Scandinavian Journal of Gastroenterology
ISSN: 0036-5521 (Print) 1502-7708 (Online) Journal homepage: http://www.tandfonline.com/loi/igas20
Rehydration of Guaiac-Based Faecal Occult Blood Tests in Mass Screening for Colorectal Cancer: An Economic Perspective A. R. Walker, D. K. Whynes & J. D. Hardcastle To cite this article: A. R. Walker, D. K. Whynes & J. D. Hardcastle (1991) Rehydration of Guaiac-Based Faecal Occult Blood Tests in Mass Screening for Colorectal Cancer: An Economic Perspective, Scandinavian Journal of Gastroenterology, 26:2, 215-218, DOI: 10.3109/00365529109025033 To link to this article: http://dx.doi.org/10.3109/00365529109025033
Published online: 08 Jul 2009.
Submit your article to this journal
Article views: 7
View related articles
Citing articles: 1 View citing articles
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=igas20 Download by: [University of Cincinnati Libraries]
Date: 23 March 2016, At: 17:54
Rehydration of Guaiac-Based Faecal Occult Blood Tests in Mass Screening for Colorectal Cancer An Economic Perspective
Downloaded by [University of Cincinnati Libraries] at 17:54 23 March 2016
A. R. WALKER, D. K. WHYNES & J . D. HARDCASTLE Dept. of Surgery, University Hospital, and Dept. of Economics, University of Nottingham, Nottingham, U.K.
Walker AR, Whynes DK, Hardcastle JD. Rehydration of guaiac-based faecal occult blood tests in mass screening for colorectal cancer. An economic perspective. Scand J Gastroenterol 1991, 26, 215-218 Owing to dehydration during storage, faecal occult blood tests have been found to lose sensitivity; accordingly, test rehydration before development has been advocated, although this practice has yet to be subjected to an economic evaluation. In this paper, the results from two major screening trials in Sweden and England, one using rehydration and the other not, are so evaluated, based on a costing model developed within the English trial. The higher sensitivity resulting from rehydration was found to be accompanied by losses in specificity, such that, although more cancers are detected, the costs of screening and of cancer detection are actually considerably higher under the rehydration regimen than with non-hydration.
Key words: Cancer; colorectal cancer; cost-effectiveness; economic evaluation; faecal occult blood test; mass screening David K . Whynes, Dept. of Economics, University of Nottingham, University Park, Nottingharn NG7 2RD, lJ.K.
Faecal occult blood tests are the principal method of mass population screening for colorectal cancer currently being evaluated in large randomized trials. Tests such as Haemoccult are performed in subjects’ homes; pea-sized stool samples are placed onto ‘windows’ of guaiac-impregnated paper, and the test is returned to a hospital laboratory for development. When such tests were first introduced, it was suggested that the storage of returncd tests for a period of time in excess of a few days would tend to decrease the test’s sensitivity, owing to dehydration. Reactions that would have appeared as weakly positive if development had taken place immediately would actually produce negative results after protracted storage. The rehydration of test samples before development counteracts this effect, however, and sensitivity is thereby enhanced, although specificity is correspondingly reduced (1). In economic terms, this specificity reduction entails
a cost in terms of the increased number of falsepositive results requiring further investigation. Although some researchers now recommend the rehydration of all tests, debate on the virtues of this procedure to date have been largely subjective-one group discussion (2) suggested that there was ‘at least an intuitive feeling’ (p. 58) that any increased yield would be outweighed by increased cost. This paper presents a more thorough economic evaluation of the evidence available by comparing two specific test protocols. MATERIALS AND METHODS The discussion is confined to the detection of asymptomatic cancers and relates to data derived from two contemporary trials, one of which uses the hydration technique and one of which does not. The Gothenburg randomized controlled trial (3) initally divided its study group into two cohorts
Downloaded by [University of Cincinnati Libraries] at 17:54 23 March 2016
216
A . R . Walker et al.
