Sociodemographic predictors of endometrial cancer mortality in South Africa (1997 to 2015): a case-control study

Abstract

South Africa is currently undergoing epidemiological and health transition that may impact on the risk factors of endometrial cancer mortality. We evaluated the sociodemographic characteristics of women that died from endometrial cancer in South Africa from 1997 to 2015.

An unmatched case control study was conducted to compare the sociodemographic characteristic of women that died from endometrial cancer cases (n = 3,955) with the characteristics of women that died from other cancers (controls, n = 66,202) using the population-based mortality data from Statistics South Africa. Unconditional binary logistic regression modelling was conducted.

The Mean age of women that died from endometrial cancer was 66.7 ± 11.9 years. The odds of death from endometrial cancer among women aged ≥50 years was about 4-fold as compared to women younger than 50 years (AOR = 3.98 95% CI: 3.14 − 5.03; p value: <.001). Conversely, high school leavers (AOR = 0.79 95% CI: 0.66 − 0.94; p value: .009), smokers (AOR= 0.53 95% CI: 0.43 − 0.65; p value: <.001), and divorced women (AOR= 0.73 95% CI: 0.59 − 0.89; p value: .002) had lesser odds of endometrial cancer mortality.

This study provides information to guide Public Health control program on endometrial cancer in South Africa.

Impact statement

• What is already known on this subject? The prevalence of endometrial cancer is increasing in low and middle income countries. After the abolition of Apartheid in South Africa, the country commenced a multi-racial government that provided socio-economic empowerment to the majority of the country. Thus, South Africa is currently undergoing epidemiological and health transition that may impact on the risks of endometrial cancer in the country.

• What do the result of the study add? This study is the first in South Africa to evaluate the sociodemographic characteristics of women that died from endometrial cancer using a large population-based data. Age, educational status, smoking status are some identified risk factors of mortality from endometrial cancer in South Africa

• What are the implications of these findings for clinical practice and/or further research? The highlighted risk factors that are associated with endometrial mortality will help in guiding health policies that will help reduce the prevalence of endometrial cancer. The study recommends a targeted national prevention program for endometrial cancer as the burden appears to be worsening with each passing year.

Keywords:

• Endometrial cancer
• reproductive cancer mortality
• risk factors
• sociodemographic South Africa

Introduction

Globally, more than 400,000 cases and 90,000 deaths occurred from endometrial cancer in 2020 (International Agency for Research on Cancer 2020). Furthermore, 39,867 cases and 13,419 (Global Cancer Observatory 2020) endometrial cancer deaths occurred in low and middle income countries (LMICs) (Bray et al. 2018; International Agency for Research on Cancer 2018b). Endometrial cancer mortality is increasing globally, and most LMICs have higher percent increases (Raglan et al. 2019; Zhang et al., 2019).

In South Africa, endometrial cancer is the second most common gynaecological cancer and the fourth most common cause of cancer deaths (International Agency for Research on Cancer 2018a). The pattern of gynaecological cancer incidence and deaths in South Africa contrasts with the female cancer patterns in high income countries where endometrial cancer is the most common cause of gynaecological cancer death (Braun et al. 2016; Howlader et al. 2016; Brooke Steele et al. 2017). Since South Africa is currently undergoing epidemiological and health transition with associated increased prevalence of obesity, coupled with a low national fertility rate, there may be an increase in the risk of endometrial cancer in the country (Gerland et al. 2017).

The pathophysiology of endometrial cancer is multi-factorial. However, the key aetiology is chronic excessive unopposed action of oestrogen on the uterine endometrium. Common risk factors of endometrial cancer include obesity, early menarche, late menopause, advanced age, and some genetic predisposition such as Lynch syndrome and Cowden syndrome (Kumar et al. 2014). While most studies have focussed on the risk factor for the incidence, few have evaluated the risk factor for the mortality of endometrial cancer (Brüggmann et al. 2020). Although, South Africa currently has national cancer control programs for breast (National Department of Health 2017a) and cervical cancer (National Department of Health, 2017 b), there is no national cancer control guideline for endometrial cancer. This study therefore evaluates the socio-demographic risk factors of mortality from endometrial cancer as compared to death from other cancers among South African women from 1997 to 2015. The result from this study will contribute to evidence for designing an endometrial cancer control program in the country.

