Cost-effectiveness of Chlamydia antibody tests in subfertile women

Transcription

1 Human Reproduction Page 1 of 8 Hum. Reprod. Advance Access published November 11, 2004 doi: /humrep/deh608 Cost-effectiveness of Chlamydia antibody tests in subfertile women A.A.A.Fiddelers 1,4, J.A.Land 2, G.Voss 1, A.G.H.Kessels 1 and J.L.Severens 1,3 1 Department of Clinical Epidemiology and Medical Technology Assessment and 2 Department of Obstetrics and Gynaecology, Academisch Ziekenhuis Maastricht, PO Box 5800, 6202 AZ Maastricht and 3 Department of Health Organisation, Policy and Economics, University Maastricht, PO Box 616, 6200 MD Maastricht, The Netherlands 4 To whom correspondence should be addressed. Afi@kemta.azm.nl BACKGROUND: For the evaluation of tubal function, Chlamydia antibody testing (CAT) has been introduced as a screening test. We compared six CAT screening strategies (five CAT tests and one combination of tests), with respect to their cost-effectiveness, by using IVF pregnancy rate as outcome measure. METHODS: A decision analytic model was developed based on a source population of 1715 subfertile women. The model incorporates hysterosalpingography (HSG), laparoscopy and IVF. To calculate IVF pregnancy rates, costs, effects, cost-effectiveness and incremental costs per effect of the six different CAT screening strategies were determined. RESULTS: pelisa Medac turned out to be the most cost-effective CAT screening strategy (e per IVF pregnancy), followed by MIF Anilabsystems (e15 108). A combination of tests (pelisa Medac and MIF Anilabsystems; e15 127) did not improve the cost-effectiveness of the single strategies. Sensitivity analyses showed that the results are robust for changes in the baseline values of the model parameters. CONCLUSIONS: Only small differences were found between the screening strategies regarding the cost-effectiveness, although pelisa Medac was the most cost-effective strategy. Before introducing a particular CAT test into clinical practice, one should consider the effects and consequences of the entire screening strategy, instead of only the diagnostic accuracy of the test used. Key words: Chlamydia antibody testing/cost-effectiveness/decision model/diagnostic test Introduction For the evaluation of tubal factor subfertility, Chlamydia antibody testing (CAT), hysterosalpingography (HSG) or laparoscopy can be applied. Because HSG and laparoscopy are invasive and expensive procedures, CAT has been introduced for large-scale screening purposes. The results of CAT are important in the total screening strategy, because its results determine what kind of further testing is indicated (HSG and/or laparoscopy). These decisions, made at the beginning of the screening strategy based on CAT, may have clinical and economic consequences. Because of the variety of tests for CAT available, characterized by different diagnostic accuracy, an important question is: which test for CAT is the most (cost) effective when used in a screening strategy for tubal pathology? In a previous study (Land et al., 2003), we compared the clinical performance of five CAT tests and a combination of tests. In clinical practice, however, the decision, regarding which CAT test is to be preferred in a particular clinical setting should not be based on diagnostic test accuracy alone. Costs related to the performance of the tests, and costs related to the consequences of the tests (of false-positive and false-negative test results in particular) have also to be taken into consideration. The first aim of this study is to compare five commercially available Chlamydia immunoglobulin G antibody tests (MIF Biomerieux, MIF Anilabsystems; ELISA Anilabsystems, pelisa Medac, and ELISA Savyon) with respect to their cost-effectiveness (costs per IVF pregnancy) in a screening strategy for tubal pathology. The second aim is to evaluate whether serial testing, with an enzyme-linked immunosorbent assay (ELISA) test as the first test and retesting of all positive serum samples with a microimmunofluorescence (MIF) test, is to be preferred from an economic point of view. A decision model technique was applied, since it is not possible to answer these questions by means of a prospective cohort study, in which all patients with subfertility problems will have all tests for CAT, HSG and laparoscopy. It is shown how a decision model technique can be used to structure the evidence on clinical and economic outcomes in a form that can help to inform decisions about clinical practices and health care resource allocations (Weinstein et al., 2003). Materials and methods In this study, five commercially available CAT tests (MIF Biomerieux, MIF Anilabsystems, ELISA Anilabsystems, pelisa Medac Human Reproduction q European Society of Human Reproduction and Embryology 2004; all rights reserved Page 1 of 8