in accordance with age (60-62 years and 62-64 years) and rehydrated the completed tests of the older group. Although compliance, positive rates, and detection rates are all known to be agerelated (thus raising the possibility of intrinsic differences between the two groups), the very small difference in age implies this effect can be presumed insignificant. The reported sensitivity of unhydrated tests for this trial was far lower than that found elsewhere (4,5), although the figure is in line with some expectations (6). In contrast, the randomized controlled screening trial in Nottingham (4) does not hydrate any of the returned tests before development. The basic data from the two trials illustrate the expected trade-off between sensitivity and specificity: hydrated tests are more sensitive (Gothenburg’s 85% against Nottingham’s 65%), whereas nonhydrated tests are more specific (Nottingham’s 99% against Gothenburg’s 95%). A costing model of faecal occult blood (FOB) screening based on the Nottingham trial data has been developed (7) and is used in the following analysis. This model simulates the costs of screening a standardized population of 75,000 asymptomatic individuals, aged 5@74 years, with a cancer prevalence of 3.9 per 1000 and a screening compliance rate of 57.8%. The model consists of three major cost elements: i) costs of FOB tests, including distribution to subjects and test development, ii) costs of follow-up investigation of subjects with positive FOB test results by colonoscopy or barium enema roentgenography, and iii) overhead costs, including administration. For a given population the expected yield of cancers can be calculated by means of prevalence and detection and positive rates, assuming no differences in prevalence between accepting and non-responding groups. Yield may then be compared with the costs of screening, including the cost of further investigation of positive test results and of administering the program, costs being expressed as ‘detection costs per cancer found’ and ‘cost per person screened’. The detection of asymptomatic cancers is assumed to represent the outcome of the screening program, although, in reality, it is only an interim measure; no mortality data are yet available from either trial to enable
final outcomes, in terms of gains in quantity and quality of life, to be calculated. RESULTS The specificity and sensitivity findings of the two trials can be translated into positive rates and detection rates; the mathematics of this procedure are presented in the Appendix. These rates serve as parameters for the costing model of hydration versus non-hydration and are presented in Table I. Both are naturally higher for the hydration case. Table I also presents the cost and detection results for the two alternative protocols. As is evident, hydration is successful in detecting more cancers than non-hydration, but at greatly increased cost: total and average screening costs are almost doubled, and cost per cancer detected is 46.8% higher in the hydrated case. A minor element in this differential is the increased test development costs incurred by virtue of the necessity of rehydrating each test, but the principal factor explaining the cost difference derives from the relatively poor specificity (and thus low positive predictive value) of hydrated tests; lower specificity entails a proportionately higher number of follow-up investigations for false-positive subjects. From the incremental point of view, hydration increases the yield of cancers by 34 at an additional cost of &208,855-that is, at an average incremental cost per cancer detected of f6,143 (approximately three times the average cost under non-hydration). Interestingly, even if hydration were to be 100% sensitive and thus capable of detecting all 169 cancers in the model population (detection rate = prevalence = 3.9 per l000), the cost per cancer detected with this technique would only fall to f2,597, which is still 24.9% higher than in the non-hydration case. O n this assumption, the non-hydration technique would still generate the lower cost per cancer detected as long as its detection rate remained above 1.95 per 1000 (50% sensitivity). One limitation of the analysis so far is that only detection costs are considered. However, it is quite possible that interval, symptomatic, cancers would cost more to treat than if detected at their asymptomatic stage, and relatively more of these
Rehydration: An Economic Perspective
217
Downloaded by [University of Cincinnati Libraries] at 17:54 23 March 2016
Table 1. Cost comparison for hydration and non-hydration of FOB tests for a standardized population of 75,000
Detection rate (per 1000) Positive rate (%) No. of cancers detected Total cost of screening (f), of which: FOB tests (2) Follow-up investigations (f) Overhead costs (f) Cost per cancer detected (f) Incremental cost per cancer detected (f) Cost per person screened (f) Incremental cost per person screened (f)
would be likely to arise under the non-hydration scenario. Preliminary estimates of the net difference in treatment costs is f612 per interval cancer ( 8 ) . For the 34 extra cancers that might present under the non-hydration protocol, therefore, an additional cost of f20,808 needs to be included. Even so, this inclusion only raises the cost per cancer detected in the non-hydration case to f2,270, which is still 25.6% below the hydration figure.
