Is Pelvic Inflammatory Disease a Risk Factor for Ovarian Cancer?

Infections and inflammation are well-established etiologic factors for several types of cancer and are estimated to contribute to the development of up to 25% of all cancer cases worldwide (1). In 1999, Ness and Cottreau (2) suggested that a number of potential risk factors for ovarian cancer, including endometriosis, talc use, and pelvic inflammatory disease (PID), may act through a common inflammatory mechanism. In addition, studies have shown that tubal ligation and hysterectomy protects against ovarian cancer (3) and this protective effect may also be related to inflammation, as these surgical procedures potentially prevent inflammatory reagents from reaching the fallopian tubes and the ovaries (2). Previously, all histotypes of ovarian cancer were believed to originate from the ovarian surface epithelium, and inflammation hereof was initially suggested to cause malignant transformation (2). However, recent studies suggest that serous ovarian cancer originates in the fallopian tube and only involves the ovary secondarily (4). Thus, inflammation of the tubes may also be important for the development of this tumor type (5).

PID is an infection-induced inflammation of the upper genital tract, including the uterus, fallopian tubes, ovaries, and pelvic peritoneum, caused by ascending pathogens from the lower genital tract (6). Knowledge on the true lifetime risk of PID among women in the Western world is rather sparse, and the few published results show great variation with cumulative incidence rates ranging from 4% to 20% (7–10).

The existing literature on the association between PID and risk of ovarian cancer is conflicting; some studies find an increased risk associated with PID (11–14), whereas others do not find an association (15–19). In addition, many of the previous studies had methodological constraints. For example, the majority used a case–control design where information about PID was based on self-reporting, thereby introducing a potential risk of recall bias and misclassification. To date, only 2 cohort studies have been performed, both including clinically verified diagnoses of PID (11, 12). However, these cohort studies had other methodological limitations such as the inclusion of lower genital tract infections in the definition of PID, a short follow-up period, and a low number of ovarian cancers. Furthermore, ovarian cancer is a heterogeneous disease with proposed differences in risk factors, origin, and genetic mutations among histotypes (4, 20), but few previous studies have investigated the association between PID and ovarian cancer risk according to histotype (15, 18). Using data from the Ovarian Cancer Association Consortium (OCAC), we have recently conducted a pooled analysis, representing the largest study on PID and risk of ovarian cancer, and no association between PID and risk of ovarian cancer overall was observed; however, indications of differences across histotypes of ovarian cancer were noted (21). Nevertheless, that study relied on self-reported information on PID and therefore has some of the same limitations as most previous studies.

The conflicting evidence regarding the possible role of PID in ovarian carcinogenesis and the methodological limitations of previous studies prompted us to conduct a large nationwide cohort study using data from the Danish health registries to investigate the association between PID and risk of epithelial ovarian cancer. The present register-based cohort study will be the largest on this research topic to date. We expect that the large number of study subjects included and the long follow-up period with up to 35 years of follow-up will enable us to evaluate risk patterns with high statistical precision and to assess histotype-specific risks as well as the effect of timing of PID and number of PID episodes.

The Danish Civil Registration System was established in 1968, and since then, all citizens have been registered with a unique personal identification number encoding date of birth and sex (22). The personal identification number is used universally in the Danish society, including in all Danish health registries ensuring accurate record linkage of information between registries. The Civil Registration System also holds information about place of birth, continuously updated information on vital status (date of death and data on migration), and identifiable information about parents and live born offspring. From this registry, we extracted data (personal identification number and information on vital status) for all women born between January 1, 1940 and December 31, 1970, who were alive and living in Denmark during the study period from January 1, 1978 to December 31, 2012 (n = 1,319,155).

The Danish National Patient Registry was established in 1977 and contains information on all diagnoses and surgical procedures performed at Danish hospitals (23). Initially, it covered only inpatient contacts, but after 1994, outpatient and emergency room contacts were also included. In the present study, PID was defined as an upper genital tract infection, including endometritis, salpingitis, oophoritis, pelvic peritonitis, and tubo-ovarian abscess and coded according to the 8th revision of the International Classification of Diseases (ICD-8) during 1977–1993 and the 10th revision of the International Classification of Diseases (ICD-10) during 1994–2011 (Supplementary Table S1). Using the personal identification numbers, the study cohort was linked to the National Patient Registry to identify all women with a diagnosis of PID from January 1, 1978 to December 31, 2011.

