This paper has reported a cost-effectiveness analysis comparing aCGH with karyotyping for detecting genomic imbalances that diagnose ILD. The average cost of aCGH was £442 per single (patient) sample and the average cost of karyotyping was £117 per sample. The majority of the cost-difference was accounted for by the array cost. Thus, from a single test perspective, aCGH is more expensive than karyotyping, explaining, in part, the hesitation by commissioners to fund aCGH in NHS diagnostic laboratories.
In reality, the situation is more complex because information regarding subsequent tests for genomic imbalance must be considered before the true cost-effectiveness can emerge. We have shown that the overall cost per diagnosis of the karyotyping route, including a single multi-telomere FISH assay (£4,957) is more expensive than that of the aCGH route (£3,118) that yields 10% more diagnoses. However, if the less conservative yield of 15% more diagnoses is correct, then the aCGH cost reduces to £2,440 per diagnosis, a figure more comparable to karyotyping plus the alternative multi-telomere assay, MLPA (£2,129 per diagnosis). Importantly, 92% of cases tested by karyotyping and a multi-telomere assay will require further tests for an eventual diagnosis. By contrast, the aCGH route, which effectively represents karyotyping, multi-telomere testing and not one, but ∼34,000 interstitial FISH tests as well as assaying the entire human genome at higher resolution is unlikely to require further genome-wide tests for genome imbalance.
Stand-alone karyotyping is the cheapest test when considered per diagnosis (£2,067), but this is at the sacrifice of missing ∼75% (12/18) diagnoses achievable by aCGH (Fig. 4). Thus, the crux of the aCGH versus karyotyping argument in ILD comes down to diagnostic capability versus cost; how much is it acceptable to spend and how many diagnoses is it acceptable to miss? aCGH clearly offers the greatest diagnostic capability, providing 10–15% more diagnoses over all other available tests.
One limitation of our study is that we do not know the full magnitude of the cost for additional follow-up tests after karyotyping. However, we do know that such costs would rapidly escalate and even then the majority of diagnoses achievable by aCGH would be missed; clinically relevant genomic imbalances found by genome-wide aCGH are rarely recurrent and therefore targeted approaches are unlikely to improve diagnostic yield (Veltman and de Vries 2006). Even if multi-telomere FISH or multi-telomere MLPA are not the tests chosen following karyotyping, costing in only five targeted tests for genomic imbalance at £100/test for every sample with a normal karyotype would raise the overall cost of testing 100 patients to ∼£59,402 (compared with £56,130 for aCGH) and offer negligible improvement in resolution overall. Thus, even without precise costing of follow-up tests, our results suggest that aCGH is the most cost-effective testing strategy in the long-term for testing ILD patients.
A further study limitation is the use of a simple outcome measure, diagnosis, rather than the more usual cost per quality adjusted life year promoted by the National Institute for Health and Clinical Excellence. However, describing and valuing health states in children is difficult and well documented (Petrou 2003). Current methodological work in health economics may give useable health states for children, but is unlikely to be suited for those with LD.
In this study, one scenario that we were unable to cost either for karyotyping or array CGH was that of further researching apparently inherited genome imbalances for which clinical relevance cannot be excluded. Such cases may reflect benign variants or may cause disease through unmasking recessive mutations, through variable penetrance or through imprinting, for example. Currently these account for up to 32% cases (see Fig. 4) and therefore follow-up tests such as sequencing would be prohibitively expensive (not all inherited imbalances are small). However, as more and more studies are performed and more data regarding benign/relevant genome imbalances and genotype/phenotype correlations are added to databases, it is anticipated that the clinical relevance of a significant number of these cases will be defined earlier, thereby minimising the need for parental testing or additional follow up tests. This in turn will lead to reduced aCGH testing costs and costs arising from doctor/counsellor time taken to discuss uncertain results. Furthermore, it may become possible to reduce costs more by employing better defined clinical ‘gatekeeping’ criteria to help clinical geneticists direct testing (thereby minimising total tests done). Databases such as The Database of Genomic Variants (http://projects.tcag.ca/variation/), The Human Structural Variation database, (http://humanparalogy.gs.washington.edu/structuralvariation/), ECARUCA (http://agserver01.azn.nl:8080/ecaruca/ecaruca.jsp) and the DatabasE of Chromosomal Imbalance and Phenotype in Humans using Ensembl Resources (DECIPHER http://www.sanger.ac.uk/PostGenomics/decipher/) have all been designed to expedite these processes. Another possibility may be to use targeted arrays as an initial screening test for paediatrician referrals, though currently these offer no cost advantage over genome-wide arrays and utility will depend on a high diagnostic pick-up rate.
In the meantime, it will continue to be important for families to be counselled in possible outcomes before taking up the test, for both parental samples to be available for testing and for any outgoing laboratory reports to be carefully designed with clearly defined results e.g. array batch, controls used, imbalances found, confirmatory method and database search results that might help inform clinical relevance.
In conclusion, our study demonstrates that in the context of ILD, genome-wide aCGH is viable for NHS diagnostic use. Indeed, where possible, it may be appropriate to replace karyotyping with aCGH as the first-line test for genomic imbalance in ILD. If needed, samples that are normal by aCGH could then be karyotyped in order to identify truly balanced rearrangements or further characterise genome imbalances. Additionally, aCGH is expected to be useful for clarifying previous equivocal karyotyping results (e.g. enabling definition of a cryptic translocation in a family where only one of two unbalanced outcomes is cytogenetically visible).
In the future, improved diagnostic yields of aCGH and reduced follow-up tests will lower the costs of clinical follow-up and additional investigations. Advances in technology will also reduce costs (e.g. automation, increased probe density, multi-sample and cheaper array production and hybridisation methods) and software improvements may reduce analysis time.
Finally, it is important to note that potential applications of microarray technology extend beyond the genetic diagnosis of ILD to include a range of other conditions with suspected genome imbalance and/or aberrant gene expression e.g. haematological malignancies, colorectal cancer and other fields including oncology, immunology, neurology and pathology. The UK Department of Health is keen for the NHS to adopt new technologies (Department of Health 2003), yet commissioners are unable to endorse implementation without considering the clinical utility and economic implications of technology adoption. Our costing, with the results divided into different testing stages provides a framework for costing array implementation in different settings. Not least, it is intended that the study will be useful for healthcare providers faced with the decision of introducing aCGH testing into NHS laboratories before the availability of substantial effectiveness information.