Cost-effectiveness of whole-exome sequencing in progressive neurological disorders of children

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


Introduction
The Online Mendelian Inheritance of Man database recognizes over 4000 clinical synopses with neurological involvement, out of which over 3000 with a confirmed molecular basis. The genetic complexity is significant: the same disease can be caused by variants in several different genes, for example Leigh disease (1) can be due to variants in more than 75 different genes, and even a single variant may cause variable symptoms in different patients such as in X-linked adrenoleukodystrophy (2). The development of next-generation sequencing (NGS) methods for the human genome dramatically improved the diagnostic approaches (3), sometimes providing targeted approaches for treatment (4). In addition, early genetic diagnosis provides tools for counseling (5) and guidance for reproductive planning (6).
Whole-exome sequencing (WES) provides data from protein-coding genes of the genome (7). Sequence analysis of all the genes, instead of single candidate genes, reduces the time required for identification of gene defects exponentially, and enables discovery of novel disease-causing genes.
Method development has also progressively reduced the costs of the analysis, making it feasible for routine diagnostics (7). However, health care providers still may consider NGS methods expensive for clinical practice (8), which calls for cost-effectiveness studies to support decision-making.
The improved and efficient diagnostic yield as a consequence of NGS-analysis might result in better health outcomes or more efficient use of health care services (3). A recent meta-analysis found that the pooled diagnostic utility, meaning the rate of definitive diagnoses achieved, for WES was 36% in children with suspected genetic diseases (9). However, current studies on cost-effectiveness and economic outcomes of WES are limited to few studies (3). In a diagnostic work up to reach a diagnosis, the largest cost drivers are found to be the costs of genetic tests and costs of WES (10), but J o u r n a l P r e -p r o o f 4 if WES is used as a near first-line test in a selected cohort of patients, overall budget increase may not be required.
Here, we report diagnostic utility and cost-effectiveness of WES as a routine diagnostic tool in progressive neurological disorders of children.

Study population and data collection
Patients with infantile-onset severe neurological disease or childhood-onset progressive neurological disorder were prospectively recruited to the WES study at Children's Hospital at Helsinki University Hospital, a tertiary care hospital, during the years 2016-2018. Exclusion criteria were nonprogressive intellectual disability or autism spectrum disorder, family history of a known genetic disorder, or otherwise clinically identifiable genetic disorder. In total, 48 non-consanguineous pediatric patients underwent the singleton WES as a routine diagnostic test ("WES group"). Short clinical descriptions, including genetic testing before recruitment to the study, are presented in Appendix 1.
The control group included 49 children, who suffered from similar disorders as the WES group, but had often undergone some conventional diagnostic tests, such as metabolic investigations, Sanger sequencing, NGS gene panels, and karyotyping, but not WES.

Patient cohort
The median age of study subjects was 2.4/0.9 years (range 0-16/0-17 years) at the beginning of the first diagnostic visit, and 63%/51% were male among WES group and control group, respectively (Table 1). There were statistically significant differences with residence district between WES and control groups (p<0.001). The collection of data was more comprehensive concerning clinical visits for the control group, who were more likely to live in our hospital district.
Both patient cohorts consisted of heterogeneous phenotypes, with the majority affected by encephalopathy (54%/61%) and neuromuscular disorders (31%/29%). Additionally, patients undergoing WES were further characterized by how many years of investigations they had had in Children's Hospital before being recruited to this study. Patients getting WES during their first year of investigations (48%) constitute our early WES patient group.
The use of health care services related to the diagnostic path of study participants was gathered retrospectively from patient records. The data consisted of all diagnostic health care visits and investigations including hospitalizations, clinical visits, laboratory tests, imaging, and genetic testing.
Only events considered relevant for the diagnostic process were included, and the events were reviewed individually by study physicians. In addition, gender and age at the first visit in the hospital, the date of diagnosis and timing of WES along the diagnostic path were recorded.

Whole-exome sequencing
WES was performed using exome capture by Agilent SureSelect V5 kit and Illumina MiSeq sequencing at the Finnish Institute of Molecular Medicine (FIMM) as described in Sainio et al. (11).
A customized exome analysis pipeline (11) was used to analyse the genetic data , and the gene J o u r n a l P r e -p r o o f 6 findings were compared to phenotype with study physicians and thus to reach a definitive diagnosis.
Sanger sequencing was used as an additional independent method to confirm findings and segregation in patient and family samples.

