Cervical screening and risk of adenosquamous and rare histological types of invasive cervical carcinoma: population based nested case-control studyBMJ 2019; 365 doi: https://doi.org/10.1136/bmj.l1207 (Published 03 April 2019) Cite this as: BMJ 2019;365:l1207
- Jiayao Lei, postgraduate student1,
- Bengt Andrae, senior consultant1 2,
- Alexander Ploner, statistician1,
- Camilla Lagheden, biomedical scientist3,
- Carina Eklund, biomedical scientist3,
- Sara Nordqvist Kleppe, database administrator1 3,
- Jiangrong Wang, postdoctoral researcher1 3,
- Fang Fang, senior researcher1,
- Joakim Dillner, professor3 4,
- K Miriam Elfström, research associate3 5,
- Pär Sparén, professor1
- 1Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, 171 77 Stockholm, Sweden
- 2Center for Research and Development, Uppsala University/Region of Gävleborg, 801 88 Gävle, Sweden
- 3Department of Laboratory Medicine, Karolinska Institutet, 141 83 Stockholm, Sweden
- 4Karolinska University Laboratory, Karolinska University Hospital, 171 76 Stockholm, Sweden
- 5Regional Cancer Center Stockholm-Gotland, 118 27 Stockholm, Sweden
- Correspondence to: J Lei @JiayaoL on Twitter) (or
- Accepted 11 March 2019
Objectives To examine the association of cervical cytology screening with the risk of adenosquamous cell carcinoma (ASC) and rare histological types of invasive cervical carcinoma (RICC), using comprehensive registry data, and to assess tumour human papillomavirus status of ASC and RICC.
Design Nationwide, population based, nested case-control study.
Participants All cases of invasive cervical carcinoma in Sweden during 2002-11 (4254 confirmed cases after clinical and histopathological review). 338 cases were neither squamous cell carcinoma nor adenocarcinoma, including 164 cases of ASC and 174 cases of RICC (glassy cell carcinoma, clear cell carcinoma, small cell carcinoma, neuroendocrine cell carcinoma, large cell carcinoma, and undifferentiated carcinoma). 30 birth year matched controls from the general Swedish population were matched to each case by applying incidence density sampling.
Main outcome measures Conditional logistic regression was used to calculate odds ratios, interpreted as incidence rate ratios, for risk of ASC and RICC in relation to screening status and screening history, adjusted for education. Human papillomavirus distribution of ASC and RICC was based on available archival tumour tissues from most Swedish pathology biobanks.
Results Women with two screening tests in the previous two recommended screening intervals had a lower risk of ASC (incidence rate ratio 0.22, 95% confidence interval 0.14 to 0.34) and RICC (0.34, 0.21 to 0.55), compared with women without any test. High risk human papillomavirus was detected in 148/211 (70%) cases with valid human papillomavirus results from tumour tissues. The risk reduction among women with tumours that were positive (incidence rate ratio 0.28, 0.18 to 0.46) and negative (0.27, 0.13 to 0.59) for high risk human papillomavirus was similar, compared with women who did not attend any test.
Conclusions Cervical screening is associated with reduced risk of ASC and RICC, and most ASC and RICC are positive for high risk human papillomavirus. This evidence provides a benchmark for evaluating future cervical screening strategies.
The primary goal of cervical screening is to prevent invasive cervical carcinoma by detecting and removing precancerous lesions (cervical intraepithelial neoplasia 3 or adenocarcinoma in situ), which if left untreated could lead to invasive cervical carcinoma. Hence, cervical screening can effectively reduce the incidence of and mortality from cervical carcinoma and concurrently detect asymptomatic invasive malignancies early, which affects the prognosis of invasive disease.12 Previous studies have shown that screening is associated with reduced risk of squamous cell carcinoma and adenocarcinoma, and to some extent of adenosquamous cell carcinoma (ASC).234 However, no studies have looked at screening and risk of rare types of invasive cervical carcinoma (RICC).