Materials and methods

An unmatched case control study was conducted using the nationally representative mortality data of all female cancer deaths in South Africa between January 1997 and December 2015 as reported by the Statistics South Africa (Stats SA). The Stats SA is a national agency that collates all notifications of death records as obtained from the Department of Home Affairs. Reporting of deaths is mandatory in South Africa (Department: Home Affairs 2020).

The women whose underlying cause of death was endometrial cancer (ICD-10 C54) were taken as the cases (n = 3,955) and coded as ‘1’. All other women whose underlying cause of death were cancers (ICD: C00–D49) other than endometrial cancer were classified as the control and coded as ‘0’. Women whose underlying cause of deaths were due to cancers with known association to oestrogen metabolism (breast, biliary, pancreatic and gynaecological cancers) (Zane et al. 2017; Parl et al. 2018; American Cancer Society 2019) were excluded from the control arm of the study. Furthermore, women whose underlying cause of deaths was from any cancer that has an association with obesity, Human Immunodeficiency Virus, smoking and Lynch syndrome were also excluded (Johnatty et al. 2017; Oliver and Chiao 2017) (See list of excluded cancers in Supplementary information Table 3). These exclusions were based on their association with endometrial cancer either as protective factor or risk factor. The excluded cancer deaths have the potential to cause bias of the odds ratio. This case control study has been utilised by other researchers to obviate referral and selection bias since both the cases and controls in this instance are drawn from the same dataset (Mallick et al. 2017).

Data analysis

Data was imported into Stata/SE version 15.1 (Stata Corp, USA). Data cleaning and validation was done. Categorical variables were summarised as frequency and percentage. Continuous variables were summarised as mean and standard deviations. Pearson’s Chi-square and Student’s independent t-test were used to assess the association between the dependent variable (Endometrial cancer deaths/non-Endometrial Cancer death) and the independent variables (age, marital status, smoking status, educational status, ‘year period of death’ and province of usual residence). Unconditional univariable and multivariable binary logistics regression was then conducted to obtain crude and adjusted odds ratio (with 95% confidence interval). Variables with univariable p-value <.2 were included in the multivariable model. Backward elimination method was used to build the multivariable model. Post regression variance inflation factor (VIF) to check for collinearity was conducted. Statistically significant variable was set at p-value <.05.

Ethical approval

The study was approved by the Research and Ethics Committee of the faculty of Health Sciences, University of Johannesburg before commencement (NHREC number 241112-035). Permission was obtained from the management of Statistics South Africa to utilise the anonymized dataset.

Results

During the study period 1997–2015, 322,979 women died from cancer in South Africa. The endometrial cancer deaths (cases) were 3,955. Thus, 319,024 women with non-endometrial cancer deaths were in the control arm. After excluding 252,822 from the non-endometrial cancer deaths, the control now became 66,202. The comparison of the sociodemographic characteristics of women that died from endometrial cancer as compared to women that died from other cancers are as shown in Table 1.

Table 1. Comparison of the sociodemographic characteristics of endometrial cancer and non-endometrial cancer female cancer deaths in South Africa (1997–2015).

