Table 5.

Avoiding hidden pitfalls to optimise using 100KGP data for cardiovascular research

1. Initial eligibility criteria for prior testing of known genes were relaxed

This occurred throughout the time course of the project for cardiomyopathies. Therefore, while the cohort is depleted for known genetic diagnoses, some recruited probands do have pathogenic/likely pathogenic variants in known disease genes.

2. Initial eligibility criteria for family history requirements changed

For cardiomyopathies, family history requirements were relaxed during the course of the project. Conversely for a minority of disease categories, such as renal and urinary tract disorders, clinical criteria were tightened.

3. Recorded age of onset is unreliable for early onset diseases

Where age of onset was not entered GEL assigned a value of zero making recorded age of onset unreliable, particularly for early onset diseases, i.e. congenital onset is indistinguishable from missing data.

4. The majority were recruited as singletons limiting scope for de novo analysis

Only 14% of the cardiomyopathy cohorts were recruited as trios (proband, mother and father), which limits the scope for de novo analysis. The proportion of probands recruited as trios is likely to vary across different disease categories and be impacted by age of onset of disease (for cardiomyopathies, children were more likely to be recruited as trios compared to adults), and by family history requirement (for cardiomyopathies, probands with a positive family history were less likely to be recruited as trios).

5. Phenotype recording varies in detail

For cardiomyopathies, it often just includes 1 or 2 features key to eligibility and is in the form of human phenotype ontology (HPO) terms. Objective measures of disease such as ejection fraction are not currently available. Deep phenotype data which would allow reverse phenotyping is typically unavailable.