Canadian CE of Dabrafenib in Metastatic Melanoma

Canadian CE of Dabrafenib in Metastatic Melanoma

Journal: PharmacoEconomics

Cost Effectiveness of Dabrafenib as a First-Line Treatment in Patients with BRAF V600 Mutation-Positive Unresectable or Metastatic Melanoma in Canada

Thomas E. Delea, MSIA;1 Jordan Amdahl, BS;1 Alice Wang, MA;1 Mayur M. Amonkar, PhD;2 Marroon Thabane, PhD3

1Policy Analysis Inc. (PAI), Four Davis Court, Brookline, MA 02445, USA

2GlaxoSmithKline, 1250 South Collegeville Road, Collegeville, PA 19426, USA

3GlaxoSmithKline, 7333 Mississauga Rd N, Mississauga, Ontario,L5N 6L4, Canada

Address correspondence to:

Thomas E. Delea, MSIA

Policy Analysis Inc. (PAI)

Four Davis Court

Brookline, MA 02445, USA

Phone: 617-232-4400

Fax: 617-232-1155

E-mail:

1

Canadian CE of Dabrafenib in Metastatic Melanoma

Electronic Supplementary Material

Inputs and Distributions for the Indirect Treatment Comparison

Hazard ratios (HRs) for progression-free survival (PFS) and overall survival (OS) for vemurafenib versus dabrafenib were obtained using the Bucher method [1, 2] as shown in Table S1.The HR for PFS for vemurafenib versus dabrafenib was estimated to be 1.03 (95 % confidence interval [CI], 0.64–1.65; p=0.91; Table S1).The HR for OS for vemurafenib versus dabrafenib was estimated to be 1.16 (95 % CI, 0.43–3.14; p= 0.765)

Table S1 Adjusted ITC of the HR for PFS and OS for vemurafenib versus dabrafenib

CI Confidence interval, HR hazard ratio, ITC indirect treatment comparison, SE standard error

a The point estimates from this ITC are the same as those from the ITC conducted by Srivastava et al. [3]. The SEs and CIs differ due to the use of the ‘corrected’ CI for the rank-preserving structural failure time-adjusted HR for OS from BRIM-3

Note that, in conducting this ITC, the variance of the ln(HR) for OS from BRIM-3 was not based on the 95% CI reported by Roche in their submission to theNational Institute for Clinical Excellence [3]. Assuming that the ln(HR) for OS is distributed normally, the 95% CI on the HR for the rank-preserving structural failure time (RPSFT) analysis of OS reported by Roche (0.53–0.78) [4] implies a standard error (SE) on the ln(HR) of 0.117, which is virtually identical to that implied by the 95% CI on the HR for the intent-to-treat (ITT) analysis of OS (0.118). Any differences between these two values are likely not a consequence of rounding.As the RPSFT method maintains the p-value of the ITT analysis, if the measure of treatment effect increases with the RPSFT analysis (as it does for the RPFST analysis of OS from BRIM-3, from a 24% reduction to a 36% reduction in hazards of death), then the SE on the treatment effect must, by definition, increase. Accordingly, the 95% confidence half width on the RPSFT HR and the corresponding SE on the ln(HR)are likely underestimated (the methods by which Roche obtained the 95% CI on the RPSFT HR were not reported). To obtain a corrected SE on the ln(HR), we solved for the SE that would protect the p-value implied by 95% CIs for the HR for OS for the ITT analysis in BRIM-3 (HR, 0.76; 95% CI, 0.63–0.93; implied p-value [based on assumption of normality of ln(HR)]=0.00572). This yielded a corrected 95% CI on the RPSFT-adjusted HR for OS of 0.47 to 0.88 (versus 0.53 to 0.78 originally).

It should also be noted that, in the model, the HRs for vemurafenib versus dabrafenib during the trial period are calculated as the ratio of the input values for the HRs for vemurafenib versus dacarbazine and for dabrafenib versus dacarbazine. For the projection period, the model applies the HR of vemurafenib versus dacarbazine directly to the dacarbazine projections (the same approach is used for OS as for PFS).This is mathematically equivalent to applying the estimated HRs for vemurafenib versus dabrafenib obtained using the Bucher method as described above. This approach is used in the model to ensure proper application of the HRs to the referent distributions for PFS and OS in the trial and projection periods. The HRs and corresponding SEs from the indirect treatment comparisons (ITCs) are reported for illustration only.