Conclusion Our evaluation of the mixed evidence suggests that both the cost per cancer detected and the cost per person screened are considerably lower under the non-hydration scenario, as compared with the hydration protocol. Specifically, additional cancer yield from hydration is obtained at approximately three times the average cost of yield from non-hydration. It must be recognized, however, that the leastcost solution to a problem is not necessarily the cost-effective solution. O u r analysis measures only costs in relation to intermediate outcomes (cancers detected) and makes no claims with regard to whether broader outcome objectives would actually be more appropriate. In fact, the use of the more expensive protocol by medical decision makers could be legitimated by the selection of alternative outcome criteria. First, we have shown that hydration succeeds in detecting more cancers at more-than-proportionately
Non-hydration
Hydration
2.54 1.25 110 228,966 119,165 64,303 45,498 2.079
3.32 5.31 144 437,821 120,387 271,936 45,498 3,051 6,143 10.13 4.85
5.28
higher cost; however, if society is willing to pay at least this amount for the detection of the additional cancers, then hydration becomes the appropriate protocol. As noted above, a more precise statement in this respect will only be possible when the trials’ final outcomes become available. Second, it is conceivable that, were a mass screening program to be introduced, a high sensitivity rate would be seen as important in maintaining public confidence in screening. In that hydration appears to offer a higher test sensitivity than non-hydration, decision makers might well opt for this alternative and be willing to pay the additional cost premium to obtain the sensitivity gains. Whether additional yield from screening under either protocol can be achieved at reduced cost remains a moot point. There are two recognized means of reducing the false-positive rate, although the Gothenburg results already embody one of these (dietary restriction). The alternative (retesting positive results before investigation) has been attempted at Nottingham, using unhydrated tests, and with some success (9, 10). To date, no published data on the application of such a strategy to hydrated tests are available.
ACKNOWLEDGEMENTS The authors acknowledge the financial support of the Medical Research Council and the helpful comments of two referees.
Downloaded by [University of Cincinnati Libraries] at 17:54 23 March 2016
218
A . R. Walker et al.
REFERENCES
APPENDIX
1. Kewenter J, Bjorck S, Haglind E, Smith L, Svanvik J, Ahren C. Screening and rescreening for colorectal cancer: a controlled trial of faecal occult blood testing in 27,700 subjects. Cancer 1988, 62, 645-651 2. Discussion of colorectal cancer screening. In: Chamberlain J, Miller AB, eds. Screening for gastrointestinal cancer. Huber, Toronto, 1988 3. Kewenter J, Haglind E , Svanvik J. Faecal occult blood screening for colorectal cancer: the Swedish experience. In: Faivre J, Hill M, eds. Causation and prevention of colorectal cancer. Excerpta Medica, Amsterdam, 1987 4. Hardcastle JD, Thomas WM, Chamberlain J, et al. Randomised controlled trial of faecal occult blood screening for colorectal cancer: results for 107,349 subjects. Lancet 1989, 1160-1164 5. Kronborg 0, Fenger C, Olsen J, Beck K, Sondergaard 0. Repeated screening for colorectal cancer with faecal occult blood tests. Scand J Gastroenterol 1989, 24, 599-606 6. Simon J. Occult blood screening for colorectal cancer: a critical review. Gastroenterology 1985,88, 82CL837 7. Walker A, Whynes DK, Hardcastle JD, Chamberlain J. The costs of screening for colorectal cancer. J Epidemiol Community Health (in press) 8. Tuck J, Walker A, Whynes DK, Pye G , Hardcastle JD, Chamberlain J. Screening and the costs of treating colorectal cancer: some preliminary results. Public Health 1989, 103, 413-419 9. Thomas WM, Pye G, Hardcastle JD, Chamberlain J , Charnley RM. The role of dietary restrictions and three-month retesting in FOB screening for colorectal cancer. Br J Surg 1989, 76, 976-978 10. Walker A, Whynes DK, Hardcastle JD, Chamberlain J. Retesting positive results in screening for colorectal cancer: a marginal analysis. Appl Econom (in press)
Consider a target population of N individuals. As a result of screening there will emerge: a true-positive results b false-positive results c false-negative results d true-negative results that is, a + b c + d = N. By definition: Prevalence = (a + c)/N Sensitivity = a/(a + c) Specificity = d/(b + d) Detection rate = a/N Positive rate = (a + b)/N From these definitions: (i) (a + c) = N x Prevalence = a/Sensitivity therefore a = N x Prevalence x Sensitivity (ii) c = (N x Prevalence - a) = N x Prevalence x (1 - Sensitivity) (iii) (b + d) = N - (a + c) = N x (1 - Prevalence), and given d = (b + d) x Specificity. d = N x (1 - Prevalence) x Specificity (iv) Given (b + d), b = N x (1 - Prevalence) x (1 - Specificity). It therefore follows that: Detection rate = Prevalence x Sensitivity Number of cancers detected = N x Detection rate x Compliance Positive rate = (Prevalence x Sensitivity) + {(I - Prevalence) x (1 - Specificity)}.
