Early Detection and Diagnosis
Decreasing incidence of cervical adenocarcinoma in Ontario: Is this related to improved endocervical Pap test sampling?
Version of Record online: 27 OCT 2006
Copyright © 2006 Wiley-Liss, Inc.
International Journal of Cancer
Volume 120, Issue 2, pages 362–367, 15 January 2007
How to Cite
Howlett, R. I., Marrett, L. D., Innes, M. K., Rosen, B. P. and McLachlin, C. M. (2007), Decreasing incidence of cervical adenocarcinoma in Ontario: Is this related to improved endocervical Pap test sampling?. Int. J. Cancer, 120: 362–367. doi: 10.1002/ijc.22171
- Issue online: 28 NOV 2006
- Version of Record online: 27 OCT 2006
- Manuscript Accepted: 31 MAY 2006
- Manuscript Received: 22 NOV 2005
- adenocarcinoma of cervix uteri;
- cervical cancer;
- endocervical brush;
- endocervical screening
In many developed countries, the incidence of cervical cancer has decreased. These reductions have been specific to squamous cell carcinoma (SCC) and have not included adenocarcinoma (AC). Incidence of AC has increased steadily over the last 20 years. The intent of this article is to examine trends in cervical adenocarcinoma incidence in Ontario over a 20-year period in relation to screening practices. All cases of cervical cancer between 1981 and 2002 were extracted from the Ontario Cancer Registry (a population-based, provincial-wide database). Age-standardized incidence rates were calculated overall, by broad age groups and by morphological type (SCC and AC). Time trends were assessed using JoinPoint methodology. In Ontario, opportunistic cervical cancer screening has been accompanied by significantly decreased rates of SCC since at least 1981. Conversely, the incidence of AC rose by 3.1% per year (95% CI: 1.6%, 4.6%) between 1981 and 1995, and subsequently declined by 4.0% per year (95% CI: −7.4%, −0.5%). From the mid- to late-1990s, instructions were distributed to clinicians, reinforcing the importance of dual specimen collection (i.e., using both spatula and endocervical brush). At the same time, laboratories routinely provided physicians with kits that included both spatula and brush. The subsequent decline in AC incidence may be due, in part, to improved specimen collection. As well, the decline may be partly due to increased awareness of AC precursors among cytopathologists and clinicians, and/or improvements in laboratory training and quality assurance. © 2006 Wiley-Liss, Inc.
Cervical cytologic screening has been successful in reducing the incidence of cervical cancer in many developed countries.1, 2, 3, 4, 5, 6, 7, 8, 9 Jurisdictions with organized screening have realized even greater reductions.10, 11 These reductions have been attributed to declining rates of the most common type of cervical cancer, squamous cell carcinoma (SCC).9, 12
Cervical adenocarcinoma (AC) is less common but increasing incidence has been reported in several developed countries around the world,13 including Australia,14 United States,3, 15 Sweden16 and Norway.17 Trends in the incidence of adenocarcinoma vary with age and geography. The literature reveals a doubling of rates in the United States3 and Sweden,16 a 3-fold increase in Norway,17 a 4-fold increase in Canada18 and a 9-fold increase in England.4 The most notable increases were among women born in the mid-1930s, with other cohorts identified in subsequent years in many countries.13 Several hypotheses have been proposed, and there has been a great deal of speculation, as to the cause of increasing rates of AC in developed countries. Some have noted cohort, period or age effects, especially among younger women.4, 13, 15, 18
It is generally accepted that cervical cytology in the context of an organized screening program is even more successful than spontaneous screening in detecting the precursors of squamous cancer but still has a limited efficacy in detecting glandular precursors. For example, in Finland,10, 11 rates of AC have remained stable in spite of the tremendous success of organized screening to reduce squamous cell carcinoma. The lack of success in detecting glandular precursors is due, in part, to the anatomy of the endocervical glands, which makes it more difficult to sample the lining cells. It is also due to the difficulty in distinguishing preneoplastic glandular cells from myriad reactive changes.19
