Mortality of Midlife Women With Surgically Verified Endometriosis

A Cohort Study Including 2.5 Million Person-years of Observation

L. Saavalainen; A. But; A. Tiitinen; P.Härkki; M. Gissler; J. Haukka; O. Heikinheimo


Hum Reprod. 2019;34(8):1576-1586. 

In This Article


The all-cause mortality in midlife was lower throughout the follow-up among women with surgically verified endometriosis compared to the reference cohort.

The absolute difference was low—four fewer deaths occurred among 1000 women over 10 years of follow-up in endometriosis patients. Endometriosis is associated with an increased risk of several common diseases, also known as common causes of death. However, even if morbidity is increased, mortality due to these conditions may be decreased. Nevertheless, even after adjustments mortality due to these conditions was decreased, i.e. deaths due to any cancer and cardiovascular conditions including ischemic heart disease and cerebrovascular disease. We also found a decreased risk of death due to alcohol-related causes, accidents and violence, and diseases of the digestive and respiratory system.

The strengths of this study include the surgically diagnosed endometriosis disease, the large, population-based cohort of women, and the long follow-up (nearly three decades of calendar time and a follow-up of 2.5 million person-years). Finland has a long history of administrative data collection. Nationwide health and social registers have provided an important data source for epidemiological research. Moreover, due to the high-quality nationwide population-based registers, the completeness and validity of the data are reliable (Gissler and Haukka, 2004; Sund, 2012; Pukkala et al., 2018). The registers also allowed us to adjust for many demographic factors that are important when assessing mortality (Forouzanfar et al., 2016; Mackenbach et al., 2016; Jensen et al., 2017; Stringhini et al., 2017). In addition, previous knowledge of the all-cause and cause-specific mortality in women with endometriosis is scarce.

Several important lifestyle factors, such as smoking, alcohol consumption, BMI, or use of medications could not be adjusted for as they do not exist in our register-based data. These risk factors contribute significantly to the development and prognosis of several illnesses and, therefore, also to deaths; therefore, the residual confounding cannot be ruled out (Danaei et al., 2009; Di Angelantonio et al., 2016; Flegal et al., 2013; . In previous studies, endometriosis diagnosis has been associated with lower BMI, but the results have been inconsistent for alcohol consumption and tobacco smoking (Parazzini et al., 2013; Bravi et al., 2014; Shafrir et al., 2018). We found lower risk of death due to alcohol-related causes in women diagnosed with endometriosis. In addition, the decreased mortality due to accidents and violence might also reflect the safer lifestyle of women with endometriosis. However, there was no significant difference in the risk of lung cancer mortality, which often reflects the smoking habits of the study population.

The present study may also be subject to selection bias, bias by indication, detection bias, or reverse causality. First, the selection bias exists as the procedural data were not collected until 1987 and, therefore, some women in the reference cohort may have undergone the endometriosis procedures prior to that. Moreover, the reference cohort is likely to include women with undiagnosed endometriosis (~2%; Zondervan et al., 2002) and endometriosis without surgical verification.

Uneven access to health care results often in another selection bias. There are also some inequalities in access to health care in Finland (Kangas and Blomgren, 2014). Moreover, the access to the specialized medical care may depend on patient's awareness and persistence. This may cause a selection bias in our study. In the analysis, we adjusted for the education level that is known to be associated with the socioeconomic status, health behavior, and risk contexts. Endometriosis is typically diagnosed after a delay of ~7 years (Nnoaham et al., 2011). Thus, women in the endometriosis cohort are likely to have been rather persistent in seeking medical advice and help. This may apply to other health issues as well. In addition, receiving an endometriosis diagnosis and medical attention might alter the overall behavior toward healthier lifestyle. It is also possible that some of the important risk factors, such as alcohol abuse, restrain women from seeking medical help. Moreover, to be eligible for operative treatment is likely to exclude several serious conditions, and preoperative evaluation might reveal other pre-existing diseases as well as increased medical attention postoperatively leads to a situation referred as selection and detection bias.

The indication bias is caused by limiting the study cohort to women with endometriosis eligible for operative treatment, although the indications were not otherwise limited as we also included incidental diagnosis of endometriosis (subsidiary diagnosis, 35% of all diagnosis). In addition, confounding by indication might also be caused by the presence of comorbidities between the cohorts, not adjusted in the present study.

Moreover, live births were taken into account as nulliparous women are known to have increased risk of death (Zeng et al., 2016). However, the data on infertility was not available. Furthermore, the former data have shown women with assisted reproductive techniques to have decreased risk of death although the recent study have shown that there is a healthy patient effect—the risk of death returns to normal after 10 years. (Braat et al., 2010; Vassard et al., 2018).

Women with endometriosis are likely to use more non-steroidal anti-inflammatory drugs and hormonal medications such as oral contraceptives. Non-steroidal anti-inflammatory drugs are known to decrease the risk of death due to ovarian, colon, and breast cancer and, moreover, also the deaths due to myocardial infarction (Din et al., 2010; Olsen et al., 2011; Huang et al., 2015; Verdoodt et al., 2017). Furthermore, oral contraceptives are reported to decrease the overall risk of death and for example deaths due to ovarian cancer (Beral et al., 2008; Hannaford et al., 2010). The use of these medications might contribute in part to the decreased mortality among women with endometriosis.

