See addendum 1 and addendum 2 of the AG report.105
Evidence from published trials
Several approaches can be taken to estimate of the proportion of patients treated with docetaxel monotherapy who will experience one or more episodes of grade 3 or 4 FN as a result of treatment. A total of eight different estimated incidence rates were identified as follows.
Assessment Group base case (TAILOR)41
Four patients in TAILOR41 were reported to have experienced grade 3 or 4 FN in the docetaxel arm, all of whom were in the subgroup of 63 patients treated every 3 weeks with high-dose docetaxel (75mg/m2 body surface area). This corresponds to an incidence rate of 6.35% (95% CI 1.79% to 13.50%) and relates to the dose and frequency of docetaxel administration most commonly used in the UK.
Decision Support Unit Report94
During the first appraisal of erlotinib versus docetaxel in second-line chemotherapy for NSCLC (NICE TA16221) the Decision Support Unit was asked to investigate the incidence of FN and its associated treatment costs. They conducted a meta-analysis of reported trials and estimated the incidence as 5.95% (95% CI 5.30% to 7.70%).
TAILOR trial41 (all patients)
If no distinction is made between high dose (3-weekly) and low dose (weekly docetaxel 35mg/m2 body surface area), the FN incidence rate is 3.85% (95% CI 1.07% to 8.28%).
Other trials (pre-epidermal growth factor testing)
Data from 17 randomised clinical trials,34,37,41,42,49,64,74,85,106–115 which included 3-weekly high-dose docetaxel monotherapy as one treatment arm, were combined to provide a weighted average incidence rate. It was not possible to carry out a formal meta-analysis because of the diversity of comparators, populations and settings of these trials. The weighted average estimate is 7.3% (95% CI 6.3% to 8.3%). Heterogeneity testing of trial incidence values identified that two of the larger trials34,49 exhibited significantly higher incidence rates than the remaining 15 trials. Therefore, two weighted average values were selected for sensitivity testing, 10.8% (95% CI 8.9% to 12.8%) and 5.0% (95% CI 4.0% to 6.2%), corresponding to these distinct data subsets. The maximum estimated incidence among all 17 trials, 12.7% (95% CI 9.0% to 16.8%) was also selected for exemplification in the decision model.
Extreme sensitivity analysis
In order to explore the impact of a very high incidence rate, the value of the greatest upper confidence limit of any of these 17 trial arms was selected – 25%.
Comment on Royal College of Physicians suggested incidence rates
In the Royal College of Physicians submission document to NICE it is stated that:
‘In clinical practice, admission rates for neutropaenic sepsis and treatment complications are 25–50% with docetaxel compared to < 5% with erlotinib’.116
Unfortunately no supporting evidence was cited for this statement. Subsequently, the Royal College of Physicians responded to the Appraisal Consultation Document citing a conference abstract by Sharma117 of an observational study of admissions in three NHS trusts to support a figure of 41%. The abstract shows that 41% is the total number of hospital admissions in second-line docetaxel treatment (9 out of 22 admissions), whereas only four of these were a result of neutropenic sepsis (i.e. 18%). In addition, it should be noted that admission rates are necessarily higher than incidence rates, as the Decision Support Unit estimated that affected patients require an average of 1.4 admissions per patient. Using this factor to adjust admission rates to incidence rates, the best estimate from the Sharma117 study is an incidence rate of 13.0% (95% CI 2.7% to 29.5%). The small numbers involved and the wide CI (which encompasses all of the eight estimates listed above) indicates that these data add nothing useful to the consideration of FN incidence rates.
Sensitivity analysis for febrile neutropenia incidence
summarises the cost-effectiveness results for the AG-revised base case and the seven alternative FN scenarios described previously. In all cases erlotinib is not cost-effective compared with docetaxel, because the cost and utility effect of varying FN incidence is not sufficient to counteract the estimated survival advantage of docetaxel. The incremental cost is zero for a FN rate of 16.2% (equal cost, but QALY gain for docetaxel). The ICER for erlotinib versus docetaxel only exceeds £30,000 cost savings per QALY lost for docetaxel FN incidence rates above 63%.
Sensitivity analysis of AG-revised base-case scenario, with alternative assumed values of the incidence rate of grade 3 and 4 FN during second-line docetaxel 3-weekly monotherapy
provides an overview of the three estimated AG base-case ICERs made available to the AC during this appraisal.
Estimated base-case cost-effectiveness estimates of erlotinib vs. docetaxel for the EGFR M– population provided by the AG during the appraisal