Characteristics	Endometrial cancer deaths N, (%)	Non-endometrial cancers deaths N, (%)	Total N, (%)	^ap-Value
Total number	3,955 (100)	66,202 (100)	70,157 (100)
Age (mean ± SD)	(67.8 ± 11.9)	(59.7 ± 18.7)	(60.1 ± 18.5)	<.001^b
	318 (8.1)	17,345 (26.2)	17,663 (25.2)	<.001
≥50	3,633 (91.9)	48,760 (73.8)	52,393 (74.8)
Marital status
Never married	903 (28.3)	18,720 (33.9)	19,622 (33.6)	<.001
Married	1,149 (36.0)	18,463 (33.4)	19,611 (33.5)
Widowed	472 (14.8)	5,922 (10.7)	6,394 (11.0)
Divorced	671 (21.0)	12,164 (22.0)	12,835 (22.0)
Smoking status
Smoker	167 (10.2)	4,271 (16.7)	4,438 (16.3)	<.001
Non-smoker	1,479 (89.9)	21,359 (83.3)	22,837 (83.7)
Highest education
No education	309 (21.3)	4,321 (18.3)	4,630 (18.5)	<.001
Primary school	410 (28.2)	4,753 (20.1)	5,163 (20.6)
High school	645 (44.4)	12,911 (54.7)	13,556 (54.1)
Tertiary	88 (6.1)	1,624 (6.9)	1,712 (6.8)
Province of usual residence
Western Cape	594 (16.8)	10,275 (17.9)	10,869 (17.8)	<.001
Eastern Cape	382 (10.8)	7,307 (12.7)	7,688 (12.6)
Northern Cape	138 (3.9)	1,931 (3.4)	2,069 (3.4)
Free State	300 (8.5)	3,848 (6.7)	4,148 (6.8)
KwaZulu-Natal	566 (16.0)	9,653 (16.8)	10,219 (16.8)
North West	179 (5.1)	3,180 (5.5)	3,359 (5.5)
Gauteng	955 (27.1)	14,671 (25.6)	15,626 (25.6)
Mpumalanga	170 (4.8)	3,077 (5.4)	3,247 (5.3;)
Limpopo	245 (6.9)	3,488 (6.1)	3,733 (6.1)
Year period of death
1997–2001	658 (16.7)	14,963 (22.6)	15,621 (22.3)	<.001
2002–2006	917 (23.2)	16,910 (25.5)	17,827 (25.4)
2007–2011	1,163 (29.4)	17,794 (26.9)	18,957 (27.0)
2012–2015	1,214 (30.7)	16,535 (25.0)	17,749 (25.3)

^aExcept otherwise stated, the p-value is for Pearson’s Chi-square. ^bp-value for student’s independent t-test. SD: standard deviation.

From Table 1, the mean age at death of women who died from endometrial cancer was higher than the mean age of women who died from other unrelated cancers (67.75 ± 11.91, and 59.65 ± 18.71 years). Nearly all the endometrial cancer deaths (91%) occurred among women older than 50 years while about three-quarters (73%) of non-endometrial cancer deaths occurred in women older than 50 years.

Both the cases and control had the highest percentage of death among those with high school level of education, (44.42 vs. 54.69%). On the other hand, the least percentage of EC deaths occurred among women with tertiary education (EC deaths was 6.06% and control was 6.88% respectively). Gauteng, Western Cape and Kwazulu-Natal were the provinces with the highest percentage of death in both the cases (27.07, 16.84, and 16.04%) and control (25.55, 17.89, 16.81%) groups. There was a progressive increase in the absolute number of deaths among both cases and control with increasing year period from 1997 to 2011. However, in the latest year period of 2012–2015, while control group recorded a decrease in absolute number of death from 17,794 to 16,535, endometrial cancer deaths increased further from 1,163 to 1,214.

Socio-demographic predictors of deaths from endometrial cancer

Having only primary school education raised the odds of EC deaths by 21% as compared to women with no formal education at univariable analysis (COR: 1.21; 95% CI: 1.03−1.41; p-value=.016). This association did not reach statistical significance after multivariable analysis (p-value=.077) (Table 2). Furthermore, women who had secondary education or tertiary education as the highest educational qualification had 30 and 24% (Table 2) lesser odds of endometrial cancer as compared to women with no formal educational. However, only the relationship with secondary education remained statistically significant after multivariable analysis. Similarly, five out of eight provinces had lesser odds of EC deaths as compared with Northern Cape Province which had the least number of EC deaths after univariate analysis. However, there was no statistically significant relationship between province of usual residence and EC deaths after multivariable analysis.

Table 2. Univariable and multivariable logistic regression of the socio-demographic characteristics of Endometrial cancer deaths.