ASC is a histological type of invasive cervical carcinoma that is composed of a mixture of malignant glandular and squamous components.5 RICC include a group of histological types that are of glandular origin, with overlapping morphology, and the histopathological classification of these types is relatively difficult.5 RICC have also been reported as highly aggressive, with a worse prognosis than squamous cell carcinoma and adenocarcinoma.1 In previous studies on screening and risk of invasive cervical carcinoma, RICC have been classified as “non-squamous cell carcinoma” or “other types” besides squamous cell carcinoma and adenocarcinoma or have simply been excluded from the analysis.24 Limited evidence is available on the human papillomavirus status of these histological types.
In Sweden, women aged 23-60 years were invited to cervical cytology screening every three years until age 50 and every five years thereafter, according to national guidelines before 2015.6 The aim of our study was to examine the association of cervical cytology screening with the risk of ASC and RICC, by using comprehensive registry data, and to assess the tumour human papillomavirus status of ASC and RICC.
We did a population based, nested case-control study in a cohort of all women born during 1909-86 in Sweden. We identified 4533 cases of cervical cancer and unspecified uterine cancer during 2002-11, through cross linkage to the Swedish Cancer Register7 (supplementary figure A). We did a thorough clinical and histopathological review of all cases to ascertain the final diagnosis. Subsequently, we excluded 279 cases for the following reasons: not primary cervical origin, not epithelial, not invasive, and recurrence of a previous cancer according to review of medical charts by a single expert gynaecologist (BA), leaving 4254 confirmed cases of primary invasive cervical carcinoma. A senior pathologist (WR) did a histopathological review of 91% of all sample slides collected from pathological laboratories in Sweden. Among the confirmed cases, 338 were classified as neither squamous cell carcinoma nor adenocarcinoma, including ASC and RICC (glassy cell carcinoma, clear cell carcinoma, and other rare types of invasive cervical carcinoma such as small cell carcinoma, neuroendocrine carcinoma, large cell carcinoma, and undifferentiated carcinoma). We used date of diagnosis as the index date for these cases.
For each case, we randomly selected 30 controls from the Total Population Register,8 using incidence density sampling and individual matching by year of birth, which corresponds to age at time of diagnosis of the case. All controls were alive, with no history of cervical cancer, and living in Sweden up to the date of diagnosis of their matched case. We used date of diagnosis of the corresponding case as the index date for the controls. We subsequently excluded controls (n=449) who had a history of total hysterectomy, because they were no longer at risk of cervical cancer after total hysterectomy, leaving a total of 9691 controls eligible for analysis. Information on total hysterectomy came from the Swedish Patient Register,9 established in 1964 and with complete national coverage of information on inpatient care since 1987.
Human papillomavirus genotyping
We retrieved archived formalin fixed, paraffin embedded blocks from the archives of the diagnosing pathology laboratory, and 2909 of the confirmed cases were genotyped for human papillomavirus (68.4%). The blocks were extracted and tested in parallel with β-globin real time polymerase chain reaction and human papillomavirus genotyping by using general primers polymerase chain reaction targeting the L1 region,10 followed by typing with Luminex for 13 high risk types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68) and non-high risk types (6, 11, 26, 30, 40, 42, 43, 53, 54, 61, 66, 67, 69, 70, 73, 74, 81, 82, 83, 86, 87, 89, 90, 91) as previously described.11 We used a blank block containing only paraffin as a control for contamination, which was sectioned and analysed in between each case block. In total, 2850 confirmed cases had valid human papillomavirus genotyping, including 211 cases of ASC and RICC, which accounted for 62% of all cases (n=338) included in this study.
