Correlation Studies

Most laboratories doing microscopy have "self-calibrated" to some internal standard. Such self-calibration is precisely the target of CLIA and other regulatory mechanisms, but is a regrettable reality in sperm analysis.

Until the advent of the SQA, it has not been possible to use an objective standard, so each population of laboratorians has converged upon its own set of judgment rules.

Because of that self-calibration, it is almost certain that when the SQA is introduced to a lab there will be significant differences between its output and that of microscopic evaluations. Fortunately, this is easy to resolve once it is accepted that the lab is offset by some factor.

Whether or not the results match, the issue is whether they are coherent and monotonic. That is, when one goes up so should the other, and vice versa. The animated graph below is an example of "good" correlation. Note that on that initially the tests are offset one from the other, and in no case did both tests achieve the same result.
 

They "correlate" because they follow one another closely. By applying a simple offset correction, they can be made to look almost identical.

On the left is an example of "poor" correlation. Some of the test pairs achieved exactly the same results, but others were so far off that a statistical analysis shows very poor performance, and there is no offset correction that would improve the fit between the curves.

The following figures reflect the results of studies by a reputable laboratory, after a statistical analysis applied the required offset. We have been very critical in presenting this information: there are better results achieved by many SQA users, but these correlations are typical.

You will note that in this particular set there was a problem in evaluating samples 4 and 5. Later analysis indicated that the error was in the microscopy rather than the SQA.