	Univariable analysis			Multivariable analysis
Factor	COR	95% CI	p-Value	Adj OR	95% CI	p-Value
Age
	Ref	Ref	Ref	Ref	Ref	Ref
≥50	4.06	3.62–4.56	<.001	3.98	3.14–5.03	<.001
Marital status
Never married	Ref	Ref	Ref	Ref	Ref	Ref
Married	1.29	1.18–1.41	<.001	0.85	0.72–1.00	.049
Widowed	1.65	1.47–1.86	<.001	0.89	0.72–1.08	.233
Divorced	1.14	1.03–1.27	.010	0.73	0.59–0.89	.002
Smoking status
Non-Smoker	Ref	Ref	Ref	Ref	Ref	Ref
Smoker	0.53	0.48–0.67	<.001	0.53	0.43–0.65	<.001
Highest education
No education	Ref	Ref	Ref	Ref	Ref	Ref
Primary school	1.21	1.03–1.41	.016	1.18	0.98–1.43	.077
High school	0.70	0.61–0.80	<.001	0.79	0.66–0.94	.009
Tertiary	0.76	0.59–0.97	.026	0.82	0.61–1.11	.197
Province of usual Residence
Northern Cape	Ref	Ref	Ref	Ref	Ref	Ref
Eastern Cape	0.73	0.60–0.89	.002	0.99	0.69–1.41	.951
Free State	1.09	0.89–1.34	.414	1.32	0.91–1.91	.142
Gauteng	0.91	0.76–1.10	.322	1.05	0.74–1.47	.797
KwaZulu-Natal	0.82	0.68–0.99	.044	1.06	0.76–1.48	.740
Limpopo	0.98	0.79–1.22	.875	1.12	0.77–1.62	.560
Mpumalanga	0.77	0.61–0.97	.029	0.97	0.66–1.43	.884
North West	0.79	0.63–0.99	.041	1.01	0.66–1.54	.968
Western Cape	0.81	0.67–0.98	.030	1.00	0.69–1.46	.980
Period of diagnosis
1997–2001	Ref	Ref	Ref	Ref	Ref	Ref
2002–2006	1.23	1.11–1.37	<.001	1.22	0.97–1.54	.091
2007–2011	1.49	1.35–1.64	<.001	1.26	1.00–1.58	.046
2012–2015	1.67	1.51–1.84	<.001	1.46	1.17–1.83	.001

COR: crude odds ratio; AdjOR: adjusted odds ratio.

After adjusting for covariates (age group category, marital status, smoking status, educational status, province of usual residence and ‘year period of death), increasing age and year of study, the age category ≥50 years and later year periods were more associated with EC death (cases) than death from other cancers in South African women (Control). Conversely, factors such as smoking, being a divorcee, having high school as highest educational attainment were associated with reduced odds of EC death.

Women who were 50 years or older were about four times more likely to die from endometrial cancer than women younger than 50 years (Adjusted odds ratio [AOR]: 3.98; 95% CI: 3.14–5.03; p value: <.001) as shown on Table 2. The odds of deaths from EC increased as the year-periods increased. For instance, compared to the 1997–2001 period, the odds of dying from EC increased from about 1.3 (AOR: 1.26; 95% CI: 1.00−1.58; p-value=.046) during the time period 2007–2011 to 1.5 (AOR: 1.46; 95% CI: 1.17–1.83; p-value: .001), during the 2012–2015 time period. The odds of EC deaths decreased among divorced women by 27% (AOR: 0.73; 95% CI: 0.59–0.89; p-value= .002) as compared to ‘never married’ women. Additionally, smoking conferred some protection as smokers had about 47% reduced odds of death from EC as compared to non-smokers (AOR: 0.53; 95% CI: 0.43–0.65; p-value: <.001). Women who had high school education as their highest educational status had a 21% decreased odds of EC deaths as compared to those with no education (AOR: 0.79; 95% CI: 0.66–0.94; p-value=.009).

The variance inflation factor (VIF) was 1.03 which showed that there was no collinearity.