2 A.A.A.Fiddelers et al. Figure 1. Decision analytic model of six Chlamydia antibody testing (CAT) screening strategies and a reference strategy to evaluate the costs per IVF pregnancy. The square in the decision tree indicates a decision node, after which only one decision is possible. The nodes represented by circles are used if subsequent outcomes are possible. The model starts at the left node, at which CAT is performed in all women at intake. Independent of the result of CAT, some of these patients will have no further evaluation of tubal function. The remaining patients will have tubal evaluation by hysterosalpingography (HSG) and/or laparoscopy (LS). In the patients with a positive CAT test, LS is done without delay. In patients with a negative CAT test, HSG is performed. In patients with a normal HSG, expectant management is proposed for,6 months. In cases where they have not conceived during this period, a LS is done to exclude tubal pathology. In patients in whom abnormalities are seen at HSG, LS is done shortly after HSG. Complications of HSG and LS have been taken into consideration in the model. In the reference strategy, all patients will have a LS at intake. Page 2 of 8

3 Cost-effectiveness in CAT and ELISA Savyon) were evaluated. The performance of these tests in predicting tubal factor subfertility has been reported previously (Land et al., 2003). In the present study, these five CAT tests, and also the combination of two tests (pelisa Medac and MIF Anilabsystems, Med þ Ani), were part of a theoretical screening strategy for tubal factor subfertility, which also included HSG, laparoscopy and IVF. Furthermore, complications of HSG and laparoscopy were taken into consideration in the model. By using data from the literature on sensitivity, specificity, probabilities and costs and our own (unpublished) clinical data, a decision analytic model was designed to predict the costs and effects (IVF pregnancies) of the six different CAT screening strategies. The seventh strategy evaluated is a reference strategy, in which the patients will have a laparoscopy immediately after intake. This strategy reflects maximal cost-effectiveness. In this model, it was decided to include CAT, HSG and laparoscopy, although there is an ongoing debate about the role and order of these tests in the evaluation of tubal function. The evaluation of different screening strategies for tubal factor subfertility was not the aim of this study, however. Decision analytic model The decision analytic model, shown in Figure 1, is based mainly on clinical guidelines used in the Department of Obstetrics and Gynaecology at the University Hospital Maastricht, and on expert opinion. Since hardly any comparative studies have been done so far, few data could be obtained from the literature (Table I). Therefore, the decision analytic model is partly hypothetical. The principles of good practice for decision analytic modelling according to Weinstein et al. (2003) were used in our model. The model starts at the left node. At this node, at intake, in all subfertile women, CAT will be performed using all six strategies. Irrespective of the result of CAT in the six CAT strategies, a proportion of these patients will have no further evaluation of tubal function because of drop out [i.e. spontaneous pregnancy, diagnosis for which tubal status is not relevant (e.g. severe male factor subfertility) and loss to follow-up]. In the patients with a positive CAT test, who are considered at high risk for tubal pathology, laparoscopy is done without delay. In patients with a negative CAT test, HSG is performed. In patients with a normal HSG, expectant management is defined for half a year. If they have not conceived during this period, a laparoscopy is done to exclude tubal pathology. In patients in whom abnormalities are seen at HSG, laparoscopy is done shortly after HSG. The model, shown in Figure 1, is applied to all six CAT strategies, using their own test characteristics such as sensitivity, specificity and costs. The seventh strategy evaluated is a direct laparoscopy strategy. We assumed that the drop out in the direct laparoscopy strategy is equal to the drop out in the other six strategies. Complications of HSG and laparoscopy have been taken into consideration in the model. After HSG, in % of cases, subfebrile morbidity occurs (Forsey et al., 1990). In our model, costs of antibiotic treatment have been included. In 1.5% of laparoscopies, surgical complications occur which usually require laparotomy (Chapron et al., 1998). The additional costs of these laparotomies have been incorporated in our model. Data decision model The data that we used as a basis for the decision analytic model are based on data obtained from an