Math problem with results, page-21

Do you mean this:
Latest analysis (announced 22/6/17) optimised to match design of US SMARTCOUGH-C study

I'm not sure you're correct in inferring a change in data collection methodology. This statement refers to the announcement of 22-Jun-2017, which has this paragraph:
The dataset and machine learning algorithms have been optimised to match the design of ResApp’s US SMARTCOUGH-C study as closely as possible. The clinical diagnosis, used as a comparator, has been adapted to match the SMARTCOUGH-C US-centric clinical case definitions. The algorithms have been fine-tuned to differentially diagnose disease from all patients recruited to the study, as required by the SMARTCOUGH-C endpoints. Following published guidance by the US FDA, the key measures of positive percent agreement (PPA) and negative percent agreement (NPA) are reported instead of sensitivity and specificity.

The changes mentioned here are:
1) Dataset
2) Algorithm
3) Clinical diagnosis

That being the case the change in PPA, NPA and overall agreements would result from using different data points and re-analysing with the modified algorithm.

You also referred to the conference call in support of your conclusion that the Australian data collection practices were changed, but I can't find any support for that conclusion. Tony certainly said that in the US the "Clinical Research Team" cannot access the patient before the "Clinical Treatment Team" give their clearance. But he was answering a question here about the differences between the Australian and US procedures. I don't believe Tony at any stage indicated directly or indirectly that the Australian procedures were changed to match the US ones.

I would urge caution building a case on the evidence presented so far. The public announcements specifically mention three types of changes, and the clinical procedures are not in that list. Nor do they imply that any changes other than the listed ones were made.