The impact of comorbidity and stage on ovarian cancer mortality: A nationwide Danish cohort study

Ovarian cancer is the leading cause of death from gynaecological cancer in western countries. It has a poor prognosis, with five-year survival rates ranging from 26% to 51% in Europe [1]. As the incidence of ovarian cancer increases sharply with age, many patients have one or more other chronic diseases, i.e., comorbidities [2, 3]. Comorbidity is an important predictor of prognosis in patients with chronic diseases, including cancer [4, 5]. Among women with ovarian cancer, the presence of comorbidity may substantially influence the diagnostic work-up, alter treatment efficacy, and affect survival.

Few studies to date have examined the impact of comorbidity on ovarian cancer survival [6‐11]. The presence of comorbidity at time of diagnosis was found to have a negative impact on prognosis and survival in two studies [7, 9], but not in others [6, 8, 10, 11]. Based on hospital discharge registry data, we recently reported that one-year mortality in a Danish regional study was twice as high in ovarian cancer patients with severe comorbidity as in those without recorded comorbidity [12]. However, the study was limited by lack of information on the stage of cancer and by our inability to exclude patients with borderline tumours.

The presence of comorbidities may influence stage at diagnosis, which in turn is a strong predictor of survival in ovarian cancer patients. It is possible that patients with comorbidities experience delay in diagnosis, resulting in a more advanced cancer stage. We thus hypothesized that poorer survival among ovarian cancer patients with severe comorbidity, compared with those without comorbid conditions, may be explained by a higher prevalence of advanced cancer at diagnosis.

The Danish Cancer Registry records stage information on incident ovarian cancers. Using data from this Registry, we conducted a nationwide study to determine the prevalence of comorbidity by stage of ovarian cancer, to estimate the impact of comorbidity on survival and mortality, controlling for cancer stage, and to examine whether the effect of comorbidity on ovarian cancer mortality varies by cancer stage at diagnosis.

We conducted this nationwide study in Denmark, which has approximately 5.4 million inhabitants, from January 1, 1995 to December 31, 2005. The entire Danish population has free access to tax-supported medical care, including hospitalization. There is practically no private inpatient ovarian cancer treatment available. In 2004, surgery for ovarian cancer took place in 52 hospital departments [13], five of which were gynecologic oncologic centers. Thus, treatment of this cancer is quite decentralized in the Danish setting.

In this population-based nationwide study, we found a higher prevalence of comorbidity in patients with an advanced stage compared with a less advanced stage of ovarian cancer. One- and five-year mortality were nearly twice as high in patients with severe comorbidity compared to those without registered comorbidity, even after adjustment for stage. Thus the increased prevalence of more advanced stage did not entirely explain the association between greater comorbidity and higher mortality. We found further that the impact of severe comorbidity on one- and five-year mortality varied by cancer stage.

Among the strengths of our study were its large size and Denmark's uniformly organized health care system, allowing a population-based design with virtually complete follow-up. However, the accuracy of our findings depends on the quality of cancer registry and hospital discharge data. The Danish Cancer Registry is known to be more than 95% complete [14], and in a previous study in which pathology records were used to confirm diagnoses, we found that ovarian cancer diagnoses were correct in 97% of the cases in the Registry, [30]. This minimal selection bias should not be related to the presence of comorbidity, however, since comorbidity was independently recorded before the cancer diagnosis.

Cancer staging is based on a combination of pathologic, operative, and clinical assessments available at the time of diagnosis. Some misclassification of stage data may occur, which could result in residual confounding. There may also have been some inaccuracy in treatment data used in the study. These data, obtained from the Danish Cancer