|
Gene expression profiling and breast cancer care: what are the potential benefits and policy implications? |
Oestreicher N, Ramsey S D, Linden H M, McCune J S, Van't Veer L J, Burke W, Veenstra D L |
|
|
Record Status This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn. Health technology The use of gene expression profiling (GEP) among premenopausal women newly diagnosed with Stage I or II breast cancer, to target adjuvant chemotherapy for those at high risk of distant recurrence, was examined. The study focused on the Netherlands Cancer Institute GEP assay.
Study population The hypothetical study population comprised premenopausal women newly diagnosed with Stage I or II breast cancer who had received treatment with mastectomy, or breast conserving surgery with radiation where indicated. The average age of the hypothetical population was 44 years. Fifty-one per cent had lymph-node positive disease, 77% had oestrogen receptor positive (ER+) disease, 47% had tumours larger than 2 cm, and 40% had high-grade tumours.
Setting The setting was primary care. The economic study was conducted in the USA.
Dates to which data relate The effectiveness data related to 2002. The cost use data related to 1991 to 2005. The price year was 2003.
Source of effectiveness data The effectiveness data were based on a review of published studies and on individual patient level data from the Netherlands Cancer Institute cohort that was used to validate the GEP assay.
Modelling A decision model was used to estimate the costs and quality-adjusted life-years (QALYs) associated with the use of GEP and NIH guidelines over a lifetime time horizon. The model consisted of two parts. There was a 6-month decision tree that modelled the prognostic categorisation and treatment of patients, and a Markov state-transition model with a cycle length of 1 year that extrapolated the QALYs and costs over the remaining cohort lifetime. The decision tree assigned patients into two categories, good prognosis (low risk of distant recurrence) or poor prognosis (high risk of distant recurrence) as identified by GEP or NIH guidelines. All women with poor prognosis were assumed to receive adjuvant chemotherapy. Patients entered the long-term Markov model in the health state "no evidence of disease". From this health state patients could progress to distant recurrence or death.
Outcomes assessed in the review The outcomes assessed included the proportion of patients that would be identified as high or low risk for distant recurrence by GEP or NIH guidelines, and the probability of distant recurrence in these groups, accounting for misidentification. These outcomes were based on an analysis of individual patient level data from the Netherlands Cancer Institute cohort. The other outcomes assessed were the relative risk for distant recurrence from adjuvant chemotherapy, mortality due to distant recurrence, mortality from other causes and the utility values for health states in the model.
Study designs and other criteria for inclusion in the review The review included a meta-analysis of adjuvant systemic therapy and studies published after 1990 that explicitly described their methods and provided numerical estimates. The authors did not provide any further details of the review, such as the sources searched to identify the primary studies.
Sources searched to identify primary studies Criteria used to ensure the validity of primary studies Methods used to judge relevance and validity, and for extracting data Number of primary studies included Approximately 16 primary studies were included in the review.
Methods of combining primary studies Where more than one study was used to inform a model parameter, the method of synthesis was not reported.
Investigation of differences between primary studies Results of the review The annual probability of distant recurrence among patients identified as low risk by GEP was estimated to be 0.018 up to year 10 and 0.009 from year 11 onwards.
The annual probability of distant recurrence among patients identified as high risk by GEP was estimated to be 0.069 up to year 10 and 0.007 from year 11 onwards.
The annual probability of distant recurrence among patients identified as low risk by NIH guidelines was estimated to be 0.020 up to year 10 and 0.000 from year 11 onwards.
The annual probability of distant recurrence among patients identified as high risk by NIH guidelines was estimated to be 0.047 up to year 10 and 0.008 from year 11 onwards.
Measure of benefits used in the economic analysis The measure of health benefit used was the QALYs. The utility values were derived from studies using standard gamble or time trade-off methods. The authors did not specify whether the preferences were obtained from patients or the general public. However, they did state that few community-based samples were available. The health states valued were 6 months after diagnosis without chemotherapy, 6 months after diagnosis with chemotherapy, no evidence of disease and distant recurrence.
Direct costs The study included the direct costs to the hospital and to the patients. The direct medical costs were derived from a review of the literature. No cost was assigned to the "no evidence of disease" health state. The cost of distant recurrence was calculated as a one-time cost that included all costs from recurrence to death. This method of costing a distant recurrence would affect the discounting, as all costs would be discounted in the year the recurrence occurs, rather than in the year the costs occur (although this would be the same in some cases). The incremental cost of adjuvant chemotherapy was calculated from reimbursement data. The resource prices were also based on reimbursement data for a managed care organisation.
The prices and the quantities were not reported separately, which may limit the generalisability of the study results. The direct patient costs were calculated by multiplying the number of hours spent travelling to treatment by the average hourly wage for women. Transportation costs were also calculated using Internal Revenue Service (IRS) reimbursement rates and parking charges. The costs were discounted at a rate of 3% per annum, and were reported in 2003 US dollars. The study focused on the incremental costs, and thus did not include the cost of primary surgery or the ongoing costs of breast cancer not related to distant recurrence.
Statistical analysis of costs Since authors did not have access to patient level cost data for the health state costs, a statistical analysis was not possible.
Indirect Costs The indirect costs were not included in the analysis. This was inappropriate given that the authors stated they were conducting the analysis from a societal perspective.
Sensitivity analysis Variability in the data was investigated by including all the model parameters in one-way sensitivity analyses, and also by conducting a full probabilistic analysis using Monte Carlo simulation. The uncertainty around probabilities and utilities was characterised using logistic normal distributions, while the uncertainty around cost parameters was characterised using a lognormal distribution. The source of the ranges tested was unclear.