A Note on the Proportional Hazards Assumptions

It should be noted that OS for dabrafenib during the projection period and PFS and OS for vemurafenib in the trial and projection periods were obtained by applying HRs to underlying log-logistic distributions. The log-logistic model is an accelerated failure time model, and the survival distribution arising from the application of a constant HR to a log-logistic distribution is not a log-logistic distribution. However, the log-logistic was used as the underlying distribution in the model based on goodness of fit to the empirical distribution, not on theoretical assumption regarding the nature of the effect of treatment on the survival distribution (i.e., proportional hazards versus proportional survival time). Furthermore, the survival distribution for dabrafenib was obtained by applying to the distribution for dacarbazine as a time-varying HR. Just as the distribution arising from the application of a constant HR to an underlying log-logistic distribution is not a log-logistic distribution, the survival distribution arising from the application of a time-varying HR to an underlying Weibull distribution (a ‘proportional hazards distribution’) is not a Weibull distribution. Accordingly, this does not represent a fundamental limitation of our approach any more than the application of HRs to the Kaplan–Meier distributions (with unknown parametric forms) is a fundamental limitation.

Furthermore, as noted above, the HR for OS for dabrafenib was varied over time. Hence, while the instantaneous hazard for dabrafenib is proportional to that of dacarbazine based on the assumed linear function describing the increasing HR over time, it is not a constant proportional hazards model per se (e.g., as is the Cox model). We do not believe this is an important limitation of the approach.

Additional Information on CostEstimates

Dosages

The assumed dosage schedules for each treatment regimen are described in Table S2. Details of each regimen include: type of administration (e.g., intravenous [IV], oral formulation), duration of infusion (if applicable), basis of daily dosage (per patient, per m2 of body surface area [BSA], per kg of body weight), days of use per cycle, length of cycle (i.e., in days), and maximum number of cycles (if applicable). Dosages for dabrafenib and dacarbazine were assumed to be the same as those employed in BREAK- 3 [5]. The dosage for vemurafenib was based on BRIM-3 [6].

Table S2Characteristics of treatment regimens for therapies considered in the model

IV Intravenous

Body Surface Area

Mean BSA and weight were calculated using patient data from BREAK-3. For each patient, BSA was calculated from height and weight using a widely used formula devised by Du Bois and Du Bois [7]. The Du Bois and Du Bois formula was derived based on the measurement of two-dimensional surface area of plaster of Paris molds of nine subjects. The age, sex, and nutrition of these subjects, studied in the middle of the First World War, are unlikely to be comparable to that of cancer patients today. Several other formulas have since been proposed, but none of those have gained widespread acceptance. The DuBois formula is shown below:

weight (kg) 0.4256 × height (cm) 0.7256 × 0.007184 [7].

The estimated mean BSA and weight are shown in Table S3.

Table S3Estimated mean BSA and weight from BREAK-3

BSA Body surface area

Unit Cost of Medications

Medication cost per day was calculated as the product of unit cost (per mg) and dosage (mg) per day. The cost per milligram for each drug was obtained by dividing the cost per unit (e.g., tablet or vial) by the milligram per unit. The unit cost of dacarbazine was based on population-weighted averages of list prices across provinces from the IMS Brogan database [8, 9]. Province-specific pricing information for each drug is presented in Table S4. Formulary prices were used when available; otherwise, wholesale prices were used. Formulary prices were available in Alberta, British Columbia, Manitoba, Ontario, and Saskatchewan for dacarbazine and in Newfoundland and Labrador, Ontario, and Québec for temozolomide. When multiple formulations of dacarbazine were available in a province, the formulation with the lowest cost per milligram was used. If pricing information for dacarbazine was unavailable in a province, the estimated unit cost for Ontario was used instead.

Medication Pricing Information for Each Canadian Province

Table S4Pricing and formulation information for medications in each province

CA Canadian, IV intravenous

The price of vemurafenib used in the model (CA$46.54 per 240-mg tablet) was the list price referenced in the pan-Canadian Oncology Drug Review’s (pCODR) final recommendation for vemurafenib [10]. The price of vemurafenib in the IMS Brogan database was CA$46.54 per 240-mg tablet across all provinces. The price of dabrafenib used in the model (CA$63.33 per 75-mg capsule) was provided by GlaxoSmithKline (Table S5).