empirical study population of 315 patients, which has been described elsewhere (Land et al., 2003). The empirical study population was selected in the University Hospital Maastricht between 1992 and 2001, and only those patients for whom cryopreserved serum was available, and who underwent laparoscopy and tubal testing with methylene blue dye as part of their fertility work-up were included. Five different CAT tests were performed after thawing the cryopreserved sera of the participating patients. This empirical cohort is not representative for the population of subfertile women at intake for their fertility work-up, since, in daily practice, not all patients in whom a CAT test is performed as part of the initial fertility work-up will undergo evaluation of tubal function by HSG or laparoscopy. Therefore, an initial cohort had to be calculated as the source of this empirical population: the source population. Based on data of the empirical population, registration data of the Departments of Medical Microbiology and Obstetrics and Gynaecology of the University Hospital Maastricht, our own unpublished data and the literature (Forsey et al., 1990; Swart et al., 1995; Chapron et al., 1998; Mol et al., 1999; Veenemans and van der Linden, 2002), it was defined that the source population would have consisted of 1715 patients, of whom 600 would have had a positive CAT test at intake and 1115 a negative CAT test (Figure 2). From the empirical population, it is known that 22% of CAT-positive patients had a laparoscopy, and that in 78% no further evaluation of tubal function was done. The main reasons for no further testing are spontaneous pregnancies, arriving at diagnoses in which evaluation of tubal status is not indicated (e.g. severe male factor subfertility) and patient drop out before HSG or laparoscopy. Therefore, of the 600 CAT-positive patients in the source population, 132 would have had a laparoscopy and 468 would have had no additional testing after CAT. From the empirical population, it is known that of the CAT-negative patients, 32% have a HSG and that 68% drop out after CAT. After HSG, 52% of the patients in the empirical population had a laparoscopy and 48% had no further evaluation of tubal function after HSG. Therefore, of the 1115 patients in the source population with a negative CAT, 352 patients would have had a HSG and 763 would have dropped out after CAT. After HSG, 183 patients would have had a laparoscopy and 169 patients would have dropped out before laparoscopy. Consequently, 484 patients (132 þ 352) of the source population would have had further evaluation of tubal function by HSG and/or laparoscopy, and 315 patients would have had a laparoscopy. This group of 315 patients corresponds to the actual study population, all of whom had a laparoscopy (Figure 2). Cost calculation The costs of each CAT test, combination of tests (Med þ Ani), HSG, laparoscopy and complications of HSG and laparoscopy have been determined by detailed cost calculations. The direct costs such as personnel costs (time spent by laboratory personnel and physicians), the material costs (cost of kits and other test materials) and the capacity costs (use of machines) were calculated by practical costing studies, done at the University Hospital Maastricht. The hospital overhead was calculated by using 35% of the direct costs according to a guideline of Oostenbrink et al. (2000). Outcome measure For this study, we defined that patients with tubal pathology have no probability of spontaneous pregnancy, but only a pregnancy probability if they have treatment by IVF. Since all patients with a diagnosis of tubal pathology were assumed to be treated by IVF, the outcome measure IVF pregnancy was used. IVF pregnancy reflects the probability of pregnancy after three cycles of IVF in patients who have had evaluation of tubal function by HSG and/or laparoscopy, and in whom tubal pathology is diagnosed. Consequently, in patients who have no evaluation of tubal function (because of spontaneous pregnancy, other diagnoses for which tubal evaluation is not indicated or drop out), tubal pathology will not be Page 3 of 8