The exposure was cervical screening measured as screening status and history of the previous two recommended screening intervals (calculated according to women’s age at the index date) more than six months before the index date (supplementary table A). We assumed Pap smear tests within six months before the diagnosis of cervical cancer to be part of the diagnostic investigation and did not consider them to be screening tests. We linked cases and controls to the Swedish National Cervical Screening Registry to retrieve information on attendance for screening and the eventual results of the screening tests.12
We categorised screening status as “no test,” “one test,” and “two tests” on the basis of attendance for screening in the defined two screening intervals (supplementary table B). Screening history was defined by the cytology (Pap smear) result of the two screening intervals according to Systematized Nomenclature of Medicine codes (supplementary table C) defined by the Swedish Association of Clinical Cytology. An abnormal smear included a diagnosis of atypical squamous cell of unknown significance or worse. We categorised screening history as “no test,” “double normal results,” “one normal result only” (including women with a normal test in one of the two screening intervals but without a test in the other interval), and “at least one abnormal result” (including women with at least one abnormal test during the two screening intervals, regardless of whether they participated or had a normal test in the other interval) (supplementary table D). We categorised mode of detection as screen detected or symptomatic cancer according to the medical charts.
We described the distribution of ASC and RICC according to age, cancer stage, and mode of detection. We categorised age at diagnosis into three groups (22-29, 30-60, and >60 years). We classified cancer stage according to the International Federation of Gynecology and Obstetrics guidelines and categorised it into IA (microinvasive), IB (localised), and II or higher (advanced).13 We included women aged 30 or above for the analysis of the associations of screening status and history with the risk of ASC and RICC, because they had two full screening intervals before the index date.
We used conditional logistic regression to calculate odds ratios with 95% confidence intervals of ASC and RICC, after adjusting for the highest completed education by the year of cancer diagnosis (classified as low, middle, and high). We retrieved this information from the Longitudinal Integration Database for Health Insurance and Labour Market Studies database, which includes everyone aged 16 and above registered in Sweden at the end of each year since 1990.14 The educational information in the database is derived from the National Swedish Education Register, established in 1985. The matching variable birth year, which corresponds to age at the time of diagnosis of the case, was automatically adjusted for in the model. Given that we tracked incident cases in a dynamic population, matched on time at event, we interpreted the odds ratios as incidence rate ratios,1516 which served as estimates for the association between cervical screening and risk of ASC and RICC. We also stratified the analysis by age at index date and cancer stage, to examine the heterogeneity of the studied association by age and cancer stage.
Among cases with available genotype data for human papillomavirus, we tabulated tumour human papillomavirus genotypes distribution by histological type, classified as positive or negative for HPV16, HPV18, and other high risk human papillomavirus. We further examined the association with screening by tumour human papillomavirus status. We also evaluated attendance for screening and risk of RICC, analysing by glassy cell carcinoma, clear cell carcinoma, and other rare types in a sensitivity analysis. All data were linked at the individual level on the basis of a sequential study number corresponding to the Swedish personal identity number, which allows for virtually 100% coverage.17 All statistical tests were two sided, and we used SAS 9.4 for data management and statistical analysis.
Patient and public involvement
No patients were involved in setting the research question or the outcome measures, nor were they involved in developing plans for design or implementation of the study. No patients were asked to advise on interpretation or writing up of results. Our findings will be disseminated through updates of the National Swedish Guidelines for Cervical Cancer Prevention, Diagnosis, and Treatment.18 These guidelines are written by expert groups and are updated regularly to reflect advances in the scientific evidence. Patients and professional organisations are involved in reviewing and commenting on these guidelines before they are adopted into practice.
Characteristics of ASC and RICC
Among the 338 cases, 164 (49%) were ASC, 43 (13%) were glassy cell carcinoma, 31 (9%) were clear cell carcinoma, and 100 (30%) were other rare types (table 1). ASC and RICC were mostly (178/338; 53%) diagnosed in women aged between 30 and 60, and only 9% (31/338) of the cases were diagnosed before age 30. Five per cent (17/338) of all cases were stage IA (microinvasive), 51% (172/338) were stage IB (localised), and 44% (149/338) were stage II+ (advanced). Most (297/338; 88%) cases were symptomatic cancers, and this proportion was even higher for CCC (30/31; 97%) and other rare types of carcinoma (97/100; 97%).
