With the recent introduction of nationwide proficiency testing in cytopathology (PTC), reconsideration of several aspects of this controversial quality assurance method becomes justified. This paper discusses various merits and demerits of the PTC system currently prescribed by federal regulations, points out perceived deficiencies, and suggests methods for improvement.

Thirty-three years ago Penner advocated six criteria for the performance of proficiency testing in cytopathology (PTC). Since that time, several further requirements have been added by other authors. The present article critically evaluates and modifies the original criteria and adds two more principles, validity and reliability, that we recognize as crucially important in the performance of PTC. The revised criteria should be taken into consideration in the planned technical redesign of the nationwide PTC.

insists that proficiency testing in cytology is worthless as a quality assurance method. 1+2 He believes that his point can be proved by comparing the relatively slower decline of cervical cancer mortality rates in the State of New York where cytologic proficiency testing was established 25 years ago, with the more rapidly decreasing rates in other states where no proficiency testing has existed.Dr. Austin's views reveal a fundamental ndivete about basic principles in epidemiology. To evaluate a population intervention, proficiency testing, he attempts to use the "mixed design" of the "ecologic" or "correlational" analysis. Ecologic studies are investigations involving groups as units of analysis. Typically, the group is a geographically defined area, such as a country, state, census tract, etc. In his editorial, Dr. Austin falls into the trap of "ecologic fallacy," which is the bias resulting from making a causal inference about a possibly unrelated, or only very superficially related, phenomenon on the basis of observations on groups. One of the major limitations of ecologic investigations in epidemiology is the total lack of ability to estimate or evaluate the effects of confounding factors. And, in this case, confounding factors abound. Cervical cancer mortality is determined by the interaction of powerful risk factors like human papilloma virus infection (related to the age at first intercourse and the number of sexual partners), education, number of births, smoking, oral contraceptives, dietary factors, immunosuppression, infectious agents other than human papilloma virus, and possibly other, not well known conditions, such as heredity. Access to screening, follow-up procedures and treatment, compliance with treatment, stage at diagnosis, histologic type of the neoplasms, proportion of hysterectomies for reasons other than cervical cancer, and the patients' general health also influence the outcome. What is the position and relative importance of proficiency testing in this very long list?It seems ascertained that cytologic screening is causatively related to the continuing decline of cervical cancer mortality, although the evidence for this association is circumstantial. * This is not surprising. It is sometimes rather difficult to prove causative relationships between risk factors and outcomes even with the most sophisticated epidemiologic methods, and we need not cite examples from the plentitude of literature available to those who are genuinely interested. One of the most elementary precepts in epidemiology is that correlation does not necessarily mean causation, and, conversely, lack of correlation does not prove that no association exists. How can it be expected that a direct causative relationship be proved or disproved between cytologic proficiency testing, which is a test of screening and interpretative ability, and the relative decline or increase of cervical cancer mortality? To prove the beneficial effects of screening itself is difficult enough! Should the CAP Interlaboratory Comparison Progra...