Based on the literature.
Strategies for improving gynecologic cytology screening
Version of Record online: 27 MAR 2009
Copyright © 2009 American Cancer Society
Volume 117, Issue 3, pages 151–153, 25 June 2009
How to Cite
Renshaw, A. A. (2009), Strategies for improving gynecologic cytology screening. Cancer Cytopathology, 117: 151–153. doi: 10.1002/cncy.20021
- Issue online: 11 JUN 2009
- Version of Record online: 27 MAR 2009
- Manuscript Accepted: 10 DEC 2008
- Manuscript Revised: 8 DEC 2008
- Manuscript Received: 20 NOV 2008
As we begin 2009, several new strategies for improving the sensitivity of gynecologic cytology screening are currently being investigated. Although the results of all these studies are not available to date, it is possible, based on the sensitivities already published in the literature, to predict the sensitivity that these different options might be able to achieve. There currently appears to be multiple methods with which to significantly improve the sensitivity of the laboratory for these important specimens.
The reason these options will be so successful is the finding that the procedure for screening currently practiced in the US is simply not very effective. Several large studies have demonstrated that the sensitivity of routine screening (not counting sampling error) using the threshold of atypical squamous cells is approximately 80% (Table 1).1 In addition, the current standard for validating this screening task and detecting errors in the process used in the US (ie, review of 10% of negative cases) is ineffective.2 It is time-consuming, insensitive, and not an independent review of the cases being screened. Together, these facts leave significant room for improvement in this task.
|Routine screening (Renshaw 20021)||80%||Relies on a single screen, no quality assurance|
|Routine screening with 10% review (Renshaw 20021)||80%||Second review not independent, not sensitive, time-consuming|
|Routine screening plus rapid rescreening||Unknown||Second review not independent, unknown sensitivity, less time-consuming|
|Two completely independent routine reviews (Renshaw 20021)||95%||Second review independent, sensitivity of 80%, time-consuming|
|Routine screening plus rapid prescreening (Djemli 20063; Deschenes 20084; Tavares 20085)||85%||Second review independent, sensitivity of 30%, less time-consuming|
|Routine screening by only excellent technologists and rapid prescreening (Deschenes 20084)||95%||Improved routine screening, second review independent, sensitivity of 30%, less time-consuming, excellent technologists not always available|
|Location-guided screening plus rapid prescreening||90%||Improved routine screening, second review independent, sensitivity of 30%, less time-consuming|
|Routine screening and prescreening using location-guided methods||95%||Second review independent, sensitivity of 85%, less time-consuming|
Previous studies have already demonstrated that with the aid of automated screening devices such as location-guided screening, the sensitivity of the process can be improved. Although the exact improvement varies from study to study as well as the type of machine used, a sensitivity of 85% for the process is not unreasonable.1 In addition, rapid prescreening of slides has also indicated that the average laboratory can reduce their error rate by at least 30%.3-5 A laboratory with a sensitivity of 80% would thus be able to improve to a sensitivity of approximately 87%. An added benefit of this method is the ability to document the performance of the laboratory in a real life as opposed to within a study setting.
However, there are several other observations that can be made from these reports. First, in studies of rapid prescreening, it is clear that there are cytotechnologists who are very, very good at routine screening and can achieve sensitivities of >95% for routine screening.4 One effective strategy that a laboratory could use to improve their performance would be to simply limit screening to these technologists. Obviously, the ability to choose this option depends on the volume of cases that are being screened and how many of these great cytotechnologists a laboratory happens to have available. This is not a realistic option for many laboratories.
In addition, one could combine rapid prescreening with automated technologies. One could expect the combination of automated screening devices and rapid prescreening to result in a combined increase in sensitivity. Together, one could expect sensitivities that approach 90%. Certainly this is a straightforward strategy for many laboratories to consider.
However, a more sophisticated use of these automated technologies, particularly location-guided screening, would be rapid prescreening rather than routine screening. In this scenario, one would use the location-guided images to rapidly prescreen the case, and the case would then be screened by hand without any knowledge of the review of the images. This method could conceivably improve the sensitivity of rapid prescreening significantly from 30% to 85%. If one assumes that routine screening by hand had an average screening sensitivity of approximately 80%, this method would result in an overall sensitivity of >95%. Studies examining this exact use of these machines are currently underway in the UK, and the results should be available soon. Looking even farther ahead, since only the images are needed for prescreening, it is possible one could have a separate laboratory do the prescreening, thus reducing overall bias even further.
Given this, it would be tempting to then suggest that the very best method would be to use location-guided screening for both the prescreening and the routine screen. However, this will not work because the 2 screening events are no longer independent events. As a result, although the study will clearly “demonstrate” excellent sensitivity, this sensitivity is not accurate because the testing method is biased by not being independent.
It may appear to be counterintuitive to spend money on automated devices, such as location-guided screening, and then only use them for what has been traditionally considered to be quality assurance. However, this perspective misses the real point, which is that prescreening as it has evolved is more than just a quality assurance method; it is essentially a completely independent second review. The reason the method works so well is that 2 independent, unbiased screenings of the cases are much better than 1 screening. Obviously, the amount of time necessary to perform 2 independent routine screening events has been prohibitive when the only option has been routine manual screening. This is why rapid prescreening has been pursued; it allows a second independent review of the case that can be performed in the amount of time that people are willing to put into a second review, and it does so in a way that is independent and thus not biased by the routine review. Directing our technologic advances to this part of the process, rather than routine screening, would appear to offer significant advantages.
Although it may be reasonable to wait for the results of the studies in the UK to finally be published, analysis of the expected benefits of this use of technology are striking. It is quite clear that the traditional method of relying on a single screening can only achieve so much and makes no effort to identify and correct mistakes that occur when a technologist has a bad day, which they clearly do.4 The evolution of “second screening” from an ineffective, time-consuming, and nonindependent review of approximately 10% of negative smears to rapid rescreening; then rapid prescreening; and then rapid prescreening using location-guided screening systems all represent major improvements in the process of gynecologic screening that result in a second review that is more effective, less time-consuming, and independent.
Conflict of Interest Disclosures
The author made no disclosures.