The study examined novel oral and intravenous chemotherapies, including vinorelbine (VNB), paclitaxel, docetaxel and gemcitabine, in the first-line treatment of non-small-cell lung cancer (NSCLC). Full details of the strategies evaluated were presented in the paper.

Treatment.

The National Institute for Health and Clinical Excellence (NICE) recommended the use of third-generation chemotherapy agents in the first-line treatment of advanced NSCLC in their clinical guideline on the diagnosis and treatment of lung cancer (February 2005). An economic model was prepared to determine the comparative costs of these agents, including a new oral VNB formulation, from the perspective of the UK National Health Service (NHS).

Cost-effectiveness analysis.

The population modelled comprised patients receiving a third-generation single agent chemotherapy regimen as first-line treatment in advanced NSCLC.

A Markov model was developed to estimate the cost per patient over a 52-week period. Thirteen therapeutic regimens using the four agents were identified and compared. Six mutually exclusive health states were defined using exclusively clinical criteria. These health states were induction, remission with dose reduction, remission without dose reduction, drop-out, progression and death. Weekly cycles were applied. Each health state was associated with a cost of treatment and dichotomous criterion of results. Transition probabilities between states were quantified using published literature.

The setting was tertiary care. The economic study was carried out in the UK.

The clinical data were derived from studies published between 1994 and 2001. The total costs were stated to be at 2004 levels.

Efficacy and toxicity data collected from the literature included:

drop-out rate;

median time to progression;

median duration of survival; dose reduction;

febrile neutropenia;

anaemia - thrombopenia;

diarrhoea - constipation; and

nausea - vomiting.

Nine publications, published between 1994 and 2001, supplied the clinical effectiveness data. The largest phase III trials identified for each agent were used as the primary source of data. Where necessary, this was augmented by smaller studies identified by the review.

The most commonly used administration schedules were established by a consensus group of 15 oncology experts, of varying age, background and experience, from different parts of the UK. A literature search from 1990 to 2004 was carried out using the MEDLINE, EMBASE, Pascal, DARE, NHS EED and HTA databases, to determine the clinical effectiveness of each single agent when used in NSCLC. Inclusion criteria were not reported.

The median duration of survival varied from 24 to 31 weeks. However, the authors stated that, in accordance with the NICE assessment, a "conservative" assumption of non-difference in therapeutic efficacy was made. Efficacy data from a VNB study were applied to all 13 regimens, while toxicity profiles for each agent remained differentiated. This, in essence, gave all regimens equal effectiveness.

The costs of drug acquisition, administration (including staff, equipment and tests, day admissions, outpatient visits, pre-therapy counselling) and hospital transport of patients were driven by the regimens and assumptions chosen. The costs of toxicity management were obtained from four publications and converted to UK 2003 prices. The cost of oral VNB, not yet available in the UK, was estimated from its price in other European countries. A pre-therapy counselling cost was included for self-administration at home. The costs of managing neurotoxicity and local reactions were not included. Transfer payments and direct non-medical costs were excluded from the analysis.

Productivity costs were excluded from the scope of the analysis.

UK pounds sterling ().

The costs were treated deterministically.

Two other options for schedules and different administration locations were investigated in the sensitivity analysis.

The reader is referred to the statement under 'Measure of Benefits Used in the Economic Analysis' above, and the relevant NICE Clinical Guideline.

The publication abstract reported annual follow-up costs of 3,746 with intravenous VNB, 5,332 with gemcitabine, 5,977 with paclitaxel and (6,766 with docetaxel. Oral VNB, with outpatient administration on day 1 and self-administration at home on other days, had a cost per patient per year of 2,888.

These results represented selected regimens from the full 13 and were the lowest costs for all agents, except gemcitabine, where another regimen was associated with annual costs of 5,081 per patient.

Costs related to administration and toxicity management were reported in addition to the total costs. Oral VNB was associated with cost-savings per patient, ranging from 2,193 to 2,444 in comparison with gemcitabine and from 3,088 to 4,009 in comparison with taxanes.