Screening status and risk of ASC and RICC
Compared with having no test in either of the previous two screening intervals, having two tests was associated with a substantially lower risk of ASC (incidence rate ratio 0.22, 95% confidence interval 0.14 to 0.34) and RICC (0.34, 0.21 to 0.55) (table 2). The risk reduction was more pronounced for women with two tests than for women with only one test. Overall, the risk reduction was greatest among women at age 30-60 (ASC: incidence rate ratio 0.21 0.13 to 0.34; RICC: 0.22, 0.12 to 0.40). Stratified by cancer stage, screening was associated with a significantly reduced risk of stage IB+ ASC through both one test (stage IB: incidence rate ratio 0.38, 0.22 to 0.66; stage II+: 0.29, 0.14 to 0.60) and two tests (stage IB: 0.23, 0.13 to 0.39; stage II+: 0.15, 0.07 to 0.33). For RICC, screening was related to a significantly reduced risk of stage II+ RICC through both one test (incidence rate ratio 0.44, 0.24 to 0.81) and two tests (0.32, 0.17 to 0.61) but a significantly reduced risk of stage IB RICC only through two tests (0.34, 0.16 to 0.75).
Screening history and risk of ASC and RICC
Women with two normal results in the previous two screening intervals had significantly lower risk of ASC (incidence rate ratio 0.16, 0.10 to 0.26) and RICC (0.29, 0.17 to 0.48). The risk reduction was also noted for women with only one test with normal result (incidence rate ratio 0.34, 0.22 to 0.53, for ASC; 0.63, 0.41 to 0.99, for RICC) (table 3). However, women with at least one abnormal test in any of the two intervals had an elevated, although not statistically significantly, risk of ASC incidence rate ratio 1.35, 0.78 to 2.37) and RICC (1.83, 0.92 to 3.66). The risk reduction in relation to two normal test results was larger for ASC than RICC and did not differ by age or cancer stage in general.
Tumour human papillomavirus status, screening status, and risk of ASC and RICC
High risk human papillomavirus was detected in 148/211 (70%) cases, among which HPV18 (79/211; 37%) was the dominant type, followed by HPV16 (47/211; 22%) and other high risk human papillomavirus types (22/211; 10%) (table 4). We observed a similar risk reduction associated with screening for cases positive (one test: incidence rate ratio 0.42, 0.26 to 0.67; two tests: 0.28, 0.18 to 0.46) and negative (one test: 0.46, 0.23 to 0.92; two tests: 0.27, 0.13 to 0.59) for high risk human papillomavirus, comparing women who attended one or both screening tests with women who did not attend any (table 5).
When we stratified by each histological type of RICC, we found that having two tests was also related to reduced risks of glassy cell carcinoma (incidence rate ratio 0.24, 0.09 to 0.68), clear cell carcinoma (0.17, 0.05 to 0.57), and other rare types (0.44, 0.24 to 0.83) (supplementary table E).
Women who routinely attended for cervical screening according to recommended intervals had a significantly reduced risk of both ASC and RICC compared with those who did not attend for screening. The risk reduction was more pronounced for women with two tests than for women with only one test. The magnitude of risk reduction in relation to cervical screening was less for RICC than for ASC. Moreover, attending screening was associated with a significantly decreased risk of advanced stage cancers and substantial down-staging for ASC and RICC.
Interpretation of findings
RICC tended to be diagnosed at a more advanced stage than the other histological types,2 which might subsequently result in a worse prognosis as previously reported.1 RICC may be a group of invasive cervical carcinomas with distinctive histological types compared with ASC, given that RICC have a rapid progression,5 and we found that the risk reduction in relation to one test in the previous screening intervals was not significant. Of note, as we defined it, RICC consist of several specific histological types, which are rare variants of adenocarcinoma. As a result, they might be more likely to have glandular precancerous lesions, which are located in the endocervical canal and typically more difficult to detect through cytology and manage through colposcopy assessment compared with squamous cell lesions.