A few studies20, 21 have concluded that screening does offer some protection against AC and AIS. This suggests that women with negative screening histories were protected from AC. The impact of screening has generally resulted in increased detection of cancer precursors and increased incidence, with an earlier diagnosis, followed by a period of decreased incidence.20, 21, 22
Conversely, other authors reported that screening does not offer effective protection against AC and/or AIS. Bulk et al.23 concluded that screening is ineffective in detecting AC and precursors in Amsterdam. Incidence of AC remained stable over a 12-year period, but risk of death with AC was 1.6 times higher than with SCC. Andersson et al.24 reported only 23% of AC was detected by screening in Sweden. Of those diagnosed, 93% had a normal Pap test in the previous 3 years, while 60% had a normal Pap test in the preceding year. Other studies have reported similar findings.3, 8, 16, 18, 25
An increased rate of adenocarcinoma had been noted in Ontario between 1979 and 1996,1 when AC increased 63%, from 1.8 to 3.0/100,000. After correcting for hysterectomy, rates increased annually by 5.8% in women aged 20–34 and 35–49. Among women over 50, a smaller annual increase of 0.8% was reported.1 During the 1990s, there was a growing recognition of the need to improve endocervical sampling and glandular cytology reporting and follow-up.
The purpose of this study was to examine incidence of cervical adenocarcinoma in Ontario before and after introduction (from the mid- to late-1990s) of new terminology, improved endocervical sampling (dual specimen collection) and quality assurance efforts.
Material and methods
Incident cases of cervical cancer26 were extracted from the Ontario Cancer Registry (OCR) for the period 1981–2002. Only women under 80 years of age were included, in part because data to estimate true population at risk (i.e., by virtue of having a cervix) were unreliable or unavailable and in part because screening activities are generally concentrated on somewhat younger women. The OCR, operated by Cancer Care Ontario (CCO), is a population-based registry containing all cases of invasive cancer (except nonmelanoma skin cancers), newly diagnosed since 1964. Details about the operation of the OCR have been published previously.27, 28, 29
Hysterectomy-corrected population at risk for cervical cancer
Ontario women at risk for cervical cancer are those who have not had a hysterectomy that includes removal of the cervix. Thus, a more accurate estimate of the rates of cervical cancer is based on total women minus the number estimated to have had a prior hysterectomy. Cumulative probabilities of having had a prior hysterectomy were estimated for Ontario women by age and year for the time period 1980–97 by Holowaty.30 For the present analysis, we extended these probabilities to 1998–2002. First, we estimated the annual age-specific incidence of hysterectomy by averaging the age-specific rates for the previous 5 years, then we added this estimate to the cumulative probability for the next younger age in the previous year. These incidence data were secured from hospital discharge data provided by the Canadian Institute for Health Information (CIHI). For example, the probability of a 50-year-old woman having a hysterectomy in 1998 was assumed to be the average of the hysterectomy incidence rates for 50-year-old women in 1993–1997. This probability was then added to the cumulative probability of prior hysterectomy for 49-year-old women in 1997 to get an estimate of the percentage of 50-year-old women who would have had a prior hysterectomy in 1998.
Morphology in the OCR is coded according to the International Classification of Diseases for Oncology (ICD-O, 1st and 2nd editions).31, 32 Morphologic types were grouped for analysis using a variant of Parkin et al.33 and others.34, 35 SCC was defined as ICD-O codes 8051–8076, and AC was defined as ICD-O codes 8050, 8140–8510, 8560 and 8570. Adenosquamous carcinoma was included with adenocarcinoma because it was too infrequent to treat it as a separate category for analysis.
Cancer incidence rates and trend analysis
All ages and age-specific incidence rates (per 100,000 women) were calculated using the SEER*Stat statistical software package (version 6.1.4).36 All rates were age-standardized to the 1991 Canadian population, by the direct method. Rates were calculated for ages 0–79 and for the specific age groups 20–34, 35–49, 50–64 and 65–79.
