Loss to follow-up of women with initial cervical cytology of ASC-US/LSIL in Tocantins, Brazil

Jucimária Dantas Galvão ORCID: https://orcid.org/0000-0002-8773-9224 Secretaria Estadual de Saúde Laboratório Central de Saúde Pública do Tocantins, Brazil E-mail: jucydg@uft.edu.br Felipe Campos ORCID: https://orcid.org/0000-0002-2724-1045 Universidade Federal da Bahia, Brazil E-mail: fcampos1205@gmail.com Fábio Russomano ORCID: https://orcid.org/0000-0001-7510-0485 Instituto Nacional de Saúde da Mulher, da Criança e do Adolescente Fernandes Figueira/FIOCRUZ, Brazil E-mail: fabio.russomano@iff.fiocruz.br


Introduction
Cancer of the uterine cervix (CCU) is an important public health problem in Brazil, with relevant regional differences. The country's North and Northeast regions show the worst performance, exceeding by two to three-fold the mean national incidence of 15.85/100,000 women (INCA, 2017). In Tocantins State, located in the North of Brazil, cervical cancer is the most frequent neoplasia in females, with an estimated incidence of 33/100,000 women, greater severity of diagnosed cases and high mortality (Galvão, 2017;INCA, 2017).
The current CCU screening strategy in Brazil is cytopathological examination every three years after two negative annual tests, for women between 25 and 64 years of age (INCA, 2016). According to the Brazilian Guidelines for Cervical Cancer Screening, women with higher likelihood of presenting precursor lesions or initial cancer (high-grade squamous intraepithelial lesions or those involving the glandular epithelium -HSIL+) should be referred immediately for colposcopy, while women with lower likelihood of presenting such lesions (atypical squamous cells of undetermined significance, possibly non-neoplastic -ASC-US, or low-grade squamous intraepithelial lesions -LSIL) should be followed at the same service, repeating the cytopathological examination after at least six months (INCA, 2016).
The control of follow-up cervical cytology is affected by organizational deficiencies and difficulties in locating women via active search, resulting in an unknown loss to follow-up rate with a possible missed opportunity for the diagnosis and treatment of precursor lesions or initial cancer, undiagnosed in the initial examination (Galvão, 2017;INCA, 2016). A possibility for reducing loss to follow-up of these women is the identification of those at increased risk of presenting more important lesions, based on screening tests such as dual stain testing of proteins involved in viral oncogenesis or presence of oncogenic human papilloma virus (HPV) DNA (Cuschieri et al., 2018;de Melo et al., 2016).
These tests assume that the presence of oncogenic viral strains or proteins resulting from viral carcinogenesis increases the likelihood of the presence of a precursor lesion or cancer (Cuschieri et al., 2018;Donà et al., 2012), thus supporting immediate colposcopy for these women. However, evidence in the literature on the usefulness of dual stain testing for p16 and Ki67 and/or HPV DNA and, to a lesser extent, its applicability to screening or triage of women with altered cytology (Cuschieri et al., 2018;Donà et al., 2012), such tests are still not available in Brazil's Unified National Health System (SUS) and require the incorporation of technology, thus meriting studies applied to the Brazilian reality.
The database containing the records on the tests performed (Information System on Uterine Cervical Cancer, or SISCOLO in the Portuguese acronym) does not allow identifying the women that underwent the tests, since each record corresponds to one test, and there is no uniform and consistent record of the women's identifiers (Dias et al., 2010;Galvão, 2017;Reis Girianelli et al., 2009). Recently, new technologies have been applied that allow revisiting databases for different purposes from those for which the base was originally created, thus permitting new studies (Galvão, 2017;Tomazelli et al., 2018). Among these possibilities, probabilistic database linkage allows individualizing and correlating the information from one or more databases (or periods), based on information as patient's name, mother's name, and patient's date of birth. Amongst other information, this data analysis strategy allows searching for each woman's test records over time (Galvão, 2017;Silva et al., 2013;Tomazelli et al., 2018). The current study aimed to identify the magnitude of loss to followup of women with ASC-US or LSIL cytology and possible factors associated with women's non-adherence to follow-up in Tocantins State, Brazil, using data available in the SISCOLO database and probabilistic record linkage.
The identification of a large volume of losses to follow-up may suggest the need for measures targeted to subgroups at increased risk of non-adherence or with higher likelihood of presenting preinvasive disease or cancer for referral to specialized services and colposcopy immediately after the initial test with diagnosis of ASC-US or LSIL.