4 A.A.A.Fiddelers et al. Table I. Parameters of the decision analytic model Model parameter Baseline value Sensitivity range References Sensitivity pelisa Medac (%) Study population d Sensitivity ELISA Anilabsystems (%) Study population d Sensitivity Med þ Ani a (%) Study population d Sensitivity MIF Anilabsystems c (%) Study population d Sensitivity ELISA Savyon (%) Study population d Sensitivity MIF Biomerieux c (%) Study population d Specificity pelisa Medac (%) Study population d Specificity ELISA Anilabsystems (%) Study population d Specificity Med þ Ani a (%) Study population d Specificity MIF Anilabsystems c (%) Study population d Specificity ELISA Savyon (%) Study population d Specificity MIF Biomerieux c (%) Study population d Evaluation of tubal function b (% of patients) Study population d Prevalence of tubal pathology b (%) Study population d Cumulative pregnancy rate after three IVF cycles (%) Reduction in pregnancy probability after 1 year delay (%) diagnosed, referral for IVF will not take place, and the IVF pregnancy probability will be zero. The number of patients with tubal pathology is equal in all strategies. Therefore, the differences between the strategies are not based on different numbers of patients treated, but on the moment of treatment, which is related to test characteristics. A CAT test with high sensitivity will identify patients at high risk for tubal pathology at intake, and tubal factor subfertility can be diagnosed without delay. Early diagnosis of tubal pathology and immediate IVF treatment was defined to have a 50% cumulative pregnancy rate after three cycles (Mol et al., 2001; Granberg et al., 2003; National Collaborating Centre for Women s and Children s Health, 2004). A longer lasting diagnostic path, in the case where CAT is false negative, was estimated to reduce the cumulative pregnancy rate by 5% per year to 47.5% (National Collaborating Centre for Women s and Children s Health, 2004). In CAT-negative patients, if HSG is abnormal, patients will have a delay of 3 months before they have a Mol et al. (2001); Granberg et al. (2003); National Collaborating Centre for Women s and Children s Health (2004) National Collaborating Centre for Women s and Children s Health (2004) and expert opinion Delay after abnormal HSG (months) Clinical observation e Delay after normal HSG (months) Clinical observation e Tubal pathology after normal HSG b (%) Swart et al. (1995); Veenemans and van der Linden (2002) No tubal pathology after normal HSG b (%) Swart et al. (1995); Veenemans and van der Linden (2002) Laparoscopy after normal HSG b (%) Veenemans and van der Linden (2002) Laparoscopy after abnormal HSG b( %) Veenemans and van der Linden (2002) Complications of HSG b (%) Forsey et al. (1990) Complications of laparoscopy b (%) Chapron et al. (1998) Cost HSG b (e) University Hospital Maastricht f Cost laparoscopy b (e) University Hospital Maastricht f Cost complications after HSG b (e) University Hospital Maastricht f Cost complications after laparoscopy b (e) University Hospital Maastricht f Cost pelisa Medac (e) University Hospital Maastricht f Cost ELISA Anilabsystems (e) University Hospital Maastricht f Cost Med þ Ani a (e) University Hospital Maastricht f Cost MIF Anilabsystems c (e) University Hospital Maastricht f Cost ELISA Savyon (e) University Hospital Maastricht f Cost MIF Biomerieux c (e) University Hospital Maastricht f HSG ¼ hysterosalpingography. a Med þ Ani is considered positive when pelisa Medac and MIF Anilabsystems both have a positive test result. b Probabilities and costs are independent of the initial Chlamydia antibody testing (CAT) screening strategies. c MIF Biomerieux and MIF Anilabsystems were considered positive if the immunoglobulin G titre was $32 (Land et al., 2003). d Values reached by using data of the empirical and source population. e Delays after normal and abnormal HSG are obtained by clinical observations at the University Hospital Maastricht. f Costs have been determined by detailed cost calculation studies, based on the costs in the University Hospital Maastricht. laparoscopy, mainly because of waiting lists. If HSG seems normal in patients who have tubal pathology at laparoscopy, expectant management is defined for half a year, which means that these patients, if they do not conceive, have a delay of 9 months before laparoscopy is done. Thus, in our model, in patients with tubal pathology, postponing IVF treatment after normal or abnormal HSG will decrease cumulative pregnancy rates by 0.6 or 1.9%, respectively, per year. It can be shown that the reduction in cumulative pregnancy rate after a delay of t months equals (1 Pe) t/12, with Pe being the reduction in cumulative pregnancy rate after 12 months delay. The moment of IVF treatment does not influence the expected costs of the different strategies, and therefore costs of IVF are considered not to be relevant for our model. Statistical analysis The baseline values of the probabilities and costs were determined and incorporated into the decision tree by using the software Page 4 of 8