Received 28 May 1990 Accepted 8 August 1990
+
ISSN: 0036-5521 (Print) 1502-7708 (Online) Journal homepage: http://www.tandfonline.com/loi/igas20
Rehydration of Guaiac-Based Faecal Occult Blood Tests in Mass Screening for Colorectal Cancer: An Economic Perspective A. R. Walker, D. K. Whynes & J. D. Hardcastle To cite this article: A. R. Walker, D. K. Whynes & J. D. Hardcastle (1991) Rehydration of Guaiac-Based Faecal Occult Blood Tests in Mass Screening for Colorectal Cancer: An Economic Perspective, Scandinavian Journal of Gastroenterology, 26:2, 215-218, DOI: 10.3109/00365529109025033 To link to this article: http://dx.doi.org/10.3109/00365529109025033
Published online: 08 Jul 2009.
Submit your article to this journal
Article views: 7
View related articles
Citing articles: 1 View citing articles
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=igas20 Download by: [University of Cincinnati Libraries]
Date: 23 March 2016, At: 17:54
Rehydration of Guaiac-Based Faecal Occult Blood Tests in Mass Screening for Colorectal Cancer An Economic Perspective
Downloaded by [University of Cincinnati Libraries] at 17:54 23 March 2016
A. R. WALKER, D. K. WHYNES & J . D. HARDCASTLE Dept. of Surgery, University Hospital, and Dept. of Economics, University of Nottingham, Nottingham, U.K.
Walker AR, Whynes DK, Hardcastle JD. Rehydration of guaiac-based faecal occult blood tests in mass screening for colorectal cancer. An economic perspective. Scand J Gastroenterol 1991, 26, 215-218 Owing to dehydration during storage, faecal occult blood tests have been found to lose sensitivity; accordingly, test rehydration before development has been advocated, although this practice has yet to be subjected to an economic evaluation. In this paper, the results from two major screening trials in Sweden and England, one using rehydration and the other not, are so evaluated, based on a costing model developed within the English trial. The higher sensitivity resulting from rehydration was found to be accompanied by losses in specificity, such that, although more cancers are detected, the costs of screening and of cancer detection are actually considerably higher under the rehydration regimen than with non-hydration.
Key words: Cancer; colorectal cancer; cost-effectiveness; economic evaluation; faecal occult blood test; mass screening David K . Whynes, Dept. of Economics, University of Nottingham, University Park, Nottingharn NG7 2RD, lJ.K.
Faecal occult blood tests are the principal method of mass population screening for colorectal cancer currently being evaluated in large randomized trials. Tests such as Haemoccult are performed in subjects’ homes; pea-sized stool samples are placed onto ‘windows’ of guaiac-impregnated paper, and the test is returned to a hospital laboratory for development. When such tests were first introduced, it was suggested that the storage of returncd tests for a period of time in excess of a few days would tend to decrease the test’s sensitivity, owing to dehydration. Reactions that would have appeared as weakly positive if development had taken place immediately would actually produce negative results after protracted storage. The rehydration of test samples before development counteracts this effect, however, and sensitivity is thereby enhanced, although specificity is correspondingly reduced (1). In economic terms, this specificity reduction entails
a cost in terms of the increased number of falsepositive results requiring further investigation. Although some researchers now recommend the rehydration of all tests, debate on the virtues of this procedure to date have been largely subjective-one group discussion (2) suggested that there was ‘at least an intuitive feeling’ (p. 58) that any increased yield would be outweighed by increased cost. This paper presents a more thorough economic evaluation of the evidence available by comparing two specific test protocols. MATERIALS AND METHODS The discussion is confined to the detection of asymptomatic cancers and relates to data derived from two contemporary trials, one of which uses the hydration technique and one of which does not. The Gothenburg randomized controlled trial (3) initally divided its study group into two cohorts
Downloaded by [University of Cincinnati Libraries] at 17:54 23 March 2016
216
A . R . Walker et al.