Methodology
This was a descriptive retrospective study based on data from the SISCOLO, provided by the administration of the Tocantins State Health Department (SES-TO) and referring to the period from January 2008 to December 2013. The data were obtained from the forms for "Cytopathological Examination -Uterine Cervix", with the respective results. In addition to the variables for the test and the women's identification, the database contains information on certain sociodemographic variables, but with uneven completion rates and thus variable reliability.
Identification of the women used probabilistic record linkage (Machado &amp; Hill, 2004) with a new variable created from the variables patient's name, mother's name, and patient's date of birth (Bastos, 2011;Galvão, 2017). Name variables were used in their phoneme format. After identification of the women and their tests, all those with a diagnosis of ASC-US or LSIL in 2011 were considered eligible. We excluded women with any cervical epithelial atypia in the three previous years (2008 to 2010), since they may have undergone a cytological examination in 2011 as a follow-up to a previous altered result.
For the women that were included, we conducted a search for a follow-up cervical cytology from 2011 to 2013, assuming that it would be a follow-up test, even if conducted after the recommended period (6 months), and within 24 months after the initial cytology with an altered result. Women that had not undergone a repeat cytology in this period were defined as losses to follow-up (or failures to adhere to the national screening guidelines), and we proceeded to analyze the variables related to such non-adherence.
Women's age was calculated based on date of birth at the time of the initial cytology in 2011 and categorized in the following ranges: 24 years or younger, 25 to 64 years, and 65 years or older. In the analysis of factors potentially associated with loss to follow-up, we compared the age range of 29 years or younger versus 30 years or older.
Information on previous cervical cytology was obtained from the variable "previous Pap smear". The first stage involved a descriptive analysis of the independent and outcome variables by initial diagnosis: ASC-US or LSIL. An exploratory bivariate analysis was performed, estimating relative risk for non-adherence to follow-up with 95% confidence interval. Although we originally planned a multivariate analysis, we did not, since there was no association with more than one dependent variable in the bivariate analysis. All analyses used the SAS software, version 9.4, with significance level set at 5% (p≤0.05).
The cervical cytology results were grouped as negative (no cytomorphological alteration), ASC-US, LSIL, and HSIL+ (for cases of high-grade squamous intraepithelial lesion -HSIL, alterations of undetermined significance, cannot rule out high-grade lesion -ASC-H, glandular cytomorphological alterations, and carcinoma).      Table 2 shows the proportion of women that returned for follow-up cytology within 24 months after the initial examination. 55.1% of the women failed to return for a follow-up cytology within this time frame. Of those who returned, most (70.9%) did so within 12 months, with no difference according to initial diagnosis (ASC-US versus LSIL). Of the 527 women who returned for follow-up cytology (44.9% of the total), 479 (90.9%) presented a negative result, 34 maintained the initial diagnosis of ASC-US or LSIL, and 14 (2.65%) were diagnosed with HSIL+ (Table 3). The attempt to identify factors possibly related to loss to follow-up only considered the variables present in the database with completion rates greater than 50%. Loss to follow-up was significantly associated with age lesser than 30 years (95%CI 1.10 to 1.35) and lack of (or inability to report) a previous Pap smear test (Table 4).

