Improving Cystic Fibrosis Screening Through a Novel Testing Design
提出一种将小型基因检测与IRT检测结合的新生儿囊性纤维化筛查设计，通过优化IRT阈值和基因检测面板，在效率和公平性上优于现行单一阈值方法，基于纽约和加州数据验证。
Newborn screening (NBS) plays a key role in detecting life-threatening genetic disorders, with cystic fibrosis (CF) being a prominent example. Having CF requires both copies of the CFTR gene (one inherited from each parent) to be pathogenic variants. With over 1,000 CF-causing variant types, a genetic test that can detect enough variant types for sufficient screening accuracy is expensive and capacity-constrained. Consequently, in the US, the first tier of testing uses an inexpensive biochemical test that measures immunoreactive trypsinogen (IRT) levels, which are often elevated in CF cases. Genetic testing is reserved for a small fraction of newborns with elevated IRT levels, based on an IRT threshold. However, setting an optimal IRT threshold that balances sensitivity with overall screening cost and genetic testing capacity is challenging, which is exacerbated by variations in IRT levels and CF prevalences among different racial/ethnic groups. As a result, the current practice of using a single IRT threshold for all newborns is not necessarily the most efficient or equitable strategy. This paper introduces a novel screening design that integrates an inexpensive small-panel genetic test with an IRT test in the first tier of testing. This strategy customizes IRT thresholds based on the number of variants detected by the small-panel genetic test. This design leads to novel optimization problems, including variant type selection for the small-panel genetic test and IRT threshold customization. We establish key structural properties of optimal IRT thresholds and develop Pareto frontiers that inform decision-makers about the efficiency versus equity trade-off, considering a specific minimax type equity measure. The case study, utilizing CF NBS data from the states of New York and California, demonstrates the potential efficiency and equity gains of this novel approach compared to current practice and race-based benchmarks. This research offers valuable insights and methodologies for improving CF NBS processes, and has implications for public policy decision-making.