Table 3 Outcomes for each scenario analysis

Base case deterministic results

0.891

0.774

$210,203

$235,924

$271,418

1. ALK+ population as proxy for ROS1+ NSCLC patients

This scenario will assume no difference between ALK+ and ROS1+ NSCLC patients and will utilize data from the PROFILE 1001 (Solomon 2014) study for PFS inputs.

1.075

0.897

$156,497

$145,642

$174,419

2. Second-best fitting curves for PFS

The second-best fitting curves were selected in this scenario for PFS for the combined analysis in each arm.

1.014

0.867

$200,571

$197,752

$231,354

3. Exponential curves for PFS

A constant risk was assumed based on an exponential distribution for PFS in this scenario for the combined analysis in each arm, to align with prior models for crizotinib.

1.131

0.952

$191,432

$169,269

$200,980

4. Individually fit curves

Rather than modelling using a common treatment parameter, the individually fitted survival curves were selected for each arm for this scenario.

1.461

1.231

$242,217

$165,825

$196,732

5. PROFILE 1001 (Shaw 2014) alone

Rather than using the combined analysis, the individually fitted curve from the PROFILE 1001 (Shaw 2014) study was used to inform efficacy inputs for the crizotinib arm.

1.242

1.063

$251,280

$202,360

$236,485

6. EUROS1 (Mazières 2015) data

Rather than using the combined analysis for the crizotinib and chemotherapy arms, individually fitted curves from the Mazières 2015 study were used to inform efficacy inputs for each treatment arm.

0.891

0.759

$160,777

$180,447

$211,837

7. No PFS difference

This scenario assumed no difference in PFS between the crizotinib and chemotherapy group. This is a conservative assumption as crizotinib appears to be associated with improved PFS outcomes. To achieve this the crizotinib PFS time-to-event values were substituted into the chemotherapy arm.

0.404

0.366

$121,524

$300,666

$331,738

8. No OS difference

To deal with uncertainty surrounding the clinical benefit of crizotinib, this scenario assumed no difference in overall survival between the two groups.

0.000

0.175

$106,390

NA

$606,640

9. No added maintenance benefit

Since uncertainty lies in whether the effect of maintenance is captured in the combined chemotherapy data, this scenario has removed the maintenance hazard ratio from both arms.

1.195

1.018

$216,744

$181,367

$212,829

10. Lower median PFS for second-line treatment (crizotinib arm)

Since there is uncertainty regarding the median PFS estimate for patients undergoing platinum-doublet chemotherapy after progressing on crizotinib, a shorter median PFS value was used from Smit et al. [37]

0.594

0.578

$206,007

$346,825

$356,140

11. Equal first-line utility (crizotinib values)

To test the uncertainty around utility value estimates, the utility value from the crizotinib arm was applied to both arms for first-line therapy.

0.891

0.714

$210,203

$235,924

$294,488

12. Equal first-line utility (chemotherapy values)

To test the uncertainty around utility value estimates, the utility value from the chemotherapy arm was applied to both arms for first-line therapy.

0.891

0.696

$210,203

$235,924

$302,179

13. First-line utility values from ALK+ population

To test the impact of using utility values from the PROFILE 1014 (Solomon 2014) trial for the chemotherapy arm (larger difference in utilities between groups)

0.891

0.864

$210,203

$235,924

$243,317

14. Proportion receiving active third-line therapy - 30%

To test uncertainty surrounding proportion of patient’s receiving active third-line therapy, a lower value (30%) was applied to the model.

0.864

0.755

$209,033

$241,917

$276,986

15. Best estimate of 3 above parameters (10, 13, 14) – pCODR Reanalysis

To test the impact of a combination of changes the above 3 parameter changes were applied together (Second line PFS of 4.2 months, ALK+ PROFILE 1014 (Solomon 2014) trial utilities and 30% receiving active third-line therapy).

0.584

0.659

$205,072

$351,140

$311,055