5 Cost-effectiveness in CAT Figure 2. Flowchart of the construction of the source population based on the empirical population. programme DATA version 3.5 (TreeAge software, Williamstown, MA). Analysing the decision tree, the path probabilities of each branch of the tree, the expected costs per patient, the IVF pregnancies and finally the expected costs per IVF pregnancy were given for each strategy. An incremental analysis was done for the seven strategies. This analysis is based on the costs per extra IVF pregnancy of a particular strategy, compared with the least expensive strategy. To test the robustness of these results, a sensitivity analysis is required (Briggs et al., 1994). This is the process of repeatedly analysing the tree by using different values for probability (e.g. values for specificity), utility (e.g. values of the outcome measure) (Krahn et al., 1997) and cost variables. In this study, univariate sensitivity analyses were performed by using the 95% confidence intervals of baseline values (sensitivity range), or a predetermined value range (Table I, column 2). The disadvantage of a univariate sensitivity analysis is that only one variable at the time can be changed. Therefore, a multivariate sensitivity analysis on accuracy numbers (sensitivity and specificity) was also done. In this analysis, a worst and best case analysis (extreme sensitivity and specificity values of the CAT tests) (Briggs et al., 1994) was performed. Finally, a threshold analysis was performed to determine if and when a variable changes. The threshold value represents the value of a variable above which another strategy is to be preferred. Results Baseline data All baseline data used in the model are presented in Table I. The sensitivity and specificity values of the five CAT tests and Med þ Ani, as observed in our empirical study population of 315 patients and source population of 1715 patients, were calculated using the outcome of laparoscopy as the reference standard. The sensitivity of the tests varies between 37.3% (ELISA Anilabsystems) and 72.5% (MIF Biomerieux). The specificity varies between 78.1% (MIF Biomerieux) and 95.8% (Med þ Ani). Costing studies showed that costs of CAT tests varied between e7.78 (ELISA Anilabsystems) and e13.05 (Med þ Ani) per serum sample tested. Results of the model Results, as presented in Table II, are expressed in four ways. Expected cost per patient per strategy. These are the expected costs per patient for CAT, HSG, laparoscopy and their complications, for all subfertile women at intake. For instance, in the pelisa Medac strategy in the decision analytic model, the expected cost per patient at intake is e This means that if couples seek help, it will cost the health care system e Expected effect per strategy. This is the probability of IVF pregnancy for the population seeking subfertility care. Thus, for instance, for the pelisa Medac strategy, an expected effect of 1.468% indicates that in a population of couples seeking subfertility care, 1468 IVF pregnancies can be expected. Cost per effect. These are the expected costs for any IVF pregnancy in patients diagnosed as having tubal pathology. For the pelisa Medac strategy, this outcome measure indicates that for each of the 1468 IVF pregnancies in couples, an investment of e is required. Incremental costs per effect. These are the costs per extra unit of effect, comparing different CAT strategies. In this study, it is the costs per extra IVF pregnancy of a particular strategy compared with the least expensive strategy. The incremental costs per effect show that two strategies (ELISA Anilabsystems and Med þ Ani) are ruled out as dominated Table II. Costs, effects, cost-effectiveness and the average costs per effect of the CAT screening strategies and the reference strategy (in which laparoscopy is done after intake) Strategy Expected cost per patient per strategy a (e) Expected effect per strategy b (%) Costs per effect c (e/effect) Incremental costs per effect d 1. pelisa Medac ELISA Anilabsystems Dominated 3. Med þ Ani e Dominated 4. MIF Anilabsystems ELISA Savyon Dominated 6. MIF Biomerieux Reference strategy a Cost per strategy includes cost of the CAT test, HSG and laparoscopy, and their complications. b Effect per strategy is the probability of IVF pregnancy for the population seeking subfertility care, compared with providing no IVF at all. c Cost per IVF pregnancy in a patient with tubal pathology. d Cost per extra IVF pregnancy in a patient with tubal pathology. e Med þ Ani is considered positive when pelisa Medac and MIF Anilabsystems both have a positive test result. Page 5 of 8

