example, the percentage range in costs might be much higher than the percentage range in treatment share estimates. For all alternative values tested in sensitivity or scenario analyses, it is necessary to provide the data source and rationale for the alternative values tested.
Scenario analyses or multiway sensitivity analyses may also be used to examine the impact of parameter uncertainty. This approach can be used when there is uncer- tainty around a set of values. For example, an alternative set of values might be plau- sible for examining changes in the treatment shares of all drugs in the treatment mix over the analysis time horizon. In this case, alternative scenarios might be tested based on inputs from budget holders or physicians. Scenario analyses or multiway sensitiv- ity analyses may also include variation in multiple types of parameters, including characteristics of the new drug such as its efficacy, safety, price, and dosing formula- tion, as well as the extent to which the current drugs are providing effective relief.
In presenting these one-way sensitivity analyses and scenario analyses, we typi- cally present the base-case analysis results using a single set of default values for both the inputs known to the budget holder and the inputs for which there is struc- tural or parameter uncertainty. We then also present the alternative sets of results by changing the uncertain input parameter values one at a time or by changing a group of parameter values. One-way sensitivity analyses and scenario analyses can also be presented for alternative base-case scenarios that include alternative feasible values for those inputs whose values are known to the budget holder.
In Box 8.2, we present an example of uncertainty analyses where the authors have presented the derivations for the data-driven ranges or alternative scenarios included in the analysis.
8 Uncertainty Analysis
Box 8.2 One-Way Sensitivity Analysis: Total Per-Patient Budget Impact of a Drug for Treatment of Chemotherapy-Induced Anemia (Rubin et al. 2008, Table 2)
Input parameter
DA Q3W total cost
EA QW total cost
EA QW – DA Q3W Drug price
Base case: AWP – 20% $8544 $8667 $123
AWP $10,606 $10,614 $8
ASP + 6% $6087 $7101 $1014
Mean dose per injection
Base case: DA Q3W 375.6 μg; EA QW 43,187 U $8544 $8667 $123 DA Q3W 283.8 μga; EA QW 43,187 U $6527 $8667 $2140 DA Q3W 467.2 μga; EA QW 43,187 U $10,555 $8667 −$1888 DA Q3W 375.6 μg; EA QW 38,044b U $8544 $7740 −$804 DA Q3W 375.6 μg; EA QW 46,307c U $8544 $9230 $686 Visit costs
Base case: $58.75 (CPT 99212 + CPT90772) $8544 $8667 $123 Visit cost + CBC panel ($58.75 + CPT 85027) $8589 $8803 $214 Visit cost for injection only $20.09 (CPT 90722) $8350 $8087 −$263 Frequency of administration visit
Base case (DA three times per week, EA once a week)
$8544 $8667 $123
DA weekly office visit; EA weekly visitd $8930 $8667 −$263 Time horizon
Base case: 16 weeks $8544 $8667 $123
12 weeks $5126 $6356 $1230
24 weeks $11,961 $13,290 $1329
Reprinted from Rubin et al. (2008) with permission from Taylor & Francis Ltd. http://www.
tandfonline.com
ASP average sales price, AWP average wholesale price, CBC complete blood count, CPT Current Procedural Terminology codes, DA darbepoetin alfa, EA epoetin alfa, Q3W every 3 weeks, QW every week
aThe minimum mean dose per infection for darbepoetin alfa Q3W was 283.8 μg Q3W based on an efficacy study of the Q3W 200 μg regimen (Taylor et al. 2005) (representing a 24.4% reduction in dose from the base case), and the maximum mean dose per injection was varied to 467.2 μg Q3W, representing a 24.4% increase from the base case (Canon et al. 2006)
bMean dose per infection was calculated based on values reported in Waltzman et al. (2005) using the following equation: mean weekly dose = mean cumulative dose/mean duration of treatment
cMean dose per injection was calculated based on values reported in Witzig et al. (2005).
A weighted average dose per injection was calculated based on dose (i.e., starting, esca- lated, and reduced dose), number of injections administered, and the number of patients on each regimen
dIt was assumed that during a 16-week duration of treatment, a patient on a QW regimen will receive 15 injections
135