Estimated benefits used in the economic analysis The use of GEP was estimated to result in 9.86 QALYs over the cohort lifetime, using a discount rate of 3% per annum. The use of NIH guidelines was estimated to result in 10.08 QALYs over the cohort lifetime, using a discount rate of 3% per annum. The side effects of adjuvant chemotherapy were included in the utility value estimated for that health state. However, the utility associated with being identified as 'high risk' or being denied the option of adjuvant chemotherapy was not included in the analysis. The use of GEP results in less patients being identified as high risk and less patients being offered adjuvant chemotherapy in comparison with NIH guidelines, and so these effects may be important.
Cost results The use of GEP was estimated to cost $29,754 over a patient's lifetime, using a discount rate of 3% per annum.
The same cost using NIH guidelines was estimated to be $32,636.
Productivity costs were not included in the analysis despite the authors stating that a societal perspective was used.
Compared with GEP, the use of NIH guidelines resulted in more women being treated with adjuvant chemotherapy, but also longer survival, and so the effect of omitting productivity costs is unclear.
Synthesis of costs and benefits The authors did not combine the costs and benefits. They stated that they did not calculate an incremental cost-effectiveness ratio (ICER) because the new technology was estimated to be less effective than using guidelines. However, as it was both less effective and less costly it would have been appropriate to calculate an ICER. The authors stated that the results were most sensitive to the test cut-off used to define women as high risk when using GEP.
Authors' conclusions The current evidence does not support the use of gene expression profiling (GEP) with its current levels of specificity and sensitivity.
CRD COMMENTARY - Selection of comparators The authors stated that NIH guidelines were not the only guidelines available for use in the study setting. Consequently, the analysis might not have contained all relevant comparators. The authors conducted an incremental analysis, so the results are only relevant if NIH guidelines are best current practice. You should decide whether these comparators are relevant in your own setting.
Validity of estimate of measure of effectiveness The effectiveness data were based on a review of the literature to inform the parameters of a decision model. The authors did not describe the methods of the review clearly, thus the included data could be subject to selection bias. A large number of the effectiveness parameters were based on a patient-level analysis of the study sample used to validate the GEP test. However, it was unclear how estimates from the primary studies were combined for other model parameters.
Validity of estimate of measure of benefit The estimation of lifetime QALYs was modelled. The use of a Markov model with utility values assigned to each health state was appropriate for such a calculation. However, it is worth noting that the authors might not have considered the decline in utility as the hypothetical cohort aged, which may overvalue the treatment associated with the greatest survival time.
Validity of estimate of costs The authors stated that a societal perspective was used, but they do not appear to have incorporated the indirect costs or justified this omission. The authors conducted an incremental analysis, and omitted the costs associated with primary treatment and ongoing costs of care not related to distant recurrence. This might have affected the study results, as the use of adjuvant chemotherapy may also affect the risks of local and regional recurrence. The direct medical costs were estimated from the literature and from databases using reimbursement rates for a managed care organisation. The costs were not reported separately from the quantities, which will limit the generalisability of the study results to other settings. The direct patient costs were estimated from IRS reimbursement rates and average US wage rates for women. It is unlikely that these costs are transferable to settings outside of the USA. A sensitivity analysis of health state costs was undertaken, but the source of the ranges tested was unclear. The authors clearly reported the price year and the discount rate used. However, the decision to calculate the costs of distant recurrence as a one-off total cost at the point of recurrence, rather than calculating the annual costs associated with distant recurrence, raises the issue of whether, given this fact, the discounting was executed incorrectly.
Other issues The authors did not compare the results of their analysis with findings from other studies. The issue of generalisability to other settings was not addressed. The authors do not appear to have presented their results selectively. The authors' conclusions assume that, as GEP was found to be less effective than NIH guidelines, it could not be considered cost-effective. However, as it was also found to be less costly than the use of NIH guidelines, an ICER could have been calculated and compared with a threshold for the willingness to accept a reduction in QALYs. The authors acknowledged that their analysis might not reflect all the aspects of the decision problem. They acknowledged that the use of GEP might influence the behaviour of physicians and patients and their attitudes toward adjuvant chemotherapy, and that the uptake of GEP might be influenced by the risk attitudes of patients. The authors also acknowledged that the Dutch study sample used to validate the GEP test did not receive treatment typical of patients in the USA, which may limit the validity of applying the effectiveness parameters in a US setting.
Implications of the study The authors suggested that research into the validation of GEP tests should proceed, as this may enhance their prognostic value.
Bibliographic details Oestreicher N, Ramsey S D, Linden H M, McCune J S, Van't Veer L J, Burke W, Veenstra D L. Gene expression profiling and breast cancer care: what are the potential benefits and policy implications? Genetics in Medicine 2005; 7(6): 380-389 Indexing Status Subject indexing assigned by NLM MeSH Adult; Breast Neoplasms /genetics /therapy; Chemotherapy, Adjuvant; Cost-Benefit Analysis; Decision Support Techniques; Disease-Free Survival; Female; Gene Expression Profiling; Health Policy; Humans; Middle Aged; Models, Econometric; Practice Guidelines as Topic; Quality-Adjusted Life Years; Radiotherapy, Adjuvant; Sensitivity and Specificity; Survival Rate AccessionNumber 22005001388 Date bibliographic record published 30/06/2006 Date abstract record published 30/06/2006 |
|
|
|