Predicting pathologic complete response to neoadjuvant chemotherapy in breast cancer using a machine learning approach

Table 3 Performance Metrics of the final model for different prediction thresholds among HR+/HER2- patients

Threshold ^a	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Patients waiving chemotherapy (%) ^b	Net reduction in intervention (%) ^c
3%	99.1	14.5	16.7	99.0	12.5	8.1
5%	96.9	34.1	20.2	98.5	29.5	20.4
7%	90.5	48.8	23.4	96.7	43.0	23.1
10%	82.3	62.8	27.6	95.4	56.2	30.1
15%	73.7	74.7	33.4	94.3	67.5	41.8

Abbreviations HR, hormone receptor; HER2, human epidermal growth factor receptor 2; PPV, positive predictive value; NPV, negative predictive value
^aHaving a predicted pCR probability lower than the threshold is predicted non-pCR, whereas greater than the threshold would be predicted pCR. The prediction model used here is the quantitative machine learning model
^bTreatment options other than chemotherapy might be considered if the patients’ predicted pCR probability estimated by the quantitative machine learning prediction model is less than the selected threshold
^cNet reduction in intervention is calculated by \(\:\text{S}\text{p}\text{e}\text{c}\text{i}\text{f}\text{i}\text{c}\text{i}\text{t}\text{y}\times\:\left(1-\text{P}\text{r}\text{e}\text{v}\text{a}\text{l}\text{e}\text{n}\text{c}\text{e}\right)-\:\left(1-\text{S}\text{e}\text{n}\text{s}\text{i}\text{t}\text{i}\text{v}\text{i}\text{t}\text{y}\right)\times\:\text{P}\text{r}\text{e}\text{v}\text{a}\text{l}\text{e}\text{n}\text{c}\text{e}\times\:\frac{Threshold}{1-Threshold}\) based on Decision Curve Analysis, where Prevalence is the rate of pCR among HR+/HER2- patients (14.7%)

ISSN: 1465-542X