Cervical cancer incidence after normal cytological sample in routine screening using SurePath, ThinPrep, and conventional cytology: population based studyBMJ 2017; 356 doi: https://doi.org/10.1136/bmj.j504 (Published 14 February 2017) Cite this as: BMJ 2017;356:j504
All rapid responses
Re: Cervical cancer incidence after normal cytological sample in routine screening using SurePath, ThinPrep, and conventional cytology: population based study
With great interest I read the article by Rozemeijer et al published recently in British Medical Journal (BMJ. 2017 Feb 14; 356:j504).
The article is intended to provoke reconsideration of the assumed similarity in the sensitivity for progressive CIN between the different types of liquid based cytology tests and conventional cytology. For this purpose the authors conducted a follow-back study identifying selected episodes that started with a normal primary screening sample taken from women within the Dutch screening program between January 2000 and March 2012. From these women the rates of invasive cancer detected after the normal screening sample was compared between the different cytology methods.
As one of the regional coordinators of the Dutch cervical screening program, responsible for region Netherlands South, I have some concerns about the data provided for the two different types of liquid based cytology and conventional cytology and the conclusions drawn from this.
Firstly, the results of this study are very interesting raising an important question that deserves closer inspection. The data from this study may point us in the right direction, but cannot finally answer the questions about long-term safety of the different cytology methods. The only way to investigate this in a meaningful and conclusive way would be a prospective study or a study with a certain well defined cohort (i.e. including all women screened in 2007/2008) with re-evaluation of all slides prior to the cancer diagnosis.
It should be taken into account that the study is based on a retrospective data analysis, which is in my opinion a serious drawbacks that could have influenced the presented results.
The impact of a number of the following factors, identified by the authors, need to be addressed in more detail.
1. Laboratories might have switched during the selected episodes from one method to another and even vice versa! The implementation of a new technology and with this the transition period can easily take, to my best knowledge, in most laboratories 3-6 months, with an additional learning curve period. To use fixed conversion dates in the study and not considering a transition period of at least 6 months, will lead to an uncontrolled bias.
2. Reading the article I assumed it described a population based screening situation? But also non-population based screening samples (diagnostic smears) taken after 4 years were labelled as primary. It is not clear what this means and which percentage of non-population based screening samples tests are labeled this way? What is the effect of misclassified “primary” samples on the results?
3. It is stated that computer assisted screening could be a cofounding factor. I agree. In my opinion another cofounding factor is HPV triage. Starting in 2007 laboratories were also allowed to use the so called HPV triage. (Siebers et al 1) In the beginning there was a slow conversion and only a few laboratories performed HPV triage testing. In 2012 nearly all laboratories performed this HPV triage testing. Therefore a fixed percentage of woman were directed back earlier to the routine program. This could be a bias factor as well.
4. The difference in follow-up time between the techniques and also the market share of the techniques used in the study could have a major impact on the results. Market share data for ThinPrep in figure 2 are 20-30% too high (personal communication from Hologic).
5. In relation to the difference in follow-up time in text and figure 1 the 72 month follow up is highlighted. We wonder if it is justified to draw conclusions of such a long period (i.e. differences in follow-up time; differences in cofounding factors) as is known that the "expiry date" of a smear is around 36 months and in that period the differences between all three technologies were very small.
6. Furthermore the stage of the cancers detected is not mentioned. These stages should be low with a very good prognosis. In addition this finding urges the need for reviewing the cytology.
7. I also noted that a remarkable amount of cancers were left out of this study. With 700 cancer cases per year in the Netherlands this study should have included 12x 700= 8400 cancer cases. Only 2007 (24%) cases are included in the study.
8. It is also known that due the lack of an individual ID number for women a significant number of data sets have to be excluded from any retrospective analysis of data from National Dutch Pathology Data Base (PALGA).
Despite these flaws, this publication describes a very interesting study with a huge amount of data. Nevertheless one should be cautious with conclusions and should not make the mistake to over interpret the results.
A.G. Siebers et al. Effectiveness of two strategies to follow-up ASC-US in cancer causes central (2014) 1141- 1149
Competing interests: No competing interests
Relevance of comparing approved Liquid Based Cytology tests for their effects on population indicators
In countries with organized cervical cancer screening programs, the last decade has seen Liquid Based Cytology (LBC) ousting Conventional Cytology (CC) as the preferred screening test. Compared to CC, LBC is reported to offer similar or better detection rates while at the same time reducing the testing failures due to low quality slides. With dramatically reduced reading times, and fewer repeat tests, LBC possibly fares favorably in cost-effectiveness as well. BD Surepath and Thinprep (Hologic) are the most widely used and approved LBC tests. Surepath has been generally accepted as a slightly better performer, providing a better cell yield and quality, but the effects of such a difference on the population indicators haven’t been studied. However, with recent gravitation towards HPV based Cervical screening test in the developed countries, the impact of such studies would be noticeably dampened.
