Atypical glandular lesions of the cervix and risk of cervical cancer

Cytology screening has been effective in reducing risks for cervical squamous cell carcinoma but less so for adenocarcinoma. We explored the association of atypical glandular cells or absence of glandular cells in cytology, and subsequent histological diagnoses and cancer risk.


| INTRODUC TI ON
Cervical cancer is the fourth most common cancer among women worldwide, with around 570 000 new cases and 310 000 deaths in 2018. 1 Well-organized cervical cancer screening programs have produced profound decreases in incidence and mortality of the disease. 2,3 In Norway, a decline in cervical cancer incidence and mortality was seen after the implementation of organized screening in 1995, but incidence rates seem to have increased during recent years. 3 The Norwegian Cervical Cancer Screening Program (NCCSP) invites all women aged 25-69 years to screening with cytological smears every 3 years. 4 The screening coverage in the target age group is 69% within 3.5 years (2013-2016). 3 Low-grade squamous cell cytological abnormalities are triaged with human papillomavirus (HPV) testing and intensified screening, whereas high-grade cytological abnormalities warrant immediate referral to colposcopy and biopsy. 5 Randomized implementation of HPV primary screening started in four Norwegian counties in 2015, 6 and will be gradually implemented in the remaining counties during 2019-2021.
In some settings, cervical screening using cytology has been reported to reduce the risk of invasive adenocarcinomas, 7,8 and lead to earlier detection of this type of cancer. 9 However, cytology has been much more successful in reducing the risk of developing squamous cell carcinoma. 2,8,10 Consequently, the relative contribution of these two main histological subtypes to the total cervical cancer burden has been changing over decades in countries with effective screening programs. 11 The proportions of cytological smears that display dysplasia and/or abnormalities vary considerably between countries (0.98%-15.5%). [12][13][14][15][16] Histological diagnosis of adenocarcinoma in situ (ACIS) can follow either glandular or squamous cytological abnormalities. In American material, about 3%-4% of women with abnormal cytology have ACIS in histology. 17,18 Of all cervical smears, 0.1%-2.1% are classified as atypical glandular cells of undetermined significance (AGUS). 17,19,20 A significant proportion of women with such smears have underlying cancer, or will develop cancer during follow-up. 20 Among women with AGUS in cytology in a US primary-care study, 19.5% had cancerous squamous or glandular lesions of the cervix or endometrium and 11.5% had pre-cancerous squamous or glandular lesions. 21 In a Swedish cohort study, women with atypical glandular cells (AGC), equivalent to AGUS, in cervical screening had a higher risk of incident cervical carcinoma, especially adenocarcinoma, than women with a high-grade squamous intraepithelial lesion. 22 The cumulative incidence of invasive cervical cancer was persistently elevated for up to 15.5 years following AGC in cytology.
Smears labelled as normal/benign cells, but with absence of endocervical or metaplastic cells (NC-NEC) in our study show normal cells, but lack endocervical or metaplastic cells. As such, the condition of endocervical or metaplastic cells cannot be evaluated in these samples.
NC-NEC smears are managed as normal smears in NCCSP. 5 A Canadian review from 2011 suggested that the absence of endocervical cells in smears does not indicate a higher risk for underlying cervical abnormalities, 23 but little is known in the Norwegian setting.
This study aimed to describe the association between atypical glandular changes in cervical cytology; AGUS, ACIS and NC-NEC, and subsequent diagnoses verified by histology. We also examined the risk of developing cervical cancer by histological subtype and other gynecological malignancies over time.

| Study population
This study included all women in Norway with AGC in their first registered cervical cytology (AGUS and ACIS) since 1992, and all women with NC-NEC in their first cervical cytology since 2005 (when this category was adopted) (n = 142 445) ( Figure 1). We

Key message
There is a high risk of cervical adenocarcinoma after atypical glandular cells in cytology. Absence of glandular cells in cytology did not entail a higher cervical cancer risk. excluded women who had a diagnosis of invasive cervical cancer or other gynecological malignancies before their first cytology (n = 2017). The Nelson-Aalen cumulative hazard function for risk of gynecological cancers, with 95% CIs, was calculated. 27 We identified all

| Characteristics of the study population
The median ages at ACIS, AGUS and NC-NEC were 43, 44 and 45 years, respectively. Within 1 and 3 years of cytology, respectively, 91% and 92% of ACIS, 76% and 78% of AGUS, and 2% and 3% of NC-NEC smears were followed by a histological diagnosis (2002-2014).

