|
Cost-effectiveness of screening for colorectal cancer in the general population |
Frazier A L, Colditz G A, Fuchs C S, Kuntz K M |
|
|
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 study investigated 22 screening strategies for colorectal cancer (CRC). All strategies began at age 55 and repeated events carried on until age 85. The strategies considered may be summarised as follows:
flexible sigmoidoscopy (SIG) with follow-up colonoscopy following a positive test result (SIG1), as a one-off, every 5 years or 10 years;
one-off rehydrated faecal occult blood testing (RFOBT) or unhydrated faecal blood testing (UFOBT), both with SIG1 every 5/10 years, with SIG2 every 5/10 years or without either;
SIG with follow-up colonoscopy if any adenomatous polyp discovered (SIG2), as a one-off, every 5 years or 10 years; and
colonoscopy (COL) every 3 years for individuals diagnosed with a high-risk polyp. COL every 10 years. Double contrast barium enema (DCBE), as a one-off, every 5 years or 10 years. No screening at all was also included.
Of the above, faecal occult blood test(FOBT), SIG every 5 years, annual FOBT + SIG every 5 years, DCBE every 5 to 10 years or COL every 10 years were recommended from the age of 55 by a recent expert panel in the USA.
Economic study type Cost-effectiveness analysis.
Study population The hypothetical study population was representative of the 50-year old US population, based on sex and race.
Setting The setting was simulated clinical practice. The economic study was carried out in the USA.
Dates to which data relate The effectiveness data were derived from studies published between 1961 and 2000. Prices were reported in 1998 dollars.
Source of effectiveness data The effectiveness data were derived from a review of the literature, augmented by the authors' assumptions.
Modelling A state-transition Markov model was used to simulate the evolution from normal colonic epithelium to adenomatous polyp to malignancy and death. It used a 1-year cycle from the age of 55 to 85. Superimposed on this was a screening mechanism that allowed for the detection and treatment of cancer.
Outcomes assessed in the review All of the following commenced at age 55 and formed the input parameters to the model. Repeated events continued until 85.
The prevalence of polyps at age 50 (PP50) and its distribution between distal (DP50) and high-risk polyps (HRP50).
Annual transition probabilities (ATP) between stages in the colorectal disease history.
CRC diagnosis probabilities (CRCDP) due to symptoms.
Annual CRC specific mortality rates (ACRCMR): localised cancer (LC), regional cancer (RC) and distal cancer (DC).
Sensitivities and specificities for polyps and cancer in both UFOBT and RFOBT.
Sensitivities and specificities for high-risk polyps (HRP), low-risk polyps (LRP) and cancer (C) in both DCBE and SIG1 or SIG2.
Patient compliance rates.
Study designs and other criteria for inclusion in the review 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 Twenty-two primary studies were included in the review.
Methods of combining primary studies Investigation of differences between primary studies Results of the review Only point estimates and ranges used for sensitivity analysis were reported (see Outcomes above for the abbreviations. Other abbreviations were: sensitivity (Sen) and specificity (Spe)).
PP50 21% (range: 11 - 42);
DP50 61% (range: 30 - 70);
HRP50 2% (range: 1 - 10);
ATP for normal epithelium to LRP age specific, LRP to HRP 0.02 (range: 0.01 - 0.04), HRP to LC 0.05 (range: 0.02 - 0.10), LC to RC 0.28 (range: 0.10 - 0.50), RC to DC 0.63 (range: 0.32 - 0.8);
CRCDP for LC 25 (range: 15 - 35), RC 55 (range: 45 - 65), DC 100 (range: 90 - 100);
ACRCMR for LC 0.002, RC 0.032, DC 0.566;
UFOBT for sensitivity for polyps (SenP) 10 (range: 5 - 20), sensitivity for cancer (SenC) 60 (range: 40 - 65), Spe 90 (range: 85 - 95);
RFOBT for SenP 10 (range: 5 - 20), SenC 60 (range: 40 - 65), Spe 90 (range: 85 - 95);
DCBE for SenLRP 30 (range: 0 - 70), SenHRP 50 (range: 50 - 70), SenC 70 (range: 60 - 90), Spe 86 (range: 80 - 98); and
SIG1/2 for SenLRP 85 (range: 80 - 85), SenHRP/C 95 (range: 85 - 90), Spe 100.