in accordance with age (60-62 years and 62-64 years) and rehydrated the completed tests of the older group. Although compliance, positive rates, and detection rates are all known to be agerelated (thus raising the possibility of intrinsic differences between the two groups), the very small difference in age implies this effect can be presumed insignificant. The reported sensitivity of unhydrated tests for this trial was far lower than that found elsewhere (4,5), although the figure is in line with some expectations (6). In contrast, the randomized controlled screening trial in Nottingham (4) does not hydrate any of the returned tests before development. The basic data from the two trials illustrate the expected trade-off between sensitivity and specificity: hydrated tests are more sensitive (Gothenburg’s 85% against Nottingham’s 65%), whereas nonhydrated tests are more specific (Nottingham’s 99% against Gothenburg’s 95%). A costing model of faecal occult blood (FOB) screening based on the Nottingham trial data has been developed (7) and is used in the following analysis. This model simulates the costs of screening a standardized population of 75,000 asymptomatic individuals, aged 5@74 years, with a cancer prevalence of 3.9 per 1000 and a screening compliance rate of 57.8%. The model consists of three major cost elements: i) costs of FOB tests, including distribution to subjects and test development, ii) costs of follow-up investigation of subjects with positive FOB test results by colonoscopy or barium enema roentgenography, and iii) overhead costs, including administration. For a given population the expected yield of cancers can be calculated by means of prevalence and detection and positive rates, assuming no differences in prevalence between accepting and non-responding groups. Yield may then be compared with the costs of screening, including the cost of further investigation of positive test results and of administering the program, costs being expressed as ‘detection costs per cancer found’ and ‘cost per person screened’. The detection of asymptomatic cancers is assumed to represent the outcome of the screening program, although, in reality, it is only an interim measure; no mortality data are yet available from either trial to enable
final outcomes, in terms of gains in quantity and quality of life, to be calculated. RESULTS The specificity and sensitivity findings of the two trials can be translated into positive rates and detection rates; the mathematics of this procedure are presented in the Appendix. These rates serve as parameters for the costing model of hydration versus non-hydration and are presented in Table I. Both are naturally higher for the hydration case. Table I also presents the cost and detection results for the two alternative protocols. As is evident, hydration is successful in detecting more cancers than non-hydration, but at greatly increased cost: total and average screening costs are almost doubled, and cost per cancer detected is 46.8% higher in the hydrated case. A minor element in this differential is the increased test development costs incurred by virtue of the necessity of rehydrating each test, but the principal factor explaining the cost difference derives from the relatively poor specificity (and thus low positive predictive value) of hydrated tests; lower specificity entails a proportionately higher number of follow-up investigations for false-positive subjects. From the incremental point of view, hydration increases the yield of cancers by 34 at an additional cost of &208,855-that is, at an average incremental cost per cancer detected of f6,143 (approximately three times the average cost under non-hydration). Interestingly, even if hydration were to be 100% sensitive and thus capable of detecting all 169 cancers in the model population (detection rate = prevalence = 3.9 per l000), the cost per cancer detected with this technique would only fall to f2,597, which is still 24.9% higher than in the non-hydration case. O n this assumption, the non-hydration technique would still generate the lower cost per cancer detected as long as its detection rate remained above 1.95 per 1000 (50% sensitivity). One limitation of the analysis so far is that only detection costs are considered. However, it is quite possible that interval, symptomatic, cancers would cost more to treat than if detected at their asymptomatic stage, and relatively more of these
Rehydration: An Economic Perspective
217
Downloaded by [University of Cincinnati Libraries] at 17:54 23 March 2016
Table 1. Cost comparison for hydration and non-hydration of FOB tests for a standardized population of 75,000
Detection rate (per 1000) Positive rate (%) No. of cancers detected Total cost of screening (f), of which: FOB tests (2) Follow-up investigations (f) Overhead costs (f) Cost per cancer detected (f) Incremental cost per cancer detected (f) Cost per person screened (f) Incremental cost per person screened (f)
would be likely to arise under the non-hydration scenario. Preliminary estimates of the net difference in treatment costs is f612 per interval cancer ( 8 ) . For the 34 extra cancers that might present under the non-hydration protocol, therefore, an additional cost of f20,808 needs to be included. Even so, this inclusion only raises the cost per cancer detected in the non-hydration case to f2,270, which is still 25.6% below the hydration figure.