Discussion
This study focused on the identification of women who failed to return to health services for follow-up cervical cytology within 24 months after an initial diagnosis ASC-US or LSIL in 2011, when they should have done so within 6 months after the first test, as specified in the Brazilian Guidelines (INCA, 2016), in order to quantify losses to follow-up. Failure to undergo follow-up cytology was more frequent in younger women. No significant differences were observed in failure to return as a function of the initial test result (ASC-US versus LSIL).
Although a few women with altered initial cytology in the public Unified National Health System (SUS) may have done their follow-up cytology in the private health system, we consider this proportion negligible for the purposes of this study.
Identification of the 1.174 women was possible with probabilistic linkage of test records from the SISCOLO database. This method was necessary, since the database contains test records without a key that would allow deterministic linkage of the records to the same women over time (Galvão, 2017;Machado &amp;Hill, 2004;Reis Girianelli et al., 2009; C. M. Ribeiro et al., 2018;Silva et al., 2013;Tomazelli et al., 2018) . This difficulty has been addressed in other studies in Brazil using the same methodology (Bastos, 2011;Dias et al., 2010;Galvão, 2017;Silva et al., 2013)  Our study's results show that the cervical cancer screening strategy is undermined in Tocantins State by nonadherence to the recommendation for a follow-up cervical cytology after a diagnosis of ASC-US or LSIL. A previous study in the same state of Brazil, aimed to assess cervical cancer screening, found that failure to undergo a follow-up examination when recommended can also be explained by local cultural and organizational factors (Galvão, 2017). This may involve women's unknowledge of the Pap smear test's importance, leading failure to return to receive the test result when the clinical complaint has been resolved and limitations imposed by individual values and difficulties in access to health units or inefficient active search of cases for performing the cytopathological examination (Amaral et al., 2014;Galvão, 2017;INCA, 2016;Silva et al., 2013).
Operational factors can also contribute to women's loss to follow-up, including difficulties in engaging health teams in the follow-up of cases that require repeat testing, as well as monitoring of the program's activities by local health administrators and laboratory services (Amaral et al., 2014;Galvão, 2017;INCA, 2016;L. Ribeiro et al., 2016).
The opportunistic model, based on women's spontaneous search for testing, poses major difficulties for controlling the disease (Galvão, 2017;INCA, 2016;L. Ribeiro et al., 2016). The data highlight the need to improve the health services' practices and management in order to guarantee follow-up of women with original diagnosis of ASC-US or LSIL, as recommended (INCA, 2016).
According to the data, among the women that underwent follow-up cytology, 1.2% had a new test result suggesting greater severity (HSIL+) (Cytryn et al., 2009). This rate is relevant, considering the importance of diagnosis and the possibility of treating preinvasive lesions and initial cancer. Studies in Brazil, Iraq, and South Korea reported the frequency of alterations of cervical intraepithelial neoplasia grades II or III (CIN II/III) ranging from 0.8% to 1.85% in cases with ASC-US in previous results (Cytryn et al., 2009;Pity et al., 2012;Song et al., 2012). In women with previous cytology of LSIL, 21.3% obtained CIN II/III as outcomes. In addition, cases with repeated results of ASC-US or LSIL may indicate higher likelihood of presenting preinvasive lesions (INCA, 2015(INCA, , 2016. Other studies have identified loss to follow-up of women who presented initial ASC-US. Galão et al. (Galão et al., 2012) observed a loss of 40.3% of all women with ASC-US in Rio Grande do Sul State, Brazil. The authors found that 3.1% of women who were retested in the subsequent three years presented HSIL+ as their new result, corroborating the need for follow-up of these cases (Galão et al., 2012).
A study conducted in a reference laboratory service in São Paulo, Brazil, reported failure to undergo follow-up cytology in 27.2% of women with initial ASC-US (Rosendo et al., 2019). Among those who underwent repeat cytology, 4.4% presented HSIL+, and 16.6% of the cases identified with biopsy (independently of having undergone repeat cytology) presented more serious lesions (Rosendo et al., 2019).
Notwithstanding differences in the frequency of more serious lesions between women with diagnosis of ASC-US versus LSIL, these studies show the importance of follow-up cytology, highlighting that the problem is not exclusive to Tocantins, but that it is also experienced in more developed regions of Brazil with lower cervical cancer incidence and mortality rates (Galão et al., 2012;Ginsburg et al., 2017;INCA, 2017;Rosendo et al., 2019).
The literature reports important local and regional cultural differences, with some regions presenting greater social vulnerability, lower levels of schooling, and worse preventive habits that can help explain the higher rates of loss to follow-up after an initial cytology with ASC-US or LSIL (Galão et al., 2012;Galvão, 2017;INCA, 2016;Rosendo et al., 2019) . These characteristics may explain the fact that we observed the highest losses to follow-up in the north of Tocantins (Médio Norte Araguaia) and the central region (Capim Dourado). Although these two regions include municipalities (counties) with outstanding farming and cattle-raising activity and better economic development, they also include towns with lower population density and lower human development indices (HDI) (Galvão, 2017).
The highest losses to follow-up were in younger women (under 29 years) and those who had not undergone a previous cytological examination or could not remember. The findings suggest that less informed women and those who seek health services with active gynecological complaints rather than for routine prevention of cervical cancer may be the ones with worst adherence to the recommendation to undergo follow-up tests. A possible strategy to reduce losses is to target activities to these women and to the health professionals involved in their care (Amaral et al., 2014;Galvão, 2017;INCA, 2016).
The study's limitations include the use of secondary data from the SISCOLO database, which showed inconsistencies and/or incompleteness in some fields in the test orders. Lack of completion can be attributed to gaps in the work processes, since the test orders and Pap smear samples were sent to the outsourced laboratory service providers, who transcribed the accompanying data on the test results, merely for billing purposes. The large number of variables with less than 50% completion compromised the analysis of factors associated with loss to follow-up and thus prevented suggesting strategies targeted more specifically to women at increased risk of failure to undergo repeat cytology.

Conclusion
More than half of the women with a cervical cytology of ASC-US or LSIL in Tocantins State, Brazil, in 2011 failed to undergo a repeat test, as recommended by the Brazilian Guidelines. Loss to follow-up is a missed opportunity for identifying and treating women with precursor lesions or CCU, thereby jeopardizing the effective control of this disease.
Measures are needed to reduce loss to follow-up, especially among younger and less-informed women and those living in regions with lower human development or located farther from referral centers.