From the National Patient Registry, we also obtained information on bilateral oophorectomy, which was used as a censoring variable in the analyses. Potential confounders were selected a priori on the basis of the current knowledge of ovarian cancer etiology and availability in the registries and included hysterectomy, tubal ligation, endometriosis, parity status, and use of oral contraceptives. Information on endometriosis, hysterectomy, and tubal ligation was obtained from the National Patient Registry. The surgical procedures were coded using the Danish Classification of Surgical Procedures and Therapies from 1977 to 1995 and from 1996 and onward, the Nordic Classification of Surgical Procedures. Information on parity status was obtained from the Fertility Database, which was established in 1980, and contains information on number of live-born children for all women born from 1930 and onward and is considered complete for women born in 1945 or later (24). Finally, data on use of oral contraceptives were retrieved from the National Prescription Registry, which contains data on all prescriptions drugs dispensed at Danish pharmacies since 1995, including data on the Anatomical Therapeutic Chemical (ATC) classification codes (Supplementary Table S1; ref. 25).

All incident cases of epithelial ovarian cancer in the study cohort were identified by linkage to the Danish Cancer Registry. Since 1943, this registry has collected information on all cases of incident cancer in Denmark (26) and the registration of gynecologic cancers is considered virtually complete (27). Until 2003, the registry was based on notification forms from the diagnosing physicians, supplemented by linkage to the Registry of Causes of Death and the National Patient Registry to ensure completeness of data. Since 2004, the registry is based entirely on recordings from several Danish health registries, mainly the Danish National Patient Registry. In the Cancer Registry, ovarian cancers were coded according to the International Classification of Diseases, 7th edition (ICD-7) from 1943 to 1977 and according to the International Classification of Diseases, 10th edition (ICD-10) and the International Classification of Diseases for Oncology, 3rd edition (ICD-O-3) from 1978 and onward. All ovarian cancer cases included during the study period were identified using the ICD-10 code C56 and an ICD-O-3 morphology code for epithelial ovarian cancer. The ICD-O-3 morphology codes were further used to classify the specific histotypes of epithelial ovarian cancer (Supplementary Table S1). For all histotypes, only invasive cancers (i.e., with “3” as the last digit in the morphology code) were included.

All women in the study cohort were followed for development of histologically verified epithelial ovarian cancer from January 1, 1978 until date of bilateral oophorectomy, date of death, date of emigration, or December 31, 2012, whichever came first. We excluded women with a diagnosis of ovarian cancer (n = 220) or bilateral oophorectomy (n = 6) before start of follow-up, leaving 1,318,929 women in the final study population (Fig. 1).

The association between PID and risk of ovarian cancer was explored using Cox regression models to estimate HRs and corresponding 95% confidence intervals (CI). We conducted analyses for overall ovarian cancer and for the following histotypes: serous, mucinous, endometrioid, clear cell, and other epithelial ovarian cancers (Supplementary Table S1). Age was used as the underlying time scale to ensure comparison of women of the same age. We included a lag-phase of 1 year from diagnosis of PID to reduce the possibility of reverse causation or early cancer symptoms being misinterpreted as an episode of PID. Therefore, time at risk as exposed started one year after the first diagnosis of PID. Analyses were adjusted for parity (0, 1, 2, and ≥3), tubal ligation (yes/no), hysterectomy (yes/no), and endometriosis (yes/no).

In addition to PID status (ever vs. never PID), we also investigated the risk of ovarian cancer according to number of PID episodes, age at first PID, and time since first PID. To count as a new episode of PID, a PID diagnosis had to be at least 2 months after the previous diagnosis. Number of PID episodes was categorized as 1, 2, or ≥3. Age at first PID was categorized as <25, 25–29, 30–34, and ≥35 years, whereas time since first PID was categorized as 1–4, 5–9, 10–19, and ≥20 years. For these categorical PID variables, tests for trend were conducted. Both the PID exposure variables and the selected confounders were included as time-dependent variables (i.e., all women in the cohort accrued person-time in the unexposed group until a diagnosis of PID and thereafter contributed person-time in the exposed group). In a subset analysis, we further adjusted for use of oral contraceptives, as this information was available from 1995 and onward in the Danish Prescription Registry. However, as this additional adjustment did not alter the results considerably, these subgroup results are not presented further. The significance of variables was tested using Wald tests. Level of statistical significance was set at 0.05. All P values were 2-sided. Statistical analyses were conducted using the statistical software package R, version 3.0.2. The study was approved by the Danish Data Protection Agency.

