Objectives To evaluate selected molecular tests in diagnosis and screening of cytomegalovirus (CMV) infection in immunosuppressed patients. Design Clinical and cost-effectiveness were assessed through a prospective two-stage trial of CMV screening regimes in a routine service setting. Different molecular test results were fed back to clinicians in each stage, plus antigenaemia results. The technical performance of the molecular methods was assessed through an independent masked comparison of each molecular test against the established (antigenaemia) test. Scientists performing a particular test were blind to the other test results for that sample. Diagnostic and therapeutic impact were recorded prospectively for all tests, to include any effect on diagnostic certainty, changes to CMV therapy and any other reported impact on patient management. The cost of each test was estimated under different laboratory conditions. Prospective patients undergoing CMV screening were compared with consecutive historical controls in the same unit. Towards the end of the study, a survey of all UK virology laboratories was undertaken to identify current CMV screening practice and test preferences. In addition, all UK renal transplant surgeons and haematology transplant centres were surveyed in order to identify current clinical practice and perceptions of the benefits of CMV screening. Setting Study patients were recruited from University Hospital Wales (UHW), Cardiff. Staff in the Cardiff Public Health Laboratory Service virology laboratory performed the tests. Participants A consecutive series of transplant patients was recruited to the prospective study over a 42-month period, totalling 98 renal and 140 haematology patients. A consecutive series of historical controls was identified, with 199 renal and 136 haematology patients who underwent transplants in the UHW during the 29 months prior to the prospective CMV screening trial. Interventions A predefined CMV screening protocol was applied to all patients in the prospective trial. Renal patients were tested every 4 weeks until 16 weeks post-transplant (five tests in total). Haematology patients were tested every 2 weeks until 12 weeks post-transplant, and then every 4 weeks until 24 weeks (10 tests in total). The assays used for CMV screening were as follows: non-molecular test, (1) pp65 antigenaemia assay; molecular tests, semi-quantitative in-house polymerase chain reaction (PCR), (2) single-round (PCR1) and (3) two-round, nested (PCR2); and qualitative commercial tests, (4) Roche Amplicor Assay (Amplicor) and (5) pp67 NASBA assay (NASBA). Main outcome measures Test failure rates, sensitivity/specificity values and positive predictive value (PPV) and negative predictive value (NPV) were measured for each assay. The laboratory cost of undertaking various CMV tests was measured and other NHS costs associated with false-positive or false-negative test results were estimated. The likelihood of CMV disease and the likely impact of positive or negative test result on therapy and further investigations were recorded. On receipt of the test result, interim outcome measures were recorded to include the impact of test result on diagnostic certainty, changes to planned patient management (e.g. therapy, investigations) and perceived benefit. All definitive diagnoses of CMV disease, prescribing of CMV therapy and interim patient outcome at the end of the screening period were recorded. Results In haematology and renal transplant patients, all tests had a similar NPV (0.976--0.997 and 0.935--0.995, respectively) when used in CMV screening. PCR1 is the least expensive molecular test (7.80-13.70 UK pounds). Commercial tests, NASBA and Amplicor, are both more expensive (22.50-34.70 UK pounds NASBA; 23.20-29.20 UK pounds Amplicor). Antigenaemia costs 12.50-27.40 pounds depending on staff grade and batch size. Quantitative PCR (COBAS) is the most expensive at around 50 UK pounds per sample. No clear link between screening test results and CMV prescribing was detected; clinicians appear to consider screening results in the context of other factors. There was no evidence that the introduction of CMV screening led to reductions in CMV deaths or improved transplant success rates. For cost per positive test result, PCR1 was the most cost-effective screening test on this indicator (renal patients 116 UK pounds per true positive, haematology patients 518 pounds). Antigenaemia was the least cost-effective screening test (renal patients 643 UK pounds per true positive, haematology patients 2475 pounds). Cost-effectiveness analysis and cost per "beneficial result" (as judged by clinicians) confirmed that PCR1 remained the most cost-effective test. Modelling outputs for targeted screening protocols also supported this. Conclusions The study findings offer some evidence that a CMV screening regime is more cost-effective than diagnostic testing alone, based on the cost per true positive detected and interim outcome such as changes in patient management. However, the study was unable to demonstrate any benefits in terms of longer term patient outcomes. If CMV screening is introduced, the use of antigenaemia pp65 is clearly less cost-effective than the use of molecular tests. The study identified the optimum test for CMV screening as an in-house molecular test (single-round PCR test). This test was less costly to perform and also resulted in lower costs linked to false positives and negatives than other tests. The in-house, semi-quantitative test was two to three times more cost-effective than the commercial molecular tests assessed; however changes to European Union legislation may mean that it may not be feasible to use in-house tests. The use of targeted screening (limiting CMV screening to high-risk transplants) as opposed to universal screening offers a significant improvement in the cost-effectiveness ratio for haematology transplant patients, but has limited impact in the case of renal transplants. Economic analyses could be expanded to model the cost-effectiveness of more frequent screening tests (as reported nationally), and screening in other "at risk" groups. Subgroup specific disease groups should be investigated across a larger population to allow more accurate modelling of the impact of CMV screening on disease progression. Further studies of CMV screening programmes should address a range of outcome measures, including patient outcomes.