| Association between cytological smears and histological diagnoses
Dysplastic lesions, premalignant lesions and cervical malignancies such as adenocarcinomas and squamous cell carcinomas were more frequent after ACIS cytology, and there were no major differences after 1 year (74.3%) and 3 years of cytology (75.4%). Following AGUS and ACIS cytology, 55.6% and 20.2% of the histological diagnoses within 3 years were normal or benign, respectively. CIN3 accounted for the majority of the CIN lesions following both AGUS and ACIS, whereas CIN1/2 were more frequent following AGUS cytology than ACIS after both 1 and 3 years. Overall, CIN1-3 accounted for 26.7% and 28.6% of the histological diagnoses within 1 and 3 years of AGUS, and 27.3% and 26.6% within 1 and 3 years of ACIS cytology (Tables 1 and 2). Following AGUS cytology, 38.7% and 41.7% of the histological diagnoses were dysplastic, premalignant and malignant (about 25% were CIN3, ACIS or more severe within 1 and 3 years).

| Association between cytological smears and histological diagnoses, stratified by age
Following ACIS cytology, the proportion of CIN1 and CIN2 histology was similar in women aged <35 and ≥35 years (Tables 3 and 4).
ACIS was the most common histology diagnosis in both age groups.
Malignant lesions had a higher proportion in women ≥35 than in women <35 years following ACIS cytology. There was a larger proportion of CIN lesions overall in women <35 than in women ≥35 years following AGUS cytology. Histology diagnosis of invasive adenocarcinoma was more common in women ≥35 years, whereas the proportion of squamous cell carcinoma following AGUS cytology was similar in women <35 and ≥35 years (Tables 3 and 4). Normal and benign was the most common histology after NC-NEC in both women <35 years (over 50% within 1 and 3 years) and ≥35 years (over 80% within 1 and 3 years) (Tables 3 and 4). There were few histological diagnoses of ACIS following NC-NEC cytology within 3 years in women <35 (0.8%) and ≥35 (0.2%) years (Table 4). Malignant lesions (squamous cell carcinoma and adenocarcinoma) were rare in both age groups.

| Cumulative risk of gynecological cancer according to cervical cytology
A total of 121 023 women were followed from 3 months after the cervical smear (Table 5)

| D ISCUSS I ON
In this nationwide population-based cohort study, we described the association between atypical glandular changes in cervical cytology and subsequent diagnoses verified by histology. We also exam-

NC-NEC A GUS A CIS
verification bias. The indication for biopsy in this group of women is not known but may relate to other risk factors such as genital symptoms or clinical findings, and the rate of normal histology in the total population of NC-NEC women is therefore likely to be higher.
Our study did not include information on hysterectomy and/or oophorectomy. The rates of hysterectomy are, however, lower in Norway (1.2 per 1000) than other western countries, such as the USA (5.4 per 1000) or Italy (3.7 per 1000). 36 Follow-up started 3 months after the cervical smear when we evaluated the cumulative risk of gynecological cancer to leave out most prevalent cancers diagnosed immediately after the index screen from the analyses. The slope of the cumulative hazard curves show that diagnostic events are frequent in the beginning of follow-up, so the absolute levels of observed risk are somewhat sensitive to the start of follow-up. Starting follow-up at 6 months instead of 3 months as a sensitivity analysis, the pattern of risk was similar, but with an up to 20% decrease in the absolute level (in cervical cancer after ACIS). Some prevalent cancers are still likely to be included in the analysis.

| CON CLUS ION
In this nationwide population-based cohort study, 78% of AGUS and 92% of ACIS smears were followed by histology within 3 years.
Of these, 42% of the histological findings following AGUS, and 75% following ACIS were dysplastic, premalignant or malignant.
Approximately 27% and 71% of the histological findings were CIN3, ACIS or more severe (including metastases) following AGUS and ACIS cytology, respectively. Only 3% of the NC-NEC smears were followed with histology within 3 years, and of these, 77% were normal or benign. During follow-up, 6.5% and 2.6% of women with ACIS and AGUS smears developed gynecological cancer, respectively, and cervical adenocarcinoma was the most common subtype. This indicates that a considerable risk is associated with ACIS and AGUS cytology, and such findings warrant careful gynecological evaluation.
According to this study, closer follow-up of women with NC-NEC may not be required.

CO N FLI C T O F I NTE R E S T
The authors have stated explicitly that there are no conflicts of interest in connection with this article.