Diagnostic yield
Effectiveness outcome was diagnostic yield, which was calculated as a proportion of definitive diagnoses to the total number of patients in both groups. It was also calculated separately for the different time-subgroups.

Cost-effectiveness analysis
Economic analysis was performed from health care provider (hospital) perspective. Costs of laboratory tests, imaging and genetic tests were obtained from the hospital (Hospital District of Helsinki and Uusimaa, HUS) and diagnostic laboratory documentation (tests performed outside the hospital). Clinical visit costs were defined according to the hospital district's outpatient product costs for specialized somatic health care visits. The costs for hospitalization periods were determined from the estimates by Finnish National Institute for Health and Welfare for the unit costs of social and health care in 2011 (12). The costs of non-WES diagnostic tests in 2019 were converted to 2018 prices in euros using the national health and social care price index by the Association of Finnish Local and Regional Authorities (13) and currency converter (14), or the current price was used, e.g. Baseline characteristics of children in WES group and control group were compared by cross tabulation and chi-square and Fisher's exact tests. Continuous variables were analysed by Wilcoxon rank-sum test. Mean diagnostic costs per patient were calculated with standard deviations, medians and 95% confidence intervals (CI). In addition, mean costs per diagnosis were calculated by dividing total costs by the total number of diagnoses in the groups.
In the cost-effectiveness analysis incremental cost-effectiveness ratio (ICER = ∆Costs/(∆Diagnostic yield)) per additional diagnosis was calculated by dividing the difference in mean costs per patient between WES and control groups by the difference in diagnostic yield (diagnosis rate) between the groups. Mean differences of the total costs per patient between WES and control groups were analyzed using Wilcoxon rank-sum test. Bootstrapping simulation with 1,000 replications was used to estimate the uncertainty of cost-effectiveness analysis. Bootstrapping resamples the data with replacement to building an empirical estimate e.g. of the mean costs or ICER of the sampling data ((15), p. 299). The early-WES subgroup was analysed separately. Since information on patient clinical visits was not comprehensive for the exome group and thus more favourable for the group, the additional analyses were done without clinical visit costs, with a third analysis with all study subjects.
Statistical significance was set at p-value <0.05. All analyses were made using Stata 15.1 (Stata, College Station, TX) except for bootstrap simulations, which were performed in Microsoft Excel.

Ethics
Ethical approval for the study was granted by the coordinating ethical committee of The Hospital District of Helsinki and Uusimaa. Informed consents were gathered from the parents of child participants.

Costs and cost-effectiveness
Mean cost per diagnosis was lower in the WES group (25,433€ vs. 40,467€). Mean costs per patient were 9,537€ (range 3,387-27,308€) in the WES group and 9,910€ (2,088-23,310€) in the control group (Table 2). WES yielded more definitive diagnoses with slightly lower costs and could therefore be considered dominant over standard care. However, the cost difference was not statistically significant (p=0.5302).
Main cost drivers were genetic tests (32%, including the price of WES) and clinical visits (26%) in the WES group. In the control group, the largest cost drivers were clinical visits (33%) and genetic tests (26%). Control patients had on the average 3.0 genetic tests (range 0-7), whereas patients in the WES group had had 1.4 tests (range 0-8) before inclusion to the study. For patients that had WES done early after manifestation, the mean was 0.6 (range 0-3). Prior to the study, 40% of the patients in the WES group had been tested for chromosomal anomalies, 42% had at least one gene analysed by Sanger sequencing, and 19% had a gene panel analysis done (corresponding to 73%, 65%, and 31% for controls). In the WES group, the mean number of clinical visits were 4.9 (ranging 1-17) and in the control group, 6.5 visits (1-13) (p<0.01).
J o u r n a l P r e -p r o o f 9 Additional analyses (Table 3) were done without clinical visit costs. When only early WES-patients were included in the treatment group, WES was dominant, meaning potentially cost-effective, as WES had a greater diagnostic yield with lower costs (mean cost per diagnosis 5,502€ vs. 6,674€).
The cost difference was not statistically significant (p=0.3309).
A third analysis showed that mean costs per patient were slightly higher in the WES group than in the control group if clinical visit costs were not included. Still, cost-effectiveness analysis showed that WES yielded the incremental cost of 2,847€ per one additional diagnosed patient.