Our finding of a greater risk reduction in women who had two tests compared with one test in the previous two screening intervals confirmed the general principle that the sensitivity of cytology is improved through repeated tests. Attending all recommended screening tests can increase the probability of existing precursors being detected. When we stratified the analyses by age at diagnosis of cancer, attending screening was related to decreased risk of ASC and RICC among women aged below 60 compared with women with no test, and we had limited power to estimate screening for women aged above 60 in this study. However, a previous study showed that the risk of invasive cervical cancer after age 60 was highly related to screening at age 51-60.19 Most (13/17; 76%) of the stage IA cases in our study were detected through screening, so it was not unexpected that women with screening tests showed an increased risk of cancer at stage IA compared with women without any tests, whose cancers are usually diagnosed at more advanced stages.
In terms of screening history, our results highlight that having normal tests in the previous two screening intervals was associated with a reduction in the risk of ASC and RICC, especially for women at age 30-60. In contrast, having at least one abnormal test was associated with increased risks of both ASC and RICC compared with women without any test. This could be due to the inadequate execution of the management practice after an abnormality.2 In addition, clinical management of abnormal smears could alter the risk of ASC and RICC after an abnormal smear, and timely assessment through biopsies or treatment (if needed) could be essential for histological types that progress rapidly. Also, the recommended management strategy might not be perfectly effective for certain scenarios of risk elevation.2021
The proportions of positivity for high risk human papillomavirus in ASC and RICC were somewhat lower than for squamous cell carcinoma and adenocarcinoma.22 Recently, a meta-analysis showed that adenocarcinoma had a lower prevalence of high risk human papillomavirus compared with adenocarcinoma in situ.23 Some variants of RICC were reported as having a low prevalence of human papillomavirus or even as not being related to human papillomavirus.2425 However, the lower proportion of positivity for high risk human papillomavirus noted in our study does not have to signify that this is not the main cause of ASC and RICC. On the contrary, it indicates that human papillomavirus might become undetectable in advanced cancers.2627 Therefore, the fact that negativity for high risk human papillomavirus is preferentially found in advanced and symptomatic cases might have resulted in a greater proportion of such cases among RICC. In another sample tested with the same method, we found that 97% of women with cervical intraepithelial neoplasia 3+ were positive for high risk human papillomavirus.28
Comparisons with other studies
Our findings on risk reduction of ASC are similar to those of UK studies.34 However, no previous studies have looked at the association between screening and risk of RICC owing to the rarity of these diagnoses and limited information on cervical screening. Compared with squamous cell carcinoma and adenocarcinoma, we saw a similar trend of risk reduction associated with cervical screening in ASC and RICC as well as a down-staging effect.2329 The risk reduction for ASC associated with cervical screening was similar to that for squamous cell carcinoma,2 whereas the risk reductions for RICC and adenocarcinoma were attenuated (supplementary table F). The smaller magnitude of risk reduction for RICC might explain why RICC tend to be diagnosed at a more advanced stage and to be symptomatic compared with the more common types.2 Similar risk reductions for ASC and squamous cell carcinoma also suggest that ASC might evolve from a squamous component and acquire glandular involvement later on,4 or it might be a glandular cancer exhibiting squamous differentiation as shown in previous studies.24
Strengths and limitations of study
Although this is the largest study to date of ASC and RICC, and we included all cases over a 10 year period, further assessment of clinical management after abnormal tests was still not feasible owing to the limited number of cases. The small number of cases resulted in reduced precision in some of the subgroup analyses. We did not have information on lifestyle factors (such as smoking status) and sexually transmitted infections. However, we had virtually complete information on educational level, which is a good proxy indicator for both socioeconomic status and lifestyle factors.30 We thus controlled for educational level in our analysis, and the association between screening and risk of ASC and RICC remained robust. Nevertheless, a healthy volunteer effect for women participating in cervical screening cannot be ruled out. If existing, it should be fairly small and not likely to account for the risk differences observed in this study. Finally, misclassification of histological types might have occurred owing to the overlap of morphology, but we used strict clinical and histopathological review to ensure accurate classification.