Compliance was estimated to be 60% for initial tests and 80% for follow-up COL.
These are the results for white men only, although white women and blacks were also studied.
Methods used to derive estimates of effectiveness The estimates of cancer progression and symptom detection were varied across clinically plausible ranges so that the stage distribution and CRC incidence predicted by the study model were similar to those reported by the Surveillance Epidemiology and End Results (SEER) programme. (See Ries et al in 'Other Publications of Related Interest' below). All other assumptions cited directly below were the authors' assumptions.
Estimates of effectiveness and key assumptions Key assumptions made in the model included the following: all cancers arose from polyps, polyps did not progress with time, the progression of polyps in the distal versus proximal colon were modelled as independent events, the sensitivity of FOBT was the same for initial and repeated tests and polypectomy would be performed at the time of an initial SIG if a polyp were found.
Measure of benefits used in the economic analysis The benefit measure used in the economic analysis was the life expectancy associated with each screening strategy from the start of screening, i.e. age 55. Life-years were derived from the decision model and a 3% discount rate was applied. The following outputs to the model were also reported, although they were only intermediate outputs: the reductions in CRC incidence and in CRC mortality.
Direct costs Only point estimates for costs and ranges for sensitivity analysis were reported. Unit costs and resources were not reported separately. Although the study perspective was stated to have been societal, only hospital costs were included. This may be due to the complexity of the model. Costs included the actual costs of medical personnel and supplies to provide the service as well as overhead costs, such as administration, charting and automated information systems. All cost data were transformed into 1998 dollars using the medical care component of the Consumer Price Index. Discounting was carried out at an annual discount rate of 3%. The model was simulated over thirty years. All cost data were obtained from a published study on one Health Maintenance Organisation. (See Taplin under 'Other Publications of Related Interest' below).
Statistical analysis of costs Only point estimates and ranges of costs were reported, which were used in the sensitivity analysis.
Indirect Costs No indirect costs were included.
Sensitivity analysis Sensitivity analyses were conducted to assess the robustness of the estimated cost-effectiveness ratios to variations in the model parameters. The parameters used were the outcomes specified above plus the discount rate, which was varied between 1 and 5%. An additional parameter identified as critical was compliance. This determined two base-case analyses: at 60% compliance and at 100% compliance. It was not explicitly stated but it is likely that 1-way sensitivity analysis was conducted. It was done for maximum/minimum values of selected variables.
Estimated benefits used in the economic analysis All screening strategies resulted in reductions in CRC incidence & mortality. The paper reported the life-expectancy, reduction in CRC incidence and mortality for each of the 22 strategies. Fifteen of the strategies were dominated by other strategies. Incremental benefits were calculated for the 7 non-dominated strategies compared to the next most cost-effective strategy. The Life Expectancy (LE) and the incremental benefits (IB) in life-days gained for the non-dominated strategies were:
No screen: (LE) 17.34810, (IB) 0;
SIG1 at 55 years: (LE) 17.3632, (IB) 5.5;
SIG2 at 55 years: (LE) 17.3654, (IB) 0.8;
SIG1 every 10 years: (LE) 17.3732, (IB) 2.8;
SIG2 every 10 years: (LE) 17.3775, (IB) 1.6;
UFOBT + SIG2 every 10 years: (LE) 17.4022, (IB) 9;
UFOBT + SIG2 every 5 years: (LE) 17.4066, (IB) 1.6; and
RFOBT + SIG2 every 5 years: (LE) 17.411, (IB) 1.6.
These are the results for white men only, although white women and blacks were also studied.