Another limitation in our study is that it fails to reliably extend into older age groups. The mean age when entering to the study cohorts was 36 years and after the follow-up was 53 years. Therefore, data on women older than 75 years of age are limited. Many diseases have their highest incidences in older ages, including many cancers, dementia, or Alzheimer's disease (Naghavi et al., 2017). Thus, our results can only be generalized to midlife women.

The potential presence of several types of bias may explain at least part—or even all—of the lower all-cause mortality seen among women with endometriosis. A difference in the overall mortality between the cohorts was present already at the time of the index surgery and persisted 24 years. This suggests the difference to be drawn by the factors other than endometriosis per se. Thus, the present results can be applied only to midlife women with surgically verified endometriosis, and caution is needed when interpreting the results in terms of causality.

During the study period the two most common causes of death among working-aged Finnish women were neoplasms and diseases of the circulatory system, followed by causes related to alcohol, accidents, and suicides. The risk of death due to any cancer was decreased among women with surgically verified endometriosis. After adjustments for potential confounders, there were 12% fewer cancer deaths in the endometriosis cohort. At baseline any gynecological organ removal (hysterectomy, unilateral or bilateral oophorectomy, or both) occurred in 38% of the endometriosis cohort and 3% of the reference cohort. Even though these procedures cannot be separated from endometriosis treatment, they account partly for the decreased cancer deaths. Indeed, the sensitivity analysis showed that the adjusted MRR for ovarian cancer was increased only in women with endometriosis who had no baseline gynecological organ removals. In addition to the sensitivity analyses, the proportion of the association between endometriosis and mortality explained by the various treatments or interventions (including organ removal) has not been addressed in this analysis. As morbidity studies to date suggest, treatments, and in particular organ removal, may play an important role on the causal pathway(s).

The association of endometriosis and favorable prognosis of ovarian cancer has been reported previously (Melin et al., 2011; Kim et al., 2014). The focus of the present study was, however, on the mortality in women with endometriosis as followed from the index surgery due to endometriosis but not from cancer diagnosis. Therefore, the results of our study cannot be interpreted in terms of cancer survival.

Mortality due to breast cancer was decreased in women with surgically verified endometriosis compared to the reference cohort before and after adjustment for important risk factors, such as parity. Parity and breast feeding are known to decrease the risk of breast cancer (Lambertini et al., 2016; Victora et al., 2016). Moreover, the risk of breast cancer and breast cancer deaths is also reduced by oophorectomy (Nichols et al., 2011; Parker et al., 2013). The sensitivity analysis included all gynecological organ removals (hysterectomy and/or oophorectomy/-ies) and showed that only women with gynecological organ removals had significantly decreased MRR for breast cancer. Hormonal replacement therapy, or the lack of it, might affect breast cancer mortality, but, unfortunately, we lacked data on the possible use of hormonal replacement therapy in our study cohort. The numbers of deaths due to other cancers are too few to reliably assess the possible differences associated to endometriosis.

Compared to the reference cohort, 45% fewer deaths due to cardiovascular diseases were reported for women with surgically verified endometriosis. This finding was further strengthened following disease-specific calculations for ischemic heart disease and cerebrovascular disease, where the adjusted risk of death was also decreased. Moreover, in the sensitivity analysis gynecological organ removals had no effect on the difference. Recent studies have shown that oophorectomy and early menopause increase the mortality for cardiovascular disease (Gong et al., 2016; Muka et al., 2016; Evans et al., 2017; Mytton et al., 2017). In addition, in a recent North American study even hysterectomy without oophorectomy when performed to women aged 35 years or under increases the risk of cardiovascular conditions (Laughlin-Tommaso et al., 2018). However, the role of parity or hormonal replacement therapy is unclear (Santen et al., 2010; Jacobs et al., 2012; Tuomikoski and Mikkola, 2014; Boardman et al., 2015; Magnus et al., 2017). No adjustments were made in this study for the major cardiovascular risk factors—hypertension, diabetes, and hypercholesterolemia—nor for lifestyle factors, due to the unavailability of such information (Mosca et al., 2011). The decrease in cardiovascular mortality might partly relate to the risk factors. Studies with longer follow-up times are likely to clarify these results as cardiovascular diseases occur more often in advanced age.

Alzheimer's disease has been linked to chronic inflammation (Nogueira et al., 2015). However, according to a recent Canadian study mortality rate due to Alzheimer's disease in women increases only after 75 years of age and the mean age at death is 86 years (Park, 2016). In the present study, the mean age of women at the end of follow-up was 54, precluding from a reliable assessment of the potential relationship between endometriosis and Alzheimer's disease.

In conclusion, the overall mortality in midlife was lower in women with surgically verified endometriosis when compared to the reference cohort. The adjusted cause-specific mortality due to cancer, circulatory diseases, cardiovascular diseases, alcohol-related causes, accidents and violence, and diseases of the digestive and respiratory systems were all decreased. We speculate that the decreased mortality is significantly due to different characteristics and factors related to women's lifestyle and/or increased medical attention and care received among women with surgically verified endometriosis. There is a need for more studies on this issue.