The risk of ovarian cancer associated with PID was analyzed for 1,318,929 women. The median length of follow-up was 35.0 years (interquartile range, 33.4–35.0), resulting in 40,336,107 person-years of follow-up. A total of 81,281 women were diagnosed with PID during the study period and 5,356 women developed epithelial ovarian cancer (2,784 serous, 651 mucinous, 738 endometrioid, 307 clear cell, and 876 other epithelial ovarian cancers). Characteristics of women in the cohort are presented in Table 1. The higher age at entry and the shorter follow-up in the exposed period reflect the study design with time-dependent variables, as all women accrued person-years in the unexposed group until a diagnosis of PID.

Table 2 shows the HRs for the association between PID and risk of ovarian cancer. The risk of ovarian cancer overall was not associated with a history of PID (HR, 1.05; 95% CI, 0.92–1.20), age at first PID, time since first PID, or number of PID episodes. However, in histotype-specific analyses, we noted a statistically significantly increased risk of serous ovarian cancer among women with PID (HR, 1.19; 95% CI, 1.00–1.41; P = 0.047). For serous cancer, we also observed a statistically significantly increased risk among women having PID before the age of 25 years (HR, 1.44; 95% CI, 1.01–2.04). The risk estimates for the remaining categories of age at PID were not statistically significant, but a decreasing risk of serous ovarian cancer with increasing age at first PID was noted (P_trend = 0.03). For analyses according to time since PID, no clear pattern in risk of serous cancer was noted (P_trend = 0.63); however, the risk estimate was highest for the category 1 to 4 years since PID. Concerning risk of serous ovarian cancer according to number of PID episodes, we observed a pattern of an increasing risk with increasing number of episodes (P_trend = 0.04). For the nonserous histotypes, no marked associations with any of the PID variables were found, except for a suggestion of an increased risk of clear cell and mucinous ovarian cancer among women with PID at the age of 35 years or older.

In this large nationwide cohort study, we observed no association between PID and risk of ovarian cancer overall. However, in histotype-specific analyses, we found a modestly increased risk of serous ovarian cancer among women with a history of PID after adjustment was performed for parity, endometriosis, hysterectomy, and tubal ligation, whereas no convincing associations were noted for any of the other histotypes of ovarian cancer.

Inflammation is a well-established contributing factor in the development of several cancer types (1) and has also been proposed to be involved in ovarian carcinogenesis (2, 28). Inflammation involves the release of free radicals, growth factors, cytokines, and prostaglandins with the potential for genetic and epigenetic changes to the DNA, including mutations in tumor suppressor genes, thus increasing the risk of neoplastic transformation (1). In addition, cell death associated with inflammation results in a compensatory increased cell proliferation, thus increasing the risk of DNA replication errors (29).

Results from the existing literature on the association between PID and risk of ovarian cancer have been inconsistent. In line with our results, 2 case–control studies (16, 17) and 1 pooled analysis of 13 case–control studies (21) also found no association between PID and risk of ovarian cancer overall. Conversely, results from 1 case–control study (14) and 2 cohort studies (11, 12) demonstrated an increased risk of ovarian cancer among women with a history of PID with relative risks ranging from 1.9 to 4.0. However, ovarian cancer is a highly heterogeneous disease with differences in risk factors, underlying genetic mutations and proposed origin between histotypes (4), and in analyses of all ovarian cancers combined, histotype-specific associations may be missed. However, only few studies have investigated the role of PID in ovarian carcinogenesis according to histotype (15, 18, 21). Results from our recent, large pooled analysis of 13 case–controls studies indicated a differential impact of PID on the various histotypes of ovarian cancer with an association suggested only for the low-grade serous subtype (21). The other 2 studies providing histotype-specific results were both included in the pooled analysis from OCAC.

