Table 2.

Screening programme performance indicators and quality measures that are primarily dependent on the test and which can be estimated at specific phases of the evaluation process

Category of screening outcome	Phase	Measure	Comment
Initial performance indicator	III
Invitation		Participation (uptake)	Fraction of invitees who complete a test.
		Non-participation	Fraction failing to complete a test. Ascertain reasons why refused.
Sample laboratory analysis		Positivity fraction	Fraction of test results that are positive. Determines colonoscopy workload.
		Test failures	Degree to which samples are unsuitable for, or otherwise fail, in the measurement whether due to collection, storage, transport or measurement difficulties.
Detection rate*		Number of CRCs detected in the population	Fraction of participants doing the test, where CRC is diagnosed at follow-up colonoscopy (per-protocol analysis). OR Fraction of invitees in whom CRC is diagnosed (intention-to-screen analysis). Stage distribution of detected CRCs to also be determined.
		Cancer stage distribution and/or detection rate by stage Number of stage I and stage II CRCs detected in the population	Shift to more favourable stage at diagnosis points to a likely benefit on CRC mortality.
		Advanced precursor lesions detected in the population	Shift to more favourable stage at diagnosis points to a likely benefit on CRC mortality.
		Any adenoma or serrated lesion detected in the population	Shift to more favourable stage at diagnosis points to a likely benefit on CRC mortality.
		Non-neoplastic pathology in the population	This will inform the variables associated with false positivity for CRC or advanced precursor lesions.
Predictive values		Positive predictive value for each category of neoplasia	The efficiency of detection: the proportion of positive tests with CRC or advanced precursor lesions.
Test accuracy		Estimates of sensitivity and specificity	Only obtainable if all participants undergo colonoscopy†
		Estimates of sensitivity, specificity relative to a comparator	Obtainable by comparing true and false positives of new and comparator test (does not require all participants to undergo colonoscopy)‡.
Burden of detection		Number needed to colonoscope to detect one lesion of interest	Inverse of the fraction of all those colonoscoped who have CRC or advanced precursors. A simple indicator of cost-effectiveness.
Subsequent performance indicator	III	Interval cancers31 32	Best ascertained with follow-up time equal at least to the recommended screening interval. This may include lesions missed by the screening tests (interval lesions among subjects with a negative non-invasive test result) and lesions missed at colonoscopy (ie, among subjects with a negative follow-up colonoscopy). Identify the optimal retest interval.
	IV	Cumulative detection rates over multiple rounds31 32	Such can also be expressed as a fraction of all test-positive participants.
		Detection and burden of detection according to interval between tests	Further inform the optimal retest interval.
		Cumulative colonoscopy workload	Over subsequent rounds, more and more people will require diagnostic assessment.
	III and/or IV	Modelling of efficacy and effectiveness	CRC mortality and incidence modelled from the above indicators. Stage distribution at diagnosis will be crucial.
		Modelling of cost-effectiveness	Costs will be specific to the jurisdiction of application of the test.
		Unexpected adverse events32	Include mortality and 30-day hospital readmission rates following colonoscopy.

Each measure can be compared between the new and a comparator non-invasive test.

*Detection rates will depend on the threshold chosen for positivity. It will be possible to estimate rates across a range of thresholds depending on the trial design (see principles 8 and 10).

†Sensitivity is the proportion of cases with a given neoplastic lesion type(s) that return a positive test. Specificity is the proportion of cases without CRC or advanced precursor who return a negative test.

‡Note that biases referred to in principle 10 will mean that the actual estimates are not necessarily reliable. These are best determined when all cases undergo colonoscopy.

CRC, colorectal cancer.