The sensitivity analysis showed that, whatever the location of administration, oral VNB remained the least costly option. This was also true when the costs of treating adverse events were multiplied by a factor up to 4.3. For an equivalent cost between gemcitabine and oral VNB, the cost of toxicities had to be multiplied by a factor of 4.4.

No synthesis was performed.

Oral vinorelbine (VNB) allowed further savings in hospital resources in comparison with other third-generation agents.

The comparators chosen represented all third-generation chemotherapy agents licensed for use as monotherapy in the first-line treatment of NSCLC in the UK, and which had recently been assessed by NICE. You should decide whether all relevant comparators for your own setting have been covered.

The model structure was described with the aid of a simplified graphical  representation.  The sources of the model parameters and the methods  used to calculate weekly transition probabilities were described, but  the values themselves were not given.  Uncertainty in the administration  schedules was investigated by modelling several different regimens with  the same agent, although the descriptions of these regimens and those  used in the sensitivity analyses were not easy to follow.  The results  of the alternative schedules used in the sensitivity analyses were  reported fully.  Only one other aspect of uncertainty was investigated:  a threshold analysis with respect to adverse event costs to find the  point at which gemcitabine and oral VNB costs became equal.

The parameters were derived from published research. A search was conducted in appropriate databases, although the search terms and inclusion criteria were not reported. The authors did not justify why only one clinical trial was selected for each agent, when others existed, instead of synthesising the evidence. It could be inferred that the trials chosen were randomised controlled trials, but this was not explicitly stated in the publication. A more thorough reporting of the studies selected and included in the review would have enhanced the validity of the estimates.

A summary measure of benefit was not derived. The authors stated that therapeutic efficacy could be "conservatively" assumed to be equivalent across all regimens, citing only the 2001 NICE assessment in justification. However, differences in toxicity profiles were accepted within the model and these helped to drive differences in costs. It was not made clear how the primary studies could have been powered to show differences in side effects when they were assumed not to show differences in primary outcomes, because the authors did not explain this approach. It was also unclear why this should be considered a conservative approach.

It was difficult to understand why patients would receive hospital transport (i.e. whether associated with chemotherapy administration or the treatment of adverse events) and the source of the resource use data was not given. No justification was given for excluding the costs of neurotoxicity and local reactions from adverse event treatment costs, which might be expected to be high. It was difficult to be sure that the costs had been inflated and standardised appropriately. The sources of the costs and prices appeared to date from 2004 (drug costs), 2003 (stated price year, after conversion, for adverse events) and 2002/2003/2004 (administration and patient transport). The total costs were then stated to be 2004 prices, although conversion or standardisation was reported only for adverse event costs and these only to 2003 levels. The unit costs were not consistently reported. It is possible that omissions might have had an impact on the results, although drug costs usually form the main cost-driver in this indication.

The authors described a search for published economic evaluations and compared the studies found with their own research. They stated that no other UK-based evaluation had been found but that their findings were aligned with other publications. The study was designed to focus on the UK NHS and should therefore be relevant to users of this database. The conclusions reflected the scope of the analysis. The authors reported no limitations to their study.

The authors suggested that, in light of the 2001 NICE guidance, which concluded that there was no significant difference in efficacy between any of the third-generation agents, the question faced by the NHS is to find the most cost-minimising interventions to implement the UK NHS Plan for Cancer. According to this study, oral VNB represents the most cost-minimising therapeutic option.

Supported by Pierre Fabre Medicament.

Because readers are likely to encounter and assess individual publications, NHS EED abstracts reflect the original publication as it is written, as a stand-alone paper. Where NHS EED abstractors are able to identify positively that a publication is significantly linked to or informed by other publications, these will be referenced in the text of the abstract and their bibliographic details recorded here for information.

This record was compiled by CRD commissioned reviewers according to a set of guidelines developed in collaboration with a group of leading health economists.