Discussion
This study evaluates the diagnostic utility and cost-effectiveness of WES as a routine diagnostic tool in pediatric patients with progressive neurological disorders. Our results show that WES provides better diagnostic yield (37.5 vs. 24.5%) compared to conventional diagnostic path utilizing clinical diagnostic means complemented with gene panel testing. First-year "early-WES" was clearly most successful (43%). Our diagnostic yield in the WES group is in line with a recently published metaanalysis of children with suspected genetic diseases (9). Considering patients that were recruited to the study even after three years of prior investigations (31%), who had been examined with a large set of standard diagnostic tools, WES resulted in previously unachievable diagnoses for four out of fifteen patients.
We chose to collect full costs of both WES and conventional diagnostic path, to elucidate the full costs related to the examinations. Previous studies have not used a similar control group of patients (16). Many of the previous studies were modeled with diagnostic scenarios in the same study cohort (17)(18)(19)(20)(21) or using a hypothetical WES trajectory (22). In addition, only a few studies were conducted in Europe (10,22). Also, previous studies mainly investigated cost-effectiveness of WES in pediatric J o u r n a l P r e -p r o o f 10 patients with any suspected monogenic disorders (6,(19)(20)(21) or with specific disorders, such as epilepsy (17) or muscle disorders (18). The finding that clinical visits and genetic tests were the main drivers of costs in both study groups are in line with previous studies, in pediatric cohorts (10) and mixed cohorts of children and adults (23) with complex neurological problems. Most of the previous studies have reached incremental cost savings per additional diagnosis when WES was used as a firstline test (18,19,21). In a population-based study by Howell et al. (17) WES also yielded cost savings per additional diagnosis only when WES was targeted early and metabolic testing was limited compared to standard care without WES in patients with severe infantile epilepsies. In other pathways, including metabolic testing, repeated magnetic resonance imaging or skin and muscle biopsies before WES, the incremental cost per additional diagnosis was $ 3,250-8,559. One of the few European studies (22)  Previously, health status or quality of life have been discussed not to necessary be the only outcome measures in health economic evaluations of genetic testing, as genetic information itself is valued and can influence one's ability to make an informed decision (24). However, there is no single threshold for interpreting the ICER result of our study, so cost-effectiveness depends on the payer's willingness to pay for one additional diagnosis. Further studies remain to be performed to estimate such willingness to pay and to outline whether payers are eager to reimburse on such outcome measures.
The importance of genetic testing cannot be over-emphasized, as it provides considerable personal benefit by ending diagnostic examinations, offering exact genetic diagnosis and counseling, providing prognosis, and sometimes directing therapy decisions.
J o u r n a l P r e -p r o o f 11 The strength of this study is a prospective cohort study design, which allowed investigation of WES as a routine diagnostic tool. In addition, the study includes a retrospectively collected control group of patients who underwent traditional diagnostic tests. However, this study also has limitations. First, living district may present a selection bias and second, the sample size is relatively small. Diagnostic yield in different studies varies based on how well the original patient population was preselected, and directly affects cost-effectiveness. The diagnostic yield could have increased if trio-analysis had been implemented. However, in WES-studies often yields from 30-40% are achieved, pointing to the value of the diagnostic tool. In a benchmark meta-study of children with heterogeneous suspected genetic conditions, diagnostic yield for singleton-WES was found to be 26.5% (95% CI: 12.9-42.9) across studies, suggesting this range of yield to be characteristic for child manifestations (25). Our study's sample size, limited due to financial capacity to do WES, potentially also widened the confidence intervals of the bootstrapped results. Third, as the purpose of the study was to clarify the costs of early WES analysis, this sample is a selected subsample. Last, infantile encephalopathies and progressive neurological disorders of childhood are a clinically heterogeneous group of patients, and tour-de-force of examinations are often initiated to gain a specific diagnosis increasing the non-WES diagnostic costs. Similar cost-effectiveness studies for different kinds of patient groups would be informative.
The results are highly interesting, as our study group was clinically broadly definedprogressive neurological disorder of childhoodand genetically heterogeneous. We propose that WES could be used in first-line diagnosis of undefined progressive neurological disorders of children, as a third of such patients would obtain a diagnosis directly, and the care could be targeted based on the specific disease. The development in NGS methods and analysis, and progressively decreasing price of WES makes the method highly valuable in diagnostic path of children. In future studies, economic J o u r n a l P r e -p r o o f 12 evaluations from the societal perspective including also costs after WES should be conducted; a recent paper (26) finds that diagnosis-related physician consultations do not decline after a negative WES. In addition, the cost-effectiveness should be studied based on other more generic effectiveness measures, such as quality adjusted life years (QALYs).