National and Subnational Trends of Incidence and Mortality of Female Genital Cancers in Iran; 1990–2016

1Non-Communicable Diseases Research Center, Endocrinology and Metabolism Population Sciences Institute, Tehran University of Medical Sciences, Tehran, Iran 2Department of Epidemiology and Biostatistics, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran 3Department of Biostatistics, Faculty of Medical Sciences, Tarbiat Modares University, Tehran, Iran 4Endocrinology and Metabolism Research Center, Endocrinology and Metabolism Clinical Sciences Institute, Tehran University of Medical Sciences, Tehran, Iran 5Reproductive Endocrinology Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran


Introduction
Gynecological cancers, including neoplasms of the cervix, corpus uteri and ovaries, were responsible for 499,727 deaths among females globally in 2016 and accounted for 3.46% of deaths in women aged 15 to 59 years. 1 The mortality rates of gynecologic cancers show variations across the globe; in 2008, it was noted that about 75% of deaths due to this group of cancers had occurred in developing countries. 2 In 2012, cervical cancer was ranked the fourth most common cancer in females and the most common gynecological cancer around the globe with an estimated 527 624 new cases. 1 Cervical cancer incidence displays a socio-economic gradient, with a distinct difference between its incidence rates in Africa and North America. Cervical cancer bears a crucial importance in public health as it mostly affects young women of low socioeconomic status, demonstrating the extent of health inequities. 3 Healthcare inequities are even present in developed countries, as variations exist in the screening and vaccination coverage of cervical cancer among different regional, ethnic and racial groups. 4,5 Although some studies have focused on time trends and regional disparities of gynecological cancers based on the national cancer registry, 6-10 there has not been a comprehensive analysis of trends of gynecological cancers in Iran at national and subnational levels. Furthermore, population-based studies on the epidemiology of gynecological cancers in Iran are out of date and incomplete, which further encouraged us to conduct this study.
As part of a larger project, in the present study, we aim to assess national and subnational incidence and mortality rates of gynecological cancers and their time trends from 1990 to 2016.

Material and Methods
Annual estimates of incidence for all cancers from 1990 to 2016 and estimates of mortality for all cancers from 1990 to 2015 at national and subnational level have been produced as part of the National and Subnational Burden of Diseases, Injuries, and Risk Factors (NASBOD) study. 11 In the NASBOD study, data regarding incidence cases of different cancers for years 2000 to 2010 were primarily obtained from the Iran Cancer Registry (except for the year 2006, which was not included in the database). Therefore, incidence data for years before 2000 and after 2010 needed to be extrapolated. The Social Security Organization Cancer Registry data and the Household Income and Expenditure Survey were used to assess rates of incompleteness in the main dataset and to provide necessary covariates for imputing missing data points and to complete Iran cancer registry data in the mentioned periods, after carefully omitting duplicated cases. For mortality data, we used different registries to obtain the most competent data; the sources are further discussed in references. Furthermore, mortality data from the death registry system was not available for the years 1990 to 1994 and 2005; therefore, mortality data for these years was extrapolated. 12 For imputation of missing data, the Amelia package of R statistical software and the multinomial imputation method were applied. In order to extrapolate data for the years before 2000 and after 2010, the Random Intercept Mixed Effects model and Age-Spatio-Temporal model were developed. Methodological details of these procedures have been further discussed in previously published articles. [11][12][13] The observed to extrapolated data ratio in this study is 10.5%.
Data on cause-specific mortality rates have been previously collected and processed under the NASBOD study and details of the applied modeling methods and data processing have been fully explained in earlier studies. 12,13 Cancers in this dataset are categorized based on the International Classification of Diseases version 10 (ICD-10). We extracted incidence data on all primary cancers that occurred at these sites: cervix (C.6.1), uterus (corpus uteri [C. 6.2] and uterine [C.6.6]), ovary (C.6.4), vulvovaginal region (C.6.7 and C.6.8) and unspecified sites and others (C.6.3 merged with placenta [C.6.5]).
The extracted data were further analyzed to calculate annual crude incidence and mortality rates, agestandardized incidence rates (ASIR) and age-standardized mortality rates (ASMR) as well as mortality to incidence ratio (MIR) for all gynecological cancers; the 95% uncertainty interval was also calculated for all the estimates.
To better understand the changes in incidence rates and mortality rates, the average annual percentage change (AAPC) throughout the study period was reported. Furthermore, decomposition analysis was performed to estimate the effect of population growth and population aging on changes in incidence rates for corpus uteri, ovarian and cervical cancers at national and subnational levels. To achieve this goal, we calculated two scenarios: In the first scenario, population age structure and age specific incidence rates for each of the mentioned cancers among women in 1990 were applied to the population size of 2016. In the second scenario, age specific incidence rates for each of these three cancers in 1990 were applied to the 2016 age structure and population size. Several different decomposition methods have been presented in the literature; [14][15][16][17] the overall idea behind the method is presented in the following formula: Overall-change = Population growth + Incidence rate changes + Age structure changes STATA/MP software version 14.0 was used for all statistical analyses, and graphs and diagrams were generated by R statistical software, v.3.1.3.

