Comment and Reply on: A comparison of the estimated costs of erlotinib, docetaxel and pemetrexed for the second-line treatment of non-small cell lung cancer from the German healthcare perspective

Dear Editor,

With the introduction of new agents, outcomes for patients with advanced non-small cell lung cancer (NSCLC) have improved and clinicians now have several effective regimens from which to choose appropriate therapy for their patients. It was therefore with interest that we read the cost comparison of erlotinib, docetaxel and pemetrexed for the second-line treatment of NSCLC that was conducted by Kotowa and colleagues1. Cost differentials between treatment options are an important factor for consideration when selecting a therapeutic strategy, but their determination must be based on sound principles. We have identified a number of methodological issues that we believe may have seriously impacted the validity of the aforementioned analysis.

The authors used a cost-minimisation approach to conclude that erlotinib offers cost advantages over docetaxel and pemetrexed in the second-line treatment of NSCLC. Cost-minimisation analyses should only be used to perform cost comparisons between two or more treatments if the outcomes of those treatments are equivalent2. In most instances, it is considered sufficient that a major efficacy endpoint be equivalent between the treatments and, where possible, data are obtained from Phase III randomised trials directly comparing the treatments. No such direct comparison exists for the treatments in question. Kotowa and colleagues therefore obtained clinical data from two randomised Phase III trials, one directly comparing pemetrexed with docetaxel and the other erlotinib with placebo3,4. There was no treatment common to both trials reducing the authors’ and readers’ ability to assess the appropriateness of the comparison. In addition, the patient populations of these trials differed in a number of respects that should preclude such a direct comparison. To be considered for entry into the erlotinib trial, patients were required to have received one or two regimens of combination chemotherapy and not be eligible for further chemotherapy, whereas patients enrolled in the pemetrexed versus docetaxel trial had received treatment with only one prior chemotherapy regimen for advanced disease and were not eligible for curative therapy. Notably, 50% of patients enrolled in the erlotinib trial had received two prior NSCLC chemotherapy regimens compared with no patients treated with either pemetrexed or docetaxel, and only 45% of patients enrolled in the erlotinib trial had stage IV NSCLC compared with 75% of pemetrexed- and docetaxel-treated patients3,5. Although the number of prior regimens did not appear to impact treatment efficacy6, stage III NSCLC was associated with increased survival in pemetrexed- and docetaxel-treated patients3, a tendency that was also seen with erlotinib5. This difference in the proportions of patients in each disease stage could therefore have influenced the relative efficacy results for erlotinib, pemetrexed and docetaxel.

With regard to the efficacy endpoint used to determine equivalence of the study therapies, Kotowa and colleagues selected median survival of patients with performance status 0/1. However, survival in this subgroup of patients was not the primary endpoint of either of the two trials from which clinical data were obtained for the cost analysis, and nor was this patient group used to obtain or derive any other data used in the cost analysis. If efficacy data are considered for the patient populations on which resource utilisation for surveillance, diagnostics and treatment were stated to be based, there does appear to be a difference in efficacy between the three treatments. The primary efficacy endpoint of the two clinical trials, overall survival, favoured pemetrexed, being 6.7 months for erlotinib-, 8.3 months for pemetrexed- and 7.9 months for docetaxel-treated patients. Therefore cost minimisation was not an appropriate analysis for these data.

The determination of costs for each therapy also requires scrutiny. The duration of treatment assumed for erlotinib (84 days) is low compared with the treatment duration reported for the pivotal trial from which resource utilisation data were stated to be derived. In that trial, 52% of patients received treatment for 84 days, 29% of patients received treatment for 168 days, 16% of patients received treatment for 252 days and 10% of patients received treatment for 336 days4, with an estimated mean duration of treatment of 125 days7. This is longer than the median treatment duration used in the pemetrexed clinical trial3 and the recommended treatment duration for docetaxel8 (4 cycles of 21 days; 84 days). Thus, the allocated treatment duration for erlotinib is possibly low compared with actual use and is definitely low compared with the efficacy data presented, presenting an exaggeratedly low cost for erlotinib. If a mean duration of 125 days of therapy is assumed for erlotinib, and the cost of erlotinib is €87.44 per 150 mg tablet/day, as we calculated was used by Kotowa and colleagues (assuming no co-medications), then the mean cost of treatment with erlotinib would actually be €10,930, a 49% increase over the cost calculated by Kotowa and colleagues. If the cost of erlotinib is assumed to be €83.98 per 150 mg tablet/day (taken from the German National drug list (Ifap 2007)), then the mean cost of treatment with erlotinib would be €10,498 (a 43% increase).

In comparison, some costs associated with laboratory tests for patients receiving pemetrexed seem to have been inflated. Although it may be correct to assume fewer laboratory tests are required for erlotinib than the comparators, frequent monitoring of blood counts should be conducted for patients receiving both pemetrexed and docetaxel. There should therefore be no major differences between these two agents in this cost element, yet laboratory costs are assumed to be more than two times as high with pemetrexed as with docetaxel. We also question the choice of 4-weekly laboratory testing, since pemetrexed and docetaxel are both administered using a 3-weekly cycle and erlotinib is administered continuously on a daily basis.

Finally, hospitalisation costs for docetaxel and pemetrexed were based on data obtained from the clinical trial3, whereas no such data were available for erlotinib. Erlotinib data were therefore assumed to be within the designated normal range for the appropriate DRG codes. Sensitivity analyses do not appear to have been conducted to assess the effect of this assumption on overall cost. The only sensitivity analyses conducted involved varying the cost of managing haematological adverse effects, with no further details offered. We suggest that Kotowa and colleagues could have also considered the impact of changes in the duration of treatment with erlotinib, particularly as treatment cost was the largest cost component considered, and the impact of laboratory tests.

Declaration on interest: JA and CL are both employees of Eli Lilly and Company.