Table S5 Estimated medication costs of dabrafenib, vemurafenib, and dacarbazine used in the model

CA Canadian, pCODR pan-Canadian Oncology Drug Review

aList price

b List price referenced in pCODR’s final recommendation for vemurafenib

c Average per milligram price over multiple formulation sizes across all provinces

d Cost per 28 days provided for comparison. Dacarbazine is administered once every 21-day cycle. Cost per 288 days for dacarbazine calculated by multiplying 21-day cost (CA$334.36) by 28/21

Dispensing and Administration Costs

The administration cost of oral medications was assumed to be immaterial and was not considered. The cost per IV administration was calculated as the sum of preparation costs, nursing and support costs, physician consultation costs, and chemotherapy chair time costs. Costs of preparation and nursing/support were from Leighl et al. [11]. Leighl and colleagues evaluated direct medical costs of docetaxel in patients with non–small-cell lung cancer from the view of the Canadian public healthcare system. In the study, the fixed cost relating to IV chemotherapy administration was estimated to be CA$13.34 (based on 1999 prices), which also included the cost of pharmacy preparation per dose. Physician chemotherapy cost was based on the cost of time spent by a physician for supervising IV infusion for treatment of malignant or autoimmune disease [12]. Cost of chair time for IV therapies was calculated as the product of hourly chair time [13] and number of hours of infusion. All costs were adjusted to 2012 prices using the Bank of Canada Inflation Calculator. Calculations of total administration costs for IV therapies are presented in Table S6.

Table S6Administration cost of IV therapy

CA Canadian, IV intravenous

Dispensing cost of oral medications was estimated based on the dispensing fee payable to most pharmacies under the Ontario Drug Benefit Program (CA$8.40 per prescription filled) [14]. Dispensing costs for IV therapy (i.e., dacarbazine) were assumed to be included in administration costs.

Costs of Treatment for Adverse Events

Costs of treatment for each adverse event (AE) were calculated by multiplying the incidence of treatment-related AEs by the expected cost of these events. Estimates of the costs of treatment of each AE were estimated by combining resource use estimates from a survey of 14 Canadian clinicians with Canadian-specific unit cost estimates. The survey included several questions about the services (i.e., outpatient visits, emergency department visits, medications, hospitalizations, or other services or procedures) required to manage patients who experience palmar-plantar erythrodysesthesia (PPE), photosensitivity, pyrexia, squamous cell carcinoma (SCC), and neutropenia while being treated with a BRAF inhibitor or dacarbazine.

Costs of treatment of AEs (per event) are assumed to be independent of treatment strategy. The costs of treating each grade 3 or 4 AE are shown in Table S7. The cost of treating photosensitivity was assigned to rash based on the assumption that the resources used to treat grade 3 or 4 rash would be similar to those used to treat photosensitivity (i.e., outpatient visits and creams). Given that keratoacanthoma is considered a form of SCC, the costs of treating grade 3 or 4 keratoacanthoma were assumed to be the same as those for treating SCC. A detailed breakdown of the component costs used to estimate the cost per each AE is listed in Table S8–S12.

Table S7Direct medical costs of treatment of grade ≥3 AEs

AE Adverse event,CA Canadian, PPE palmar-plantar erythrodysesthesia, SCC squamous cell carcinoma

Table S8Calculation of the costs of managing PPE

AE Adverse event, CA Canadian,N/A not available, PPE palmar-plantar erythrodysesthesia

aEstimates of the utilization of services and the units of service received (i.e., number of clinic visits, length of hospital stay, etc.) based on survey of Canadian physicians; medications and other services were specified by survey respondents

Table S9Calculation of the costs of managing photosensitivity

AE Adverse event, BID twice daily, CA Canadian,CBC complete blood count, N/A not available, NSAID non-steroidal anti-inflammatory drug

aEstimates of the utilization of services and the units of service received (i.e., number of clinic visits, length of hospital stay, etc.) based on survey of Canadian physicians; medications and other services were specified by survey respondents

Table S10Calculation of the costs of managing pyrexia

AE Adverse event,CA Canadian, CBC complete blood count, Cr creatinine, LDH lactate dehydrogenase, LFT liver function test, NSAIDs non-steroidal anti-inflammatory drugs

aEstimates of the utilization of services and the units of service received (i.e., number of clinic visits, length of hospital stay, etc.) based on survey of Canadian physicians; medications and other services were specified by survey respondents