Discussion

The study aimed to evaluate the sociodemographic risk factors of death from endometrial cancer in South Africa. Although there are studies that investigated the sociodemographic characteristics of incident cases of endometrial cancer, (Ali 2014; Allen et al. 2015; Collaborative Group on Epidemiological Studies on Endometrial Cancer 2015) to our knowledge, our study is the first in South Africa and possibly sub-Saharan Africa to utilise population based data to evaluate the sociodemographic characteristics of deaths from endometrial cancer. This study revealed some independent risk factors of endometrial cancer deaths such as age older than 50 years, high school educational qualification, marital status, tobacco smoking and time period. Province of residence was not associated with the risk of endometrial cancer.

Age and endometrial cancer

Age of death in women with endometrial cancer was higher by close to a decade than women’s age at death in non-endometrial cancer related death. This pattern is in keeping with the findings from previous studies that endometrial cancer occurred mainly among post-menopausal women (Setiawan et al. 2013; Decherney et al. 2019). On the other hand, our result may imply older affected women had poorer prognosis of death as compared to younger women. A cohort study may assist to firmly conclude on this.

Educational attainment and endometrial cancer death

Our result suggests that women with high school educational qualification (>15 years of formal education) had lesser odds of having endometrial cancer as compared to women with only a primary school education (<7 years of formal education). A similar pattern was reported by Seidelin et al in Denmark who found that women with short education had higher hazard of death from endometrial cancer (Seidelin et al. 2016).

These findings may be explained by a direct relationship between educational status and health seeking behaviour among females (Peltzer and Phaswana-Mafuya 2012; van Hedel et al. 2018). However, our study showed that women who had tertiary level education tend to have slightly higher odds of endometrial cancer deaths as compared to women who had only high school education. This suggests that the association between educational qualification and risk of death from endometrial cancer may be complex and not entirely linear as other studies have reported more advanced stages of disease with lower degree of education (Svanvik et al. 2019; Njoku et al. 2021).

Women with tertiary education are more likely to delay child birth while schooling and have low parity. Furthermore, women with tertiary education are more likely to engage in jobs that promote sedentary life-style with higher socioeconomic status (Parry and Straker 2013). Such women may adopt westernised diets and are subsequently predisposed to obesity (Wu et al. 2015). The aforementioned factors are key drivers of endometrial cancer incidence (Raglan et al. 2019) and can increase the risk of the disease among women who had tertiary education.

Although women with tertiary education had a higher risk of endometrial cancer mortality than women with high school education. Women with tertiary education still had lesser risk than women with no formal education for endometrial cancer mortality. Early detection and prompt treatment of endometrial cancer is potentially curative and can increase its survival rate (Seidelin et al. 2016). Thus, good health seeking behaviour among women with tertiary education could have reduced the risk of mortality among them as compared to women without formal education. All women irrespective of educational status should therefore be targeted for screening for endometrial cancer based on other prevailing factors (Seidelin et al. 2016).

Marital status and endometrial cancer death

Divorced marital status was associated with a reduced risk of endometrial cancer death by about 27%. High parity is known to be a strong protector against endometrial cancer occurrence in women (Andarieh et al. 2016; Ray et al. 2019). Marital status has been shown to affect a woman’s fertility with increased spousal separation being associated with lesser number of children (Houle et al. 2016). However, in South Africa a divorce state may be associated with higher fecundity as there is usually prompt remarriage and delivery of more children in the new marriage (Hertrich 2017). This leads to higher parity in such women and may lead to some form of protection against endometrial cancer occurrence.

It has also been suggested that late hospital presentation is common among unmarried women (Akinyemi et al. 2019), hence it would have been expected that the endometrial cancer deaths would be higher among them due to poor health seeking behaviour but it was not so in this study.

Smoking status and endometrial cancer death

A reduced risk of endometrial cancer death was associated with smoking in this study. This is consistent with different studies which reported reduced risk of endometrial cancer deaths among women who are smokers (Felix et al. 2014; van den Berg and van Duursen 2019). The protective effect of tobacco smoking has been attributed to its hypo oestrogenic effect, thereby leading to prevention of excessive proliferative growth of endometrial lining that usually lead to endometrial cancer (van den Berg and van Duursen 2019).