Cost results The total intervention costs for the non-dominated interventions were as follows (1998 dollars):
No screen: 1,052;
SIG1 at 55 years: 1,070;
SIG2 at 55 years: 1,095;
SIG1 every 10 years: 1,218;
SIG2 every 10 years: 1,288;
UFOBT + SIG2 every 10 years: 1,810;
UFOBT + SIG2 every 5 years: 2,034; and
RFOBT + SIG2 every 5 years: 2,448.
The discount rate was 3%. The study period covered 30 years. No adverse effects were stated.
Synthesis of costs and benefits The cost-effectiveness (C/E) ratio used was in dollars per life-year gained units. An incremental analysis was performed with incremental C/E ratios calculated for the non-eliminated strategies. There were 7 remaining strategies in total. The incremental C/E ratios follow for a compliance rate of 60%:
No screen: 0;
SIG1 at 55 years: $1,200;
SIG2 at 55 years: $11,000;
SIG1 every 10 years: $15,800;
SIG2 every 10 years: $16,100;
UFOBT + SIG2 every 10 years: $21,200;
UFOBT + SIG2 every 5 years: $51,200; and
RFOBT + SIG2 every 5 years: $92,900.
These are the results for white men only, although white women and blacks were also studied. However, the authors reported that the same trend of cost-effectiveness across the different strategies occurred for all men and women.
In addition, RFOBT compared with UFOBT had an incremental C/E ratio of $45,000/life-year saved.
At a compliance rate of 100%, screening more frequently than every 10 years became prohibitively expensive. COL once at 55 years replaced SIG every 10 years on the cost-effective frontier.
Maximum-minimum sensitivity analysis was reported. RFOBT + SIG2 was most sensitive to compliance with follow-up COL. Slight changes in UFOBT caused it to become non-dominated with a C/E ratio of less than $20,000. COL was the most sensitive test but also the most expensive. If the cost of screening were less than the costs of averted treatment then it was less costly to screen than not to screen.
Compliance has been stressed in this study. The authors suggested emphasizing trial obtained levels of compliance even though, ideally, we may wish for full compliance.
Authors' conclusions The study found that screening for CRC is as cost-effective as other forms of cancer screening. For example, it compares favourably with annual Papanicolau testing beginning at 20 years ($99,000/life-year saved). RFOBT + SIG2 every 5 years was the most effective strategy with an 80% reduction in CRC mortality and an incremental C/E ratio of $92,900/life-year saved. Significant progress can be made with a single screen.
CRD COMMENTARY - Selection of comparators The choice of comparators used was thorough and appropriate. They included the screening strategies currently recommended by a panel of experts, other related alternatives and no screen at all. You, as the user of this database should decide if they are valid comparators in your setting.
Validity of estimate of measure of effectiveness The estimates of measures of effectiveness were based on several published studies. The authors appear to have included an appropriate list of inputs in the model. Sensitivity analysis was used to investigate ranges in the parameters of the model. All these points enhance the internal validity and generalisability of the estimates.
Narrative methods were used to present the findings of the studies. The authors did not report the quality of the parent studies, e.g. the design, or how the estimates of measures of effectiveness were determined from the literature estimates. The authors made a number of assumptions that were used in the Markov model. Of these, the authors noted that the following may not be realistic: all cancers arose from polyps; polyp progression did not depend on time; and the progression of polyps in the distal and proximal colon was modelled as independent. Further, it was assumed that the sensitivity of FOBT was the same for initial and repeated tests. This may bias the results towards FOBT in the long-term. These points limit the internal validity and generalisability of the estimates.
Validity of estimate of measure of benefit Life expectancy from the start of screening at age 55 was used as the benefit measure for each strategy. This seems to be appropriate for a screening programme of CRC. Furthermore, it permits the comparison of other health care benefits in the health care system. A Markov model was used to estimate benefits and appropriate discounting was performed.