Trend analysis was carried out on annual age-standardized rates for all ages 0–79 and for specific age groups.37 This allows estimation of the annual percentage change (APC) in incidence rates and tests for significant changes in trend using a number of contiguous linear segments and “join points.” Permutation tests were used to identify the best-fitting join point regression, initially testing the null hypothesis (no join points) versus the alternative (3 join points), then decreasing the number of join points tested if the null was not rejected. Each test was performed with a Bonferroni correction (α/3) and an overall significance level of 0.05. The maximum number of join points tested was 3 in each analysis. Joinpoint software (version 3.0; National Cancer Institute, Bethesda, MD) was used to conduct the trend analysis.
In Ontario, overall incidence of cancer of the cervix (in women 0–79 years of age) fell by 1.6% per year (95% confidence interval [CI]: −2.1%, −1.0%) to 1996 and by 4.6% per year (95% CI: −6.8%, −2.4%) thereafter, for a total drop of 39% over the 22-year period (Fig. 1).
About 72% of cervical cancers are squamous cell carcinoma (SCC), whose incidence fell by 3.2% annually throughout the period 1981–2002 (95% CI: −3.6%, −2.8%). In contrast, the incidence of AC (including adenosquamous cancers), which represents 20% of cervical cancers, rose by 3.1% per year (95% CI: 1.6%, 4.6%) to 1995 and fell by 4.0% annually thereafter (95% CI: −7.4%, −0.5%). As a result, the ratio of incidence rates for SCC to AC dropped from over 6 in the early 1980s to less than 3 by 2000–2002.
The remaining cervical cancers are of other or unspecified morphologies, accounting for 7.5–8.5% of cervical cancers during the study period. The incidence rate for this group declined steadily over time (Table I), with an estimated annual percent change (EAPC) of −1.5% (95% CI: −2.6%, −0.3%). The trend in the incidence rate of AC plus other and unspecified morphologies parallels that of AC alone (Table I).
|Adenocarcinoma (including adenosquamous carcinoma)||2.2||2.6||3.0||2.4|
|Other and unspecified morphologies||1.1||1.0||0.9||0.9|
In the most recent quinquennium (1998–2002), 595 Ontario women (0–79 years) were diagnosed with cervical AC (Table II). The estimated percentage of women who had undergone a hysterectomy, and were therefore not at risk of cervical cancer, was 12% overall, ranging from 1% of young women (ages 20–34) to 39% in the age group 65–79. These estimates are very close to the self-reported prevalence of prior hysterectomy,38 derived from the Ontario portion of the 2003 Canadian Community Health Survey: 1% for ages 20–34; 9% for ages 35–49; 26% for ages 50–64 and 34% for ages 65–79.
|Age group||No. of cases||Person years1||% prior hysterectomy||Age-standardized rate2|
|All ages (0–79)||595||28,557,570||12||2.4|
Among young women (ages 20–34), the incidence of AC rose by 3.3% annually (95% CI: 1.4%, 5.3%) (Fig. 2 and Table III); rates appear to have leveled off beginning in the late 1990s but no join point is detectable. Rates rose significantly in 35- to 49-year-old women until 1992 (EAPC: 5.2%; 95% CI: 1.3%, 9.2%) and then declined, although the fall is not quite significant at the p < 0.05 level (EAPC: −3.1; 95% CI: −6.0%, 0.03%). There is a lot of noise in the data for the older age groups, presumably due to small numbers, although rates for both appear to be declining in recent years. Only the rise in incidence for 65- to 79-year-old women up to 1999 is statistically significant (EAPC: 2.2%; 95% CI: 0.2%, 4.2%).