6 A.A.A.Fiddelers et al. Table IIIA. Univariate sensitivity analysis Baseline value Sensitivity range Threshold value Specificity MIF Anilabsystems (%) , MIF Anilabsystems Specificity ELISA Anilabsystems (%) , ELISA Anilabsystems Specificity pelisa Medac (%) ,90.2, MIF Anilabsystems Specificity Med þ Ani a (%) , Med þ Ani Sensitivity MIF Anilabsystems (%) , MIF Anilabsystems Sensitivity ELISA Anilabsystems (%) , ELISA Anilabsystems Sensitivity pelisa Medac (%) ,49.8, MIF Anilabsystems Sensitivity Med þ Ani a (%) , Med þ Ani Reduction in pregnancy probability , MIF Anilabsystems after 1 year delay (%) MIF Anilabsystems (cut-off titre , MIF Anilabsystems for positive test) Cost MIF Anilabsystems (e) ,12.26, MIF Anilabsystems Cost ELISA Anilabsystems (e) ,6.93, ELISA Anilabsystems Cost pelisa Medac (e) , MIF Anilabsystems Cost Med þ Ani a (e) ,12.28, Med þ Ani Cost laparoscopy (e) , MIF Anilabsystems, and.1291 Med þ Ani Only characteristics with a threshold value are given. a Med þ Ani is considered positive when pelisa Medac and MIF Anilabsystems both have a positive test result. by pelisa Medac, since they are more expensive and less effective. In the MIF Anilabsystems strategy, 1469 IVF pregnancies are expected in couples, which is an extra effect of one pregnancy compared with 1468 IVF pregnancies in the pelisa Medac strategy. However, for this extra pregnancy, an investment of e is required. The ELISA Savyon strategy is ruled out as dominated by the MIF Anilabsystems strategy, since it is more expensive and less effective. In the MIF Biomerieux strategy, 1473 IVF pregnancies are expected in couples, which is an extra four compared with MIF Anilabsystems. For each of these four pregnancies, an investment of e is required (Table II). In the reference strategy, in which the patients will have a laparoscopy immediately after intake, 1480 IVF pregnancies are expected in couples, which is an extra seven pregnancies compared with the MIF Biomerieux strategy. For each of these seven pregnancies, an investment of e is required. Sensitivity results Results of both the univariate and multivariate sensitivity analyses are given in Table III. In Table IIIA, the baseline values and corresponding sensitivity ranges are given for all characteristics in which a threshold value is exceeded. It turned out that the pelisa Medac strategy remained the most cost-effective screening strategy, even when baseline values were altered: when the prevalence of tubal pathology was 5 40%, when 10, 20 or 40 serum samples were tested simultaneously, when the cumulative pregnancy rate after three cycles of IVF was 20 70%, and when the delay after normal and abnormal HSG was 6 12 and 0 6 months respectively. However, the MIF Anilabsystems strategy became the most cost-effective CAT screening strategy if the annual decrease in cumulative pregnancy rate was. 19.0%, in the case where a cut-off value of 64 instead of 32 was used for a positive test, if the cost of the MIF Anilabsystems Page 6 of 8 Table IIIB. Multivariate sensitivity analysis Lowest specificity in all tests Highest specificity in all tests Lowest sensitivity in all tests Highest sensitivity in all tests Lowest specificity and sensitivity in all tests Highest specificity and sensitivity in all tests Preferred strategy MIF Anilabsystems pelisa Medac pelisa Medac pelisa Medac MIF Anilabsystems pelisa Medac test became, e12.26 or if the cost of pelisa Medac became. e8.90. MIF Anilabsystems was also most costeffective if the specificity or sensitivity of MIF Anilabsystems increased, or if the specificity or sensitivity of pelisa Medac decreased. In the case where costs of a laparoscopy were between e1068 and e1291, MIF Anilabsystems was also the most cost-effective CAT screening strategy, and in the case where costs of a laparoscopy were higher than e1291, Med þ Ani was the most cost-effective screening strategy. Med þ Ani also became the most cost-effective strategy if the specificity and sensitivity of Med þ Ani became.96.7 and 59.1%, respectively, or in the case where the costs of Med þ Ani were,e In the multivariate sensitivity analysis, pelisa Medac remained the most cost-effective CAT screening strategy in four of the six strategies (Table IIIB) that were tested. However, when the specificity of all tests had the lowest value (Table I), or when the specificity as well as the sensitivity of all tests had the lowest value (Table I), MIF Anilabsystems became most cost-effective. Discussion In this study, we have evaluated six CAT strategies (five CAT tests and one combination of tests), and a reference strategy in which the patients had a laparoscopy immediately after intake, from a health-economic perspective. It was

