Morphological Bone Score: Assessment Tool for Predicting Downstream Processing Success for Decalcified Tissue Samples: A Cost-Saving Approach in Precision Oncology
Authors
- BostonGene, Corp., Waltham, USA
Abstract
Although decalcification of tumor samples involving bone is often required to soften tissues prior to histological processing, it can lead to loss of nucleic acid integrity, resulting in next-generation sequencing (NGS) failures that impede diagnostic solutions for patients. We developed a morphological Bone score tool to evaluate the suitability of decalcified tissue samples for successful downstream processing.
Methods
Morphological features linked with either overly aggressive or insufficient decalcification of samples containing bone tissue were used as assessment criteria by experienced pathologists: bone basophilia, osteocyte presence, matrix structure, nuclear features, and stromal features. The sum of feature scores was calculated for a total Bone score from 0 to 11, reflecting low to high tissue damage (n=169 samples, BostonGene laboratory). Samples were divided into two groups based on processing outcome: successful or failed extraction/NGS. Confusion matrices evaluated the performance of Bone score cutoffs in predicting processing success. To assess economic advantages of using our tool, we determined the approximate processing expense per sample was $1,500, including labor and reagent costs (extraction, NGS and bioinformatics analysis).
Results
To use our tool as a screening test, we utilized a precision-recall curve analysis to select the Bone score cutoff in classifying sample processing outcomes (successful vs. failed). While high Bone score cutoffs of 6 and 8 improved sensitivity (0.94 and 1), PPVs were low (0.79 and 0.72, respectively). Conversely, lower cutoffs of 4 or 2 yielded higher PPV (0.89 and 0.91), but substantially reduced sensitivity (0.78 and 0.6, respectively). Thus, the selected cutoff can be adaptable to either prevent sample loss or minimize unnecessary processing and related costs.
A Bone score cutoff of 5, coinciding with the intersection of precision-recall curves, yielded the best tradeoff. Among 169 samples, this cutoff achieved sensitivity=0.89 and specificity=0.6, and accurately classified 80% of samples (106 true positives, 30 true negatives). This classification led to positive predictive value=0.84 with 20 false positives and negative predictive value=0.7 with 13 false negatives. Overall, the Bone score cutoff performed well in correctly assigning samples as successful or failed (F1 score=0.87). Exclusion of decalcified samples of poor quality from downstream processing using the Bone score can reduce sequencing costs (~$45,000 for 30 true negatives).
Conclusions
The morphological Bone score tool evaluates tissue damage from decalcification to aid laboratories in predicting downstream processing success, which may prevent ineffectual cost and time expenditures that delay value-based care to patients.
Latest publications