Conclusion Our evaluation of the mixed evidence suggests that both the cost per cancer detected and the cost per person screened are considerably lower under the non-hydration scenario, as compared with the hydration protocol. Specifically, additional cancer yield from hydration is obtained at approximately three times the average cost of yield from non-hydration. It must be recognized, however, that the leastcost solution to a problem is not necessarily the cost-effective solution. O u r analysis measures only costs in relation to intermediate outcomes (cancers detected) and makes no claims with regard to whether broader outcome objectives would actually be more appropriate. In fact, the use of the more expensive protocol by medical decision makers could be legitimated by the selection of alternative outcome criteria. First, we have shown that hydration succeeds in detecting more cancers at more-than-proportionately
Non-hydration
Hydration
2.54 1.25 110 228,966 119,165 64,303 45,498 2.079
3.32 5.31 144 437,821 120,387 271,936 45,498 3,051 6,143 10.13 4.85
5.28
higher cost; however, if society is willing to pay at least this amount for the detection of the additional cancers, then hydration becomes the appropriate protocol. As noted above, a more precise statement in this respect will only be possible when the trials’ final outcomes become available. Second, it is conceivable that, were a mass screening program to be introduced, a high sensitivity rate would be seen as important in maintaining public confidence in screening. In that hydration appears to offer a higher test sensitivity than non-hydration, decision makers might well opt for this alternative and be willing to pay the additional cost premium to obtain the sensitivity gains. Whether additional yield from screening under either protocol can be achieved at reduced cost remains a moot point. There are two recognized means of reducing the false-positive rate, although the Gothenburg results already embody one of these (dietary restriction). The alternative (retesting positive results before investigation) has been attempted at Nottingham, using unhydrated tests, and with some success (9, 10). To date, no published data on the application of such a strategy to hydrated tests are available.
ACKNOWLEDGEMENTS The authors acknowledge the financial support of the Medical Research Council and the helpful comments of two referees.
Downloaded by [University of Cincinnati Libraries] at 17:54 23 March 2016
218
A . R. Walker et al.
REFERENCES
APPENDIX
1. Kewenter J, Bjorck S, Haglind E, Smith L, Svanvik J, Ahren C. Screening and rescreening for colorectal cancer: a controlled trial of faecal occult blood testing in 27,700 subjects. Cancer 1988, 62, 645-651 2. Discussion of colorectal cancer screening. In: Chamberlain J, Miller AB, eds. Screening for gastrointestinal cancer. Huber, Toronto, 1988 3. Kewenter J, Haglind E , Svanvik J. Faecal occult blood screening for colorectal cancer: the Swedish experience. In: Faivre J, Hill M, eds. Causation and prevention of colorectal cancer. Excerpta Medica, Amsterdam, 1987 4. Hardcastle JD, Thomas WM, Chamberlain J, et al. Randomised controlled trial of faecal occult blood screening for colorectal cancer: results for 107,349 subjects. Lancet 1989, 1160-1164 5. Kronborg 0, Fenger C, Olsen J, Beck K, Sondergaard 0. Repeated screening for colorectal cancer with faecal occult blood tests. Scand J Gastroenterol 1989, 24, 599-606 6. Simon J. Occult blood screening for colorectal cancer: a critical review. Gastroenterology 1985,88, 82CL837 7. Walker A, Whynes DK, Hardcastle JD, Chamberlain J. The costs of screening for colorectal cancer. J Epidemiol Community Health (in press) 8. Tuck J, Walker A, Whynes DK, Pye G , Hardcastle JD, Chamberlain J. Screening and the costs of treating colorectal cancer: some preliminary results. Public Health 1989, 103, 413-419 9. Thomas WM, Pye G, Hardcastle JD, Chamberlain J , Charnley RM. The role of dietary restrictions and three-month retesting in FOB screening for colorectal cancer. Br J Surg 1989, 76, 976-978 10. Walker A, Whynes DK, Hardcastle JD, Chamberlain J. Retesting positive results in screening for colorectal cancer: a marginal analysis. Appl Econom (in press)
Consider a target population of N individuals. As a result of screening there will emerge: a true-positive results b false-positive results c false-negative results d true-negative results that is, a + b c + d = N. By definition: Prevalence = (a + c)/N Sensitivity = a/(a + c) Specificity = d/(b + d) Detection rate = a/N Positive rate = (a + b)/N From these definitions: (i) (a + c) = N x Prevalence = a/Sensitivity therefore a = N x Prevalence x Sensitivity (ii) c = (N x Prevalence - a) = N x Prevalence x (1 - Sensitivity) (iii) (b + d) = N - (a + c) = N x (1 - Prevalence), and given d = (b + d) x Specificity. d = N x (1 - Prevalence) x Specificity (iv) Given (b + d), b = N x (1 - Prevalence) x (1 - Specificity). It therefore follows that: Detection rate = Prevalence x Sensitivity Number of cancers detected = N x Detection rate x Compliance Positive rate = (Prevalence x Sensitivity) + {(I - Prevalence) x (1 - Specificity)}.
Received 28 May 1990 Accepted 8 August 1990
+