Results
In 1990, gynecological cancers accounted for 3.7% of new cases of cancer among women; with an ASIR of 4.5 (95% UI: 1.9-8.1) cases per 100 000 women. In 2016, gynecological cancers contributed to 8.0% of all new cases of cancer among women. An increasing trend was observed from 1990 to 2016, leading to an ASIR of 12.3 (9.3-15.7) cases per 100 000 women in 2016 ( Figures  1 and 2). The national AAPC during this period was 4%, while it ranged from 3% to 6% across the country ( Table 1). Proportion of incidence by age groups and types were illustrated in Figure 3. Except for two provinces that saw increasing trends from 1990 to 2009 and stable rates thereafter, the highest provincial ASIR was approximately 4.5-times bigger than the lowest provincial ASIR in 1990, while in 2016, the highest provincial ASIR was 5.7-times bigger than the lowest provincial ASIR.   In 1990, the highest provincial ASIR was approximately 10 times greater than the lowest provincial ASIR; this rate dropped to 4.5 in 2016.
Corpus Uteri Cancer Both the number of new annual cases and incident rates of corpus uteri cancer displayed a consistent increasing pattern at the national level. The number of new cases increased from 121 (42-244) in 1990 to 2026 (1399-2737) in 2016 ( Figure 2). Furthermore, ASIR indicated a similar time trend with a 5.8-fold increase from 1.7 (0.6-3.0) to 9.9 (6.8-13.4) during this period ( Figure 1).
Nevertheless, analysis of data on a provincial level revealed a gradual mild to moderate inclination of incidence rates in all provinces, with an AAPC range of 5-10% among provinces, while the national AAPC during the study was 7% (Table 1)
The ASMR of ovarian cancer slightly rose between 1990 and 1995, rising from 2.2 (1.9-2.5) to 3.0 (2.7-3.3) deaths per 100 000 women and then showed a consistent declining trend, reaching 1.0 (0.9-1.2) deaths per 100 000 women in 2015. The ASMR of corpus uteri cancer showed a declining trend during the study, dropping from 6.5 Decomposition Analysis Decomposition analysis showed that changes in age specific incidence rates contributed to 88% and 81% of the increase in incidence rates of corpus uteri and ovarian cancers, respectively. Furthermore, population growth and aging contributed to 3.5% and 8.3% of the rise in incidence rates of corpus uteri cancer and 7.3% and 12% of the rise in incidence rates of ovarian cancer, respectively (Tables 2 and 3).
In the case of cervical cancer, the alterations in incidence rate were mainly due to population growth, which contributed to 61.4% of the overall change. Moreover, changes in age structure attributed to 38.0% of the overall change, while changes in age specific incidence rates were responsible for less than 1% of the total change in incidence rate (Table 4).