Validity of estimate of costs The analysis of the costs was conducted from the perspective of a health organisation and it appears that all relevant categories of costs were included in the economic evaluation. Sensitivity analyses were conducted on the cost data, which were treated deterministically. The price year was reported, thus simplifying reflation exercises to other settings. As lifetime costs were assessed, appropriate discounting was performed and the impact of a different discount rate was tested in the sensitivity analyses. All these points enhance the internal validity and generalisability of the estimates.
Little information was provided about how each cost category was calculated. Unit costs and quantities were not reported separately. These points limit the internal validity and generalisability of the estimates.
Other issues The authors compared this model for only two strategies (UFOBT and UFOBT + SIG every 5 years) with slight adjustment to the assumptions to two previous models (see Eddy, and Wagner et al in 'Other Publications of Related Interest' below), which evaluated those strategies. The life-expectancy gains for this study were intermediate to those published previously, and the costs were consistently higher.
The authors acknowledged the limitations of their assumptions regarding the disease path and treatment protocol, which were listed under the 'Validity of Estimate of Effectiveness' section above.
The model was data-intensive given the approach used. However, inclusion of patient direct costs and indirect costs would have enabled the paper to keep to its stated societal perspective.
The extensive sensitivity analysis enhances the generalisability of the findings to other settings/populations. The study population was representative of the US population, which also helps generalise the results.
The authors noted that the compliance rate for screening with CRC is only 20% in the USA. The authors argued that it is better to follow clinical trial values (60%) than perfect values (100%). However, clinical trial values often achieve greater compliance than in a clinical setting. Since compliance has been highlighted as an important parameter, it would also be of value to investigate lower values of compliance. Both compliance rates (patient behaviour) and the treatment protocol for polypectomy, given that a polyp is diagnosed (clinical practice), may vary considerably across locations and so must be considered for the reader's country or institution if the results are to be generalised to those settings.
Implications of the study All of the dominant strategies have a cost-effectiveness that is within society's willingness to pay when compared to the cost-effectiveness of other cancer screening programmes. Compliance, however, is an important factor and policy should encourage increased compliance. The authors also suggest that the choice of screening test for CRC be based not only on the cost-effectiveness analysis but also on other factors such as: local expertise, availability of providers and patient preferences.
Source of funding Supported in part by grant T320K07703 from the National Cancer Institute, grant 5 K07 CA62252-03 from the National Cancer Institute and grant HS07038 from the Agency for Health Care Policy and Research, and by the Nurses' Health Study.
Bibliographic details Frazier A L, Colditz G A, Fuchs C S, Kuntz K M. Cost-effectiveness of screening for colorectal cancer in the general population. JAMA 2000; 284(15): 1954-1961 Other publications of related interest Ries L A G, Miller B A, Hankey B F, et al. SEER cancer statistics review, 1973-1991: tables and graphs. Bethesda, Md: National Cancer Institute; 1994.
Eddy D M. Screening for colorectal cancer. Annals of Internal Medicine 1990;113: 373-384.
Wagner J, Tunis S Brown M, Ching A, Almeida R. Cost-effectiveness of colorectal cancer screening in average-risk adults. In: Young GP, Rozen P, Levin B, editors. Prevention and early detection of colorectal cancer. London: W B Saunders Co; 1996. pp. 321-356.
Taplin S H, Barlow W, Urban N, et al. Stage, age, comorbidity, and direct costs of colon, prostate, and breast cancer care. Journal of the National Cancer Institute 1995;87:417-426.
Indexing Status Subject indexing assigned by NLM MeSH Colonic Polyps /prevention & Colonoscopy /economics; Colorectal Neoplasms /economics /epidemiology /prevention & Cost-Benefit Analysis; Humans; Life Expectancy; Life Tables; Male; Markov Chains; Mass Screening /economics /methods; Middle Aged; Occult Blood; Patient Compliance; Risk Factors; Sigmoidoscopy /economics; control; control AccessionNumber 22000008306 Date bibliographic record published 31/10/2003 Date abstract record published 31/10/2003 |
|
|
|