7 Cost-effectiveness in CAT defined that tubal disease is of equal severity in all patients, and that these patients have no chance of pregnancy if IVF is not done. Since our outcome measure is IVF pregnancy, and the above-mentioned definitions are equal for all strategies, they do not influence the outcome of our model. Comparison of the five CAT strategies in the source population showed that pelisa Medac is the most cost-effective strategy, followed by MIF Anilabsystems. The differences between the strategies were small, however, mainly because 72% of the patients had no further evaluation of tubal function after CAT (because of spontaneous pregnancy, diagnosis for which tubal status is not relevant or no more visits to the hospital), and had no additional costs. By performing sensitivity analyses on the baseline values of the source population, it was possible to evaluate in which circumstances MIF Anilabsystems became more cost-effective than pelisa Medac. The baseline value of the annual decrease in cumulative pregnancy rate after postponement of IVF was considered to be 5%. The sensitivity analyses showed a threshold value of 19% decrease in cumulative pregnancy rate (from 50 to 40.5%), above which the MIF Anilabsystems strategy was to be preferred. In clinical practice, however, it is unlikely that the cumulative pregnancy rate after three IVF cycles will decrease more than 19% if treatment is postponed for 12 months. MIF Anilabsystems also became most costeffective if the cut-off titre of a positive test was 64 instead of 32. Although a cut-off titre of 32 is recommended by the manufacturer, a cut-off titre of 64 might be preferred from a clinical point of view, if one prefers to have higher specificity and less false-positive test results (Land et al., 1998). Finally, if specific test characteristics such as sensitivity, specificity and costs of pelisa Medac and MIF Anilabsystems changed, MIF Anilabsystems became the most costeffective screening strategy. pelisa Medac remained the most cost-effective screening strategy in the case where the prevalence of tubal pathology was between 5 and 40%. In a primary, secondary and tertiary care population, the prevalence of tubaperitoneal disorders has been estimated to be 11% (Snick et al., 1997; Evers, 2002), 20% (Hull et al., 1985; Evers, 2002) and 30% (Collins et al., 1995; Evers, 2002), respectively. The probability that the prevalence of tubal pathology will exceed 40% is unlikely, even in a tertiary care population. Cost prices were calculated for testing 10, 20 and 40 serum samples simultaneously in one laboratory session. The baseline number of samples tested simultaneously was 20, based on the clinical experience in the University Hospital Maastricht, where, 300 CAT tests are done yearly, and CAT test results are available within 2 3 weeks. In a smaller fertility centre (,150 CAT tests per year), it may be preferred to test 10 samples simultaneously, and in a larger fertility centre (.600 CAT tests per year) the preferred number might be 40. In cases where 10, 20 or 40 serum samples were tested simultaneously, pelisa Medac remained most cost-effective. pelisa Medac was also the most cost-effective screening strategy in the case where the cumulative pregnancy rate after three IVF cycles was between 20 and 70%. These values are extremes, and it is unlikely that the cumulative pregnancy rate will be lower than 20% or higher than 70%. Apart from the five CAT tests, a combination of two CAT tests (Med þ Ani) was evaluated. For this combination of tests, pelisa Medac was chosen as a first test, since it can be automated, and the more laborious MIF Anilabsystems was chosen as the second test, because of its diagnostic accuracy. It was concluded that if pelisa Medac is performed on all samples, and only those samples with positive test results (i.e. 20% of all samples) are retested with MIF Anilabsystems, the predictive value of the set is comparable with the predictive value of MIF Anilabsystems as a single test (Land et al., 2003). Performing Med þ Ani might be preferred from a clinical point of view, since it is less laborious than testing all samples with MIF Anilabsystems, and a titre is available for positive test results. A titre gives additional information, since the height of a titre correlates with the risk for tubal pathology (Akande et al., 2003). From a cost-effectiveness point of view, however, Med þ Ani turned out to be less cost-effective than pelisa Medac or MIF Anilabsystems as a single test. Only in cases where the sensitivity, specificity or cost of the Med þ Ani test were changed did Med þ Ani become the most cost-effective strategy. Finally, the reference strategy, in which the patients had a laparoscopy after intake, was found to have the greatest effect (1.48% IVF pregnancies). From a cost-effectiveness point of view, however, the reference strategy turned out to be inferior to the pelisa Medac strategy, because of the high cost of the high number of laparoscopies in the reference strategy. Although the different strategies were evaluated in the University Hospital Maastricht, the results of this study can be used in other settings. The sensitivity analyses, based on clinical, economic and hospital characteristics, show that the model results are very robust, indicating that pelisa Medac is the most cost-effective screening strategy in most