1. Computation of probability needs to be based on the assumption that the trials (denominator) are independent of each other and the outcomes (numerator) are independent of each other. In this case two stool samples taken from a child is incorrectly considered as independent samples. Therefore, number in the denominator i.e., 1286 which comes from 643 children (i.e., two samples per subject) are not independent; and the outcomes (growth of virus) on two samples from a subject are also not independent. If a child is truly a case of polio then the chances of getting virus grown on both the samples is more, though not always. The authors have considered that "both sample results are independent of each other". For the 148 cases that had one sample negative for WPV, a sample testing positive for pild polio virus (WPV) in Round I implies that the sample from the same AFP case was negative in Round II and vice versa. Therefore, it is incorrect to consider two stool samples taken from the same child as independent samples(2).

2. For those 148 children where one of the two samples was negative, labelling these 148 samples as ‘false negative’ is incorrect. While labelling a sample as false negative there is always a reference. These 148 samples can be labelled as false negative only when there is a superior method than the method in question to show that these 148 samples actually had the virus that was not grown by the candidate method. It is incorrect to label one of the stool samples as ‘false negative’ by retrospectively considering the virological status of child (which is based on any of the stool samples being positive). For any analysis of the performance of a diagnostic test the units of study/analysis must remain the same i.e. both the test result and the disease status have to be on the same unit of study. In this case, there are two units (child and the stool sample) and the authors have used these two units interchangeably. While labelling a stool sample as false negative they have compared the growth of virus at the stool level in the second sample and the true disease status at the child level. Two stool samples are sent for the culture of polio virus to increase the sensitivity of the test. Other examples of similar approach are: sputum testing in RNTCP and stool examinations to diagnose worm infestation. In the two-by-two analysis, for sensitivity and specificity all four cells, including the one which includes disease negative and test negative subjects/samples, should be consi-dered(3). The authors have not included ‘both negatives’ (24,771 cases); the probability estimates would therefore be biased upwards with compromised validity.

3. The authors have retrospectively computed ‘two’ sensitivity figures–one, if a single sample was received and another, if two samples are received. However, it is not feasible to estimate the sensitivity of the yield of polio virus from the stool collected during AFP surveillance in its current form because no gold standard is used for the purpose. Proxy markers like enterovirus isolation rates indicate well functioning surveillance system.