Given the hypothesis that serous ovarian cancer originates in the fallopian tube and involves the ovary secondarily (30), it is likely that salpingitis—which is included in the definition of PID—has a particular impact on the risk of serous tumors (5). In support hereof, chronic salpingitis has been found more often in fallopian tube specimens from women with serous ovarian tumors than in specimens from women with non-serous ovarian tumors or non-neoplastic ovaries (31). Furthermore, chronic salpingitis has also been implicated in the pathogenesis of serous fallopian tube cancers (32). In accordance with these observations, we also observed an increased risk of the serous subtype among women with a history of PID. However, the association was modest and only just statistically significant. Unfortunately, we had no information on grade and could therefore not investigate the risk of high- and low-grade serous cancers separately. We also observed suggestions of an increased risk of clear cell and mucinous ovarian cancer with PID after 35 years of age. These are likely spurious findings, as PID status was not associated with an increased risk of these histotypes and may be due to chance.

Strengths of the present study include the population-based design using nationwide high-quality registries, the large study population, the large number of ovarian cancers, the long follow-up, and the information on physician-verified PID diagnoses from registries rather than self-report thus eliminating recall bias, which was a limitation in most previous studies. Moreover, the completeness of the registries resulted in virtually no loss to follow-up and identification of all incident cases of histologically verified ovarian cancer, therefore our results are not influenced by selection bias. Furthermore, we were able to adjust for several important potential confounders, including parity, endometriosis, tubal ligation, hysterectomy, and oral contraceptive use.

However, some limitations should also be mentioned. First, some misclassification of PID exposure is present in our data due to the lack of information on PID diagnosed before 1978, but we tried to minimize misclassification caused by left truncation of the registries by restricting the study population to women born between 1940 and 1970. In addition, we had no information on PID treated in non-hospital settings and in outpatient clinics before 1995. Furthermore, some PID episodes are subclinical and will not be captured. Our unexposed population therefore contains some women with a PID likely resulting in an underestimation of the association. However, primarily having hospitalized women means that only the most severe cases of PID are included. As severity of PID has been shown to be related to risk of tubal factor infertility (33), it seems reasonable to assume that severe PID is associated with a greater inflammatory response and therefore perhaps also a greater risk of malignant transformation. Thus, as we mostly lack information on mild cases of PID, this may not have had a large impact on our results. Second, an episode of PID may have been caused by an undiagnosed ovarian cancer or early cancer symptoms may have been misinterpreted as an episode of PID. However, we believe this is not a major problem in our study, as most women referred to hospital for PID will have an ultrasound scan performed, thereby revealing a potential underlying ovarian cancer. Furthermore, we included a lag-phase of 1 year from first PID episode to reduce the influence of reverse causation. Third, we cannot rule out that our results may have been influenced by unmeasured confounding from important ovarian cancer risk factors that we had no information on (e.g., menopausal status). Finally, despite the large sample size, we still had limited power in the analyses for non-serous ovarian cancers.

In conclusion, this study demonstrated a modestly increased risk of serous ovarian cancer among women diagnosed with PID. In contrast, no convincing associations between PID and risk of the other histotypes of ovarian cancer were observed. Although the increased risk of serous ovarian cancer associated with PID was statistically significant, the association was modest in strength, indicating that PID is not a strong risk factor for ovarian cancer.

No potential conflicts of interest were disclosed.

Conception and design: C.B. Rasmussen, A. Jensen, V. Albieri, K.K. Andersen, S.K. Kjaer

Development of methodology: C.B. Rasmussen, A. Jensen, K.K. Andersen, S.K. Kjaer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.B. Rasmussen, A. Jensen, V. Albieri, K.K. Andersen, S.K. Kjaer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.B. Rasmussen, A. Jensen, V. Albieri, K.K. Andersen, S.K. Kjaer

Writing, review, and/or revision of the manuscript: C.B. Rasmussen, A. Jensen, V. Albieri, S.K. Kjaer

Study supervision: S.K. Kjaer

This work was supported by the Mermaid project (MERMAID III).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.