To the best of our knowledge, this is the first population based study to examine cervical screening and risk of ASC and RICC and to present the prevalence of human papillomavirus in tumours. All cases were strictly reviewed according to a cervical cancer case audit protocol. We used complete screening data extracted from the Swedish National Cervical Screening Registry with limited selection and information bias. Moreover, all controls were selected randomly from the Swedish Total Population Register and individually matched to the cases on year of birth, further eliminating selection bias and confounding by age. High quality information on hysterectomy, death, and migration limited bias due to loss of follow-up. We used a xylene-free extraction method that has been shown to be robust and result in improved detectability of human papillomavirus compared with standard xylene methods.31 This method is highly sensitive for human papillomavirus genotyping in formalin fixed, paraffin embedded material,11 and it was used with strict quality control to avoid contamination with human papillomavirus and ensure the accuracy of the detected genotypes. Our results should be generalisable to countries or regions with similar screening programmes and settings.
As the great majority of ASC and RICC positive for high risk human papillomavirus were HPV16/HPV18 positive, vaccination against human papillomavirus with bivalent or quadrivalent vaccines will be a significant strategy for prevention. With the nine-valent human papillomavirus vaccine, which covers five additional virus types, additional prevention of ASC and RICC could be expected in future. European guidelines from 2015 recommend human papillomavirus as a primary testing method for organised, population based screening in women at age 30 and above.32 Continued monitoring of ASC and RICC is necessary in the post-human papillomavirus vaccination era and after the switch to primary human papillomavirus testing.
Cervical screening is associated with a reduced risk of both ASC and RICC, especially for advanced stage cancers, which gives a benchmark for evaluating future cervical screening strategies. As most of ASC and RICC were positive for high risk human papillomavirus, the switch to primary human papillomavirus screening and prevention by vaccination are also expected to decrease the risk of ASC and RICC, but this will need to be monitored.
What is already known on this topic
Evidence on cervical screening and risk of invasive cervical carcinoma that is neither squamous cell carcinoma nor adenocarcinoma is limited
Cervical screening is related to lower risk of adenosquamous cell carcinoma (ASC), but literature on the rare histological types of invasive cervical carcinoma (RICC) is absent
The distribution of human papillomavirus varies between different histological types of invasive cervical carcinoma, and its distribution for ASC and RICC is not well known
What this study adds
Women who attended screening according to routinely recommended intervals had a significantly reduced risk of both ASC and RICC
The magnitude of risk reduction in relation to cervical screening was less for RICC than for ASC
Most of the ASC and RICC tumours were positive for high risk human papillomavirus
We thank all study participants who contributed data to our research. We acknowledge Walter Ryd for histopathological review and Pouran Almstedt for data management.
Contributors: JL did the literature search and review. BA, JD, and PS conceived the research questions and hypotheses and designed the study. PS, JD, and SNK collected diagnostic samples. BA reviewed medical charts and supported histopathological classification of all cases. CL and CE did the laboratory analysis on human papillomavirus genotyping and validation. JL and AP did statistical analysis, and JL, BA, AP, JD, FF, KME, and PS interpreted the data. JL wrote the original draft of this manuscript. All authors were involved in the critical revision of article and have read and approved submission of the manuscript. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. PS is the guarantor.
Funding: This work was supported by the Swedish Foundation for Strategic Research (KF10-0046 and RB13-0011), the Swedish Cancer Society (11 0569, 14 0665, and 16 0699), the Swedish Research Council (2014-03732 and 2017-02346), the China Scholarship Council (201507930001), and the Centre for Research and Development, Uppsala University/Region of Gävleborg. The study sponsors did not participate in study design, data collection, analysis, interpretation of data, writing of the article, or the decision to submit it for publication.
Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: JD has obtained grants to his institution from Roche and Genomica for research on human papillomavirus tests; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work.
Ethical approval: This study was approved by the Regional Ethical Review Board in Stockholm, which determined that, owing to the population based nature of the study, informed consent from the study participants was not needed (Dnr 2011/1026-31/4; Dnr 02-556; Dnr 2012/1028/32; Dnr 2011/921-32).
Data sharing: Data from the study are available on request from corresponding author.
Transparency declaration: The lead author (PS) affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.