Discussion
In the present study, we provided an overview of national and subnational incidence and mortality rates as well as trends of different categories of gynecological cancers from 1990 to 2016, demonstrating rising trends in the incidence rates of corpus uteri, ovarian and vulvovaginal cancers throughout the study. The results of this study show that national MIR, an indicator of access to and quality of care, have declined for all gynecological cancers between 2000 and 2015.
During this study, cervical cancer incidence showed an increasing trend from 1990 to 2004 , and then following a 5-year period of steady rates (between 2004 and 2009), the incidence rates declined until 2016, which could be associated with the onset of the national screening program using Pap smears in the 1990s. However, the starting point of this declining trend varied across the country, possibly due to unequal healthcare coverage among the provinces, as well as disparities among the provinces in compliance to regular screening and necessary treatments. The trend observed in this study was in accordance with the data provided by the Institute for Health Metrics and Evaluation (IHME). However, in the IHME report, the incidence of cervical cancer had a steady trend between 1990 and 2012 and a declining trend in incidence was not  observed until 2013. 18,19 In the present study, more data points were used than IHME.
Moreover, we observed an increasing incidence of ovarian cancer from 1990 to 2008 and steady rates afterwards. However, incidence rates had a linear increasing trend until 2016 in the IHME report, reaching a similar incidence rate in 2016 -as presented in this study. 18 The uterine cancer trend reported by the IHME is similar to ours; however, the incidence rates were underestimated in the IHME report. 19,20 The rising trends of uterine and ovarian cancer in Iran can be explained by lifestyle changes in the past three decades, for instance, the rise of obesity prevalence and increased age of first childbirth. [20][21][22] The main strength of this study was benefitting from different national datasets for cancer incidences and mortalities and establishing a statistical model to resolve issues such as incompleteness of data. Although the observed to extrapolated data ratio in this study is not high, it is acceptable compared to previous ratios in international studies; for instance 5.6% in a cholesterol study, 23 16.6% in a study on BMI 24 and 13.6% in a study on systolic blood pressure. 25 As both data of incidence and mortality before the year 2000 were not real data and were extrapolated, the MIR, which is a computational measure, was not reliable before 2000. Therefore, MIR values were only reported after 2000 to be more precise and dependable. Although statistical models have been applied to omit the effects of missing data, junk codes and misalignments, the necessity for a higher quality national cancer registry with better coverage should not be overlooked. Furthermore, lack of reports on the coverage of the cervical cancer screening program at subnational and national levels may leave us with some doubts concerning the influence of the cervical cancer screening program on the incidence and mortality rates. Another limitation that we faced was the sparsity of covariates that were included in the cancer registry, leading to restrictions in investigating the risk factors influencing the incidence of specific cancers.
The introduction of the Papanicolaou (Pap) test as a screening tool for cervical cancer (which is currently acknowledged as the most effective tool for controlling this cancer) has affected the incidence of this neoplasia differently across countries and regions. 5 Cervical cancer screening programs have been initiated in several countries starting from the 1970s and have affected mortality and  morbidity rates of cervical cancer globally; yet in many middle-income developing countries, ongoing programs are not succeeding. 26 The first cervical cancer screening program guideline in Iran was established in 1990. It was an opportunistic screening program that encouraged women aged 20-65 years to take a free Pap smear test annually for cytology when visiting public health clinics for routine health checkups or for maternal-child health and family planning services. After conducting a cost-effective study on this guideline, the policy toward cervical cancer screening changed and the Pap smear screening was eliminated from the public health services' coverage in 2000. 27 In the first years following the initiation of the national screening of cervical cancer, an increase in incidence was detected, probably due to the increase in detection of unknown cases of cervical cancer. Following this period, we can see the impact of the screening program leading to a steady trend and then a mild decline afterwards, probably due to the detection of pre-cancerous lesions. Although the national screening program was ceased in 2000, some insurance companies still partially covered the expenses of Pap smear cytology assessment and women could receive screening tests upon their own request or in case their health-care provider recommended it. 28 Recent studies suggest that a shift from Pap smear to HPV DNA testing for cervical cancer screening would be beneficial in Iran, considering its convenience and effectiveness as well as its cost-effectiveness. 29,30 According to a cost-effectiveness analysis study conducted in Iran, starting at the age of 35 years, and at 5-years intervals, HPV DNA testing would be the most cost-effective screening technique based on national incidence rates and incomes. 31 It is also important Trend of Female Genital Cancers in Iran: 1990-2016 to notice the declining pattern of MIR for gynecological cancers, which could be explained by an increase in their early detection and better survival due to better access to treatment and better quality of care. Since cervical cancer is an HPV-related cancer, population coverage with HPV vaccine can also result in reduction of its incidence and mortality, as observed in highincome societies with high-prevalence of cervical cancer, for instance, New Zealand, Canada and Denmark. 4,[32][33][34] However, studies in Iran have shown that due to the high cost of HPV vaccination and the low incidence of cervical cancer, a national obligatory vaccination program would not be cost-effective: in countries like Iran, an organized screening program would be more effective. [35][36][37] In conclusion, this study presents the gynecological cancer profile in general as well as each specific cancer in Iran from 1990 to 2016, depicting the rising trends in incidence of gynecological cancers, specifically ovarian and uterine cancers. Furthermore, it assessed disparities among provinces by observing the trends of incidences and mortalities over time, which showed a converging pattern, indicating advancements toward health equality.
The results provided by the current study can be beneficial for policy-makers in evaluating earlier screening program guidelines as well as other executive plans to decrease the effects of known risk factors for gynecological cancers.
Authors' Contribution SE collaborated in data processing; collaborated in analysis; wrote the manuscript. NE performed statistical analysis. FF, SSM and AS initiated, conceptualized and designed the study; edited and critically reviewed manuscript. MS, FM, BM, KG, NA, and SMF collaborated in collection of data; critically reviewed manuscript. FRT edited and critically reviewed manuscript. All authors read and approved the final manuscript.

Conflict of Interest Disclosures
The authors have no conflict of interests to disclose.

Ethical Statement
The study protocol has been approved by the Ethics Committee of the National Institute for Medical Research Development (IR.NIMAD. REC.1396.192).

Funding
This study was funded by the National Institute for Medical Research Development (Grant number: 963348).