settings. In Table III, we have given the threshold values, which enables one to compare the baseline values used in this study with one s own specific situation. Before introducing a CAT screening strategy into clinical practice, one should take the baseline values in that particular setting into account, and not take decisions on differences in diagnostic accuracy of the CAT tests only. In the methodology of the economic evaluation of diagnostic tests, it is important to evaluate the effects and costs of a complete strategy, instead of only individual cost prices of the tests and separate clinical outcome measures. In conclusion, only small differences were found between the CAT screening strategies for tubal factor subfertility evaluated, concerning cost-effectiveness. pelisa Medac was the most cost-effective strategy, followed by MIF Anilabsystems. A combination of tests (pelisa Medac as the first test and retesting of all positive serum samples with MIF Anilabsystems) did not improve the cost-effectiveness of pelisa Medac or MIF Anilabsystems as single tests. Acknowledgements G.Grauls, W.Mullers (Department of Medical Microbiology, University Hospital Maastricht) are gratefully acknowledged for Page 7 of 8

8 A.A.A.Fiddelers et al. providing information for the cost analysis. The study was supported by a grant from ZON MW ( ), The Netherlands. References Akande VA, Hunt LP, Cahill DJ, Caul EO, Ford WCL and Jenkins JM (2003) Tubal damage in infertile women: prediction using chlamydia serology. Hum Reprod 18, Briggs A, Sculpher M and Buxton M (1994) Uncertainty in the economic evaluation of health care technologies: the role of sensitivity analysis. Health Econ 3, Chapron C, Querleu D, Bruhat MA, Madelenat P, Fernandez H, Pierre F and Dubuisson JB (1998) Surgical complications of diagnostic and operative gynaecological laparoscopy: a series of cases. Hum Reprod 13, Collins JA, Burrows EA and Wilan AR (1995) The prognosis for live birth among untreated infertile couples. Fertil Steril 64, Evers JLH (2002) Female subfertility. Lancet 360, Forsey J, Caul E, Paul ID and Hull MGR (1990) Chlamydia trachomatis, tubal disease and the incidence of symptomatic and asymptomatic infection following hysterosalpingography. Hum Reprod 5, Granberg M, Strandell A, Thornburn J, Daya S and Wikland M (2003) Economic evaluation of infertility treatment for tubal disease. J Assist Reprod Genet 20, Hull M, Glazener CM, Kelly NJ, Conway DI, Foster PA, Hinton RA, Coulson C, Lambert PA, Watt EM and Desai KM (1985) Population study of causes, treatment, and outcome of infertility. Br Med J 291, Krahn MD, Naglie G, Naimark D, Redelmeier DA and Detsky AS (1997) Primer on medical decision analysis: part 4 analysing the model and interpreting the results. Med Decis Making 17, Land JA, Evers JLH and Goossens VJ (1998) How to use Chlamydia antibody testing in subfertility patients. Hum Reprod 13, Land JA, Gijsen AP, Kessels AGH, Slobbe MEP and Bruggeman CA (2003) Performance of five serological chlamydia antibody tests in subfertile women. Hum Reprod 18, Mol BWJ, Collins JA, Burrows EA, Veen van der F and Bossuyt PMM (1999) Comparison of hysterosalpingography and laparoscopy in predicting fertility outcome. Hum Reprod 14, Mol BWJ, Collins JA, Veen van der F and Bossuyt PMM (2001) Cost-effectiveness of hysterosalpingography, laparoscopy, and Chlamydia antibody testing in subfertile couples. Fertil Steril 75, National Collaborating Centre for Women s and Children s Health (2004) Commissioned by the National Institute for Clinical Excellence. Fertility assessment and treatment for people with fertility problems, Clinical Guidelines, pp Accessible at Public/Fertility_full.pdf Oostenbrink JB, Koopmanschap MA and Rutten FFH (2000) Handleiding voor kostenonde rzoek, methoden en richtlijnprijzen voor economische evaluaties in de gezondheidszorg. College voor zorgverzekeringen, Amstelveen. Snick HK, Snick TS, Evers JLH and Collins JA (1997) The spontaneous pregnancy prognosis in untreated subfertile couples: the Walcheren primary care study. Hum Reprod 12, Swart P, Mol BWJ, Veen van der F, Beurden van M, Redekop WK and Bossuyt PMM (1995) The accuracy of hysterosalpingography in the diagnosis of tubal pathology: a meta-analysis. Fertil Steril 64, Veenemans L and van der Linden PJQ (2002) The value of Chlamydia trachomatis antibody testing in predicting tubal factor infertility. Hum Reprod 17, Weinstein MC, O Brien B, Hornberger J, Jackson J, Johannesson M, McCabe C and Luce B (2003) Principles of good practice for decision analytic modelling in health-care evaluation: report of the ISPOR task force on good research practices modelling studies. Value Health 6, Submitted on December 18, 2003; resubmitted on August 5, 2004; accepted on October 15, 2004 Page 8 of 8