This particular study by Rozemeijer et al sought to compare the crude cumulative incidence and hazard rates of cervical cancer diagnosis within 72 months after a normal screen done using Surepath, Thinprep, or CC, by performing a retrospective population based cohort study. The study featured an essentially conventional design, with reasonable adaptations where required. Using the data from 2000 to 2013, available from a nationwide cytopathology database, the cumulative incidence of invasive cervical cancer detected within 72 months after a normal screen result was assessed, taking in account the calendar time, follow-up time distributions, age, socioeconomic status and screening history as confounders. Determinant labs were included as random effects Stratification according to the test type was done by correlating conversion dates of individual labs with the test dates. Kaplan-Meier curves of each stratum were charted, and Cox regression analyses performed to assess the hazard rates of invasive cervical cancer after a normal screening test
Surepath was associated with not only significantly lower crude cumulative incidence [44.6 per 100,000 samples (95% CI 37.8 to 52.6)] than CC and Thinprep, but also with lower hazard [hazard ratio 0.81 (95% confidence interval 0.66 to 0.99)] of developing an invasive cancer as compared to CC. Thinprep was insignificantly associated with higher crude cumulative incidence [66.8 (56.7 to 78.7)] and a higher hazard [hazard ratio 1.15, (0.95 to 1.38)] of an invasive cancer as compared with CC. Surepath, as a primary test, showed higher detection rate [94.4% (95% confidence interval 68.9 to 120.6)] as compared to CC. The detection rates between Thinprep and CC were essentially similar, and varied only towards the approach of next scheduled screen.
The outcome measures focused on incidence and hazard ratios for each screening test which may not translate to actual survival benefits due to low mortality associated with early invasive cervical cancer[1,2]. Furthermore, the recent estimates of cervical cancer incidence in developed regions, and in western europe, are comparatively low . The fact that greater proportion of incident cases were diagnosed clinically versus the scheduled screen lends itself well to the possibility that the outside-the-scheduled-screen examinations occurred with sufficient frequency to serve as an adequate safety net. The morbidity incurred due to higher and earlier detection, may outweigh the low risk of mortality from early stage cervical cancer, thereby questioning the value of detection gains offered by Sure Path over CC or Thinprep.
Other weak points are directly attributable to a few necessary assumptions and bulk trends, and have been adequately detailed by the authors. Since the type of test used has been indirectly deduced from lab conversion dates, data near those dates is likely to suffer from misclassification. The extent of such misclassification could not be ascertained. The proportional shift from CC to LBC tests was concentrated in 2006-2008 period; any misclassified data sets near that period could easily contaminate the whole design without being readily visible as a confounder. Additionally, the type of test used differs not only at the baseline but also within/at the end of the episode. While the authors did adjust for its confounding effects with reported insignificance while performing sensitivity analyses, it could have also been just as easily accommodated as another level of stratification; especially since the study does include such a classification but for simple proportional analysis. Confounding due to automated vs manual reading and mortality/migratory trends weren’t ascertained or adjusted for. The trend towards greater proportion of Thinprep testing vs others could warrant a second look.
The implications of outcomes obtained from such studies would be best utilized by regions with established cervical cancer screening programs which are in the process of switching from CC to LBC methods. For regions where screening programs are still in nascent stages, like densely populated and resource-scarce countries in Africa and Asia, direct switching from lower quality methods like VIA to only one suitable LBC test would offer immediate survival benefits and long-term cost benefits. The impact of similar studies in the regions with advanced cervical screening programs is likely to be dulled both due to switching trends towards HPV based cervical screening and the increasing coverage of immunization against HPV.
ICMR-National Institute of Cancer Prevention and Research
Department of Health Research
Noida, U.P., India
1. Howlader N, Noone AM, Krapcho M, Miller D, Bishop K, Altekruse SF, Kosary CL, Yu M, Ruhl J, Tatalovich Z, Mariotto A, Lewis DR, Chen HS, Feuer EJ, Cronin KA (eds). SEER Cancer Statistics Review, 1975-2013, National Cancer Institute. Bethesda, MD, http://seer.cancer.gov/csr/1975_2013/, based on November 2015 SEER data submission, posted to the SEER web site, April 2016.
2. Cancer Research UK, http://www.cancerresearchuk.org/health-professional/cancer-statistics/st..., Accessed February, 2017
3. Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S, Mathers C, Rebelo M, Parkin DM, Forman D, Bray, F.
GLOBOCAN 2012 v1.0, Cancer Incidence and Mortality Worldwide: IARC CancerBase No. 11 [Internet].
Lyon, France: International Agency for Research on Cancer; 2013. Available from: http://globocan.iarc.fr, accessed on 27/02/2017.
Competing interests: No competing interests
In a recent analysis of all available studies, cancer screening has never been shown to “save lives”. 
The cumulative incidence of women with cervical cancer was only 0.09% after 15 years, in this randomised cohort study in the Netherlands. 
Younger generations of women will be rendered low risk through ongoing HPV vaccination programs.
Even if invasive cervical cancer is present and colposcopically visible, stage IA1, affected women can expect survival rates of 99%. 
Even delayed treatment for diagnosed cervical cancers, at Stage I, was not associated with worse survival. 
Reported over-diagnosis and over-treatment would significantly increase morbidity in completely healthy women.
Statistically, sexually active women will experience vaginal ailments that necessitate clinical colposcopic examination sooner than 5-10 years, so any potential cervical lesions would be detected BEFORE formal screening periods.
Competing interests: No competing interests