|Age group||Years||EAPC (95% CI)1|
|20–34||1981–2002||3.3 (1.4, 5.3)|
|35–40||1981–1992||5.2 (1.3, 9.2)|
|1992–2002||−3.1 (−6.0, +0.0)|
|50–64||1981–2002||−0.4 (−2.1, 1.3)|
|65–79||1981–1999||2.2 (0.2, 4.2)|
|1999–2002||−23.8 (−45.4, 6.2)|
|All ages (0–79)||1981–1995||3.1 (1.6, 4.6)|
|1995–2002||−4.0 (−7.4, −0.5)|
Similar to other developed countries, our study shows that the incidence of adenocarcinoma (AC) among Ontario women under the age of 80 years increased significantly by 3.1% per year until 1995 (95% CI: 1.6%, 4.6%). Unlike other jurisdictions, this was followed by a significant decrease, by 4.0% (95% CI: −7.4, −0.5%) annually thereafter. This decline in incidence may be a result of various quality assurance initiatives that were implemented in Ontario in the early- to mid-1990s.
Pap test sampling techniques have changed significantly since the 1980s, largely with the addition of sampling techniques designed to improve endocervical specimens. For many decades, specimen collection for Pap test screening depended on clinician's use of just one collection device—a spatula. This instrument was primarily designed to collect cells from the outer part of the cervix. Since this is where SCC occurs, Pap testing has been most effective for detection of squamous precursors. Consequently, decreased incidence of cervical cancer has been specific to squamous cell carcinoma.
AC arises in the endocervical canal (above the transformation zone), which is not easily accessible with the spatula. Furthermore, most of the cells lining the endocervical crypts are below the surface and not directly accessible to sampling. In the 1990s, brush devices were introduced to improve sampling of the endocervix. It has been hypothesized that improved sampling, recognition and awareness of glandular lesions in cytology screening resulted in higher AIS rates, and that the delayed impact of screening would eventually translate into decreased AC rates.8
An Australian study by Schoolland et al.40 examined the effect of endocervical brushes on collection of cells from the transformation zone. Schoolland referenced unpublished data indicating that the percentage of Pap tests with endocervical cells had increased to 85–90% from an estimated 50–60% prior to introduction of new collection devices. Seven studies suggested the potential benefits of improved sampling due to use of endocervical brush and better identification of precursor lesions, thereby reducing progression to adenocarcinoma.8, 13, 21, 40, 41, 42, 43
In Ontario, the endobrush was first recommended in 1989 through the publication of “The Adequate Pap Smear,” which detailed proper collection techniques.44 Many clinicians had already adopted this format several years prior to the recommendations.45 A 1992 survey46 by Laboratory Proficiency Testing Program reported that 51% of Ontario laboratories were receiving Pap tests that were obtained using endobrush sampling, in combination with a spatula. By 1997, dual specimen collection with both the extended-tip spatula and an endocervical brush was the standard of practice in Ontario. Since the 1997 inception of the Ontario database (which captures about 85% of Pap test results in the province), ∼85% of Pap tests have contained transformation zone component.47
In 1996, modified Bethesda terminology was published and distributed to all Ontario cytology laboratories. Revisions to the US Bethesda terminology included a diagnosis of atypical glandular cells of undetermined significance and atypical glandular cells, consistent with adenocarcinoma in situ. Recommended follow-up for both these diagnoses was referral to colposcopy. Furthermore, physicians, cytotechnologists and cytopathologists have been more alert to the diagnosis of AC and its precursors, as a result of quality assurance efforts in Ontario and widespread distribution of instructions on specimen collection, as well as screening and follow-up guidelines. All 3 factors—dual specimen collection, modified terminology and quality assurance efforts—coincide with a temporary increase followed by a decrease in AC rates. There may be a connection between these events and the declining incidence of AC.
These 3 factors may have been responsible for both the increased incidence of the disease (up to 1995), as well as improved detection (and follow-up) of glandular precursors and AIS, thereby preventing progression of abnormalities to AC. The subsequent declining incidence was initially modest in 1997, but seems to have accelerated over time (Fig. 2). A significant decrease was noted only when all ages are combined; this is likely due to small numbers in each age range. The trend (of increase followed by a decrease) may be due to a combination of some or all of these factors (1): increased detection of prevalent cases of cancer and detection of more cancer precursors and (2) true increasing rates until the mid-1990s, accompanied by improved screening, with detection of more cancer precursors.