This study further corroborates the fact that endometrial cancer is one of the few cancers whose risk does not increase with smoking. South African women currently have one of the highest female smoking rates in medium to high development index countries (American Cancer Society 2020). However, there is a declining trend in the smoking rates in South Africa due to the vigorous tobacco control program initiated by the government.

At the population and ecological level, our finding may suggest that the tobacco control program in South Africa can increase the prevalence of endometrial cancer rates among women who stopped smoking (Reddy et al. 2015). Thus, tobacco quitters should be monitored for early features of endometrial cancer, especially if they have other risk factors (Foundation for a smoke-free world 2019). Nonetheless, tobacco smoking has been associated with the evolution of multiple cancers, cardiovascular diseases and poor health outcomes and should not be encouraged (American Cancer Society 2020).

Province of usual residence and endometrial cancer deaths

This study shows that women who had lived in Eastern Cape, Mpumalanga, Northwest Cape, Western Cape and KwaZulu-Natal had 27, 23, 21, 19, and 0.9% reduction in odds of dying from EC compared to other female cancers on univariable analysis. Western Cape has the highest smoking prevalence in South Africa (Reddy et al. 2015). Thus the reduction is consistent with the protective effect of smoking on endometrial cancer occurrence in this province. Other provinces (Northern Cape and Free State) with similarly high smoking prevalence did not record such reduction in EC death odds. Besides, these reductions were not significant after multivariable analysis signifying no difference in the endometrial cancer deaths or other female cancers deaths.

Some authors have recommended that community-specific interventions are important for policy making based on the variations in socioeconomic characteristics of different provinces (Made et al. 2017; Tuoane-Nkhasi and van Eeden 2017). However the difference in socioeconomic characteristics has not been seen among the provinces between endometrial cancer deaths and other female cancer deaths in this study.

Year period of death and endometrial cancer death

Both endometrial cancer death and other female cancer death increased with each advancing year period with the later years recording more deaths than the earlier years. This is in keeping with the upward trend already anticipated for cancers in LMICs like South Africa (Lortet-Tieulent et al. 2018). It must however be noted that endometrial cancer death odds increased by 67% in the latest year period of 2012–2015 as compared to death during the earliest period of the study (1997−2015). Lortet-Tieulent et al suggested an increasing disease burden of endometrial cancer particularly in countries undergoing urbanisation and westernisation (Lortet-Tieulent et al. 2018). Although all cancers are said to be on the increase, endometrial cancer seem to mirror obesity prevalence rise (Kengne et al. 2017).

Strengths and limitations

The limitation of the study included the retrospective nature of the study and the non-availability of some other variables of interest such as body mass index since secondary data was utilised for these analyses. Also, information on the histological types of the endometrial cancer was not available. However, the impact of obesity on our findings will be blunted since we excluded cancers that are associated to obesity from the control arm of the study. The large sample size and robust statistical analysis is a strength of this study.

Conclusion

This study has identified some predictors of EC deaths that can assist in endometrial cancer control. While women older than 50 years had higher odds of endometrial cancer as compared to younger women, smokers are likely to have reduced risk of death from endometrial cancer. Likewise, higher educational attainment tends to be protective against death from endometrial cancer. The odds of death from endometrial cancer is also increasing per year. This highlights the need for a more comprehensive approach in national cancer prevention policies in South Africa, with inclusion of strategies targeted towards the reduction of endometrial cancer deaths.

Acknowledgement

Statistics South Africa is acknowledged for providing the cancer mortality dataset.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data is available with the corresponding author on reasonable request.

Additional information

Funding

The Department of Environmental Health, Faculty of Health Sciences, University of Johannesburg is acknowledged for providing funding to AIE to present abstract of this study at Public Health Association of South Africa conference. GO is funded by Glaxo Smithkline (GSK) Africa Non-Communicable Disease Open Lab through the DELTAS Africa Sub-Saharan African Consortium for Advanced Biostatistics (SSACAB) training programme. (Grant number: D1702270-01). The views expressed in this publication are those of the author(s) and not necessarily those of GSK.