Over the same time period (1981–2002), increased survival rates for AC have been noted.39 Improved survival rates support the hypothesis that improved screening has contributed to the decreased incidence of adenocarcinoma. Screening could result in both earlier diagnosis, which would have better prognosis, and some lead time bias, both of which would result in better survival. Improvement in treatment is a less likely explanation for better survival given that survival for squamous cell carcinoma remains largely unchanged.
Some13 have speculated that there is a relationship between adenocarcinoma and long-term use of oral contraceptives (OC), especially high dose estrogen that was more common when OC were first introduced. Some24, 48 have reported no correlation between adenocarcinoma and oral contraceptives (OC). Another study43 reported that risk for AIS doubled among those who ever used OC, but that risk rose more than 5-fold with long-term use of oral contraceptives (>11 years). While we have no data on consumption or duration of OC specific to the Ontario population, national survey data49, 50, 51, 52 and provincial drug plan data53 suggest that use has remained constant over time. Consequently, it is unlikely that changing trends in adenocarcinoma incidence are a function of changing use of oral contraceptives over time.
Our review includes a population that is fairly well-screened. Based on self-reported data, about 80% of Ontario women had a Pap test in the most recent 3-year period.54 Compared to other jurisdictions, there is a relatively low incidence of cervical cancer, 7.8/100, 00039 (women of all ages, uncorrected for hysterectomy). We have documentation of specific events in history that seem to have influenced decreasing incidence of AC, e.g., endocervical brush, quality assurance programs and guideline distribution. Even though we are aware of these specific events in time, we cannot infer any causal relationships.
AC is uncommon. Therefore, in spite of a large population base, our review is based on a relatively small sample. In the future, it will be necessary to monitor incidence rates of AC in other jurisdictions that have implemented dual specimen collection, to see if these findings are replicated. Similarly, close monitoring of AC incidence in Ontario and other jurisdictions that have implemented LBC will also provide important information either to support or refute the hypothesis outlined in this study, regarding the association of dual specimen collection and declining rates of AC. With the introduction of liquid-based cytology (LBC) in Ontario in 2001, it is likely that the benefits of dual specimen collection via conventional cytology will persist. Several studies suggest that LBC is equally or more effective in detecting glandular lesions,55 even though still difficult to detect with any cytology method.19, 56, 57, 58
The strengths of this study relate to the fact that the OCR is a well-established population-based cancer registry for a population of 5.7 million females, up to 79 years of age. The ability to correct for long-term changes in hysterectomy rates enables reporting of true rates; the lack of correction in other studies is a limitation resulting in incidence rates that are too low and trends that are possibly biased. The estimates for hysterectomy used in this study are very similar to self-reported Ontario data from a recent national survey (unpublished data). The long history of spontaneous cervical cancer screening in Ontario with appropriate quality controls (enabled by an organization that is at “arms-length” from the laboratories and provincial government) implies that the OCR probably contains fairly accurate morphology data.
The lack of information about the incidence of AIS over the same time period is a limiting factor. Ideally, it would have been helpful to monitor the incidence of AIS to assess the trend in comparison to the incidence of AC.
This study suggests that 3 factors—improved (dual) specimen collection, improved cytological diagnosis, as well as increased awareness of AC and precursors among clinicians and laboratory personnel—have been associated not only with arresting accelerated incidence (noted in many jurisdictions) but also with decreasing incidence of AC. This is further substantiated by the increased survival rates for AC, which may be due to earlier diagnosis. The benefits of regular Pap testing that includes endocervical sampling for prevention of AC of the cervix may have been underestimated. Continued monitoring of Pap test quality and ongoing education to include improved glandular detection should be a focus of cervical screening programs. While the outcomes of this study are not conclusive, these results add substance to the current literature, in a way that is more concrete than other speculative hypotheses.
The authors would like to thank Diane Nishri, Senior Research Associate, and Sandrene Chin Cheong, Junior Research Associate, of the Surveillance Unit, Division of Preventive Oncology, Cancer Care Ontario, for their assistance with analysis and presentation of results. The information in this study was analyzed and interpreted by staff (RIH, LDM and MI) of Cancer Care Ontario; co-authors external to Cancer Care Ontario (BR and CMM) contributed to this article independently. No external funding was received to complete this analysis/article. The authors certify that they have not entered into any agreement that could interfere with their access to the data on the research, nor upon their ability to analyze the data independently, to prepare manuscripts and to publish them.
- 1Cervical cancer in Ontario 1971–1996. Toronto: Cancer Care Ontario, 1999., , , .
- 12International Agency for Research on Cancer.IARC handbooks of cancer prevention, vol. 10: Cervix cancer screening. Lyon: IARC Press, 2005.
- 20Changing trends in adenocarcinoma and squamous cell carcinoma by age, race and stage: a 25-year population-based study. Gynecol Oncol 2001; 80: 295–296., , .
- 26World Health Organization. International classification of diseases—ninth revision, Geneva: World Health Organization, 1977.
- 27Cancer incidence in Ontario: trends and regional variations in the 1980s. Toronto: The Ontario Cancer Treatment and Research Foundation, 1995., , .
- 28Geographic distribution of cancer in Ontario, vol. 2: Atlas of cancer incidence, 1980–1991. Toronto: The Ontario Cancer Treatment and Research Foundation, 1995., , , et al.
- 29Cancer survival in Ontario. Toronto: The Ontario Cancer Treatment and Research Foundation, 1995., , .
- 30The natural history of cervical dysplasia. Dissertation. Toronto: University of Toronto, 1996..
- 31World Health Organization. Manual of the international classification of diseases for oncology, 1st edn. Geneva: World Health Organization, 1976.
- 32PercyC, Van HoltenV, MuirC, eds. International classification of diseases for oncology, 2nd edn. Geneva: World Health Organization, 1990.
- 33ParkinDM, ShanmugaratnamK, SobinL, et al., eds. Histological groups for comparative studies. Lyon, France: International Agency for Research on Cancer. 1998. IARC Technical Report No. 31.
- 35Morphologic classification of human cancer. In: SchottenfeldD, FraumeniJF,Jr, eds. Cancer epidemiology and prevention. New York: Oxford University Press, 1996. 28–44..
- 36National Cancer Institute. Surveillance research program. SEER, Stat software, version 6.1.4. Bethesda, MD: National Cancer Institute, 2005.
- 38Statistics Canada. Canadian community health survey (CCHS) 2003 (Cycle 2.1), share file. Ottawa: Statistics Canada, Health Statistics Division, 2005.
- 39Cancer Care Ontario. Insight on Cancer: News and Information on Cervical Cancer. Toronto: Canadian Cancer Society (Ontario Division), 2005., , , .
- 44Adequate “Pap” smears. Laboratory Proficiency Testing Program, Toronto, Ontario, Canada. Quality Management Program, Laboratory Services, 1989..
- 46Cytology, Pattern of Practice Survey Y-92. Laboratory Proficiency Testing Program, Toronto, Ontario, Canada: Quality Management Program, Laboratory Services, 1992.
- 47Cytobase Reports, INSCYTE Corporation. Available at www.inscyte.org/publications/reports.html. (Accessed 1 November 2005).
- 49Statistics Canada. National Population Health Survey (NPHS) 1994/1995, public use microdata file. Ottawa: Statistics Canada, Health Statistics Division, 1995.
- 50Statistics Canada. National Population Health Survey (NPHS) 1996/1997, public use microdata file. Ottawa: Statistics Canada, Health Statistics Division, 1998.
- 51Statistics Canada. National Population Health Survey (NPHS) 1998/1999, public use microdata file. Ottawa: Statistics Canada, Health Statistics Division, 2000.
- 54Statistics Canada. Canadian community health survey, Cycle 1.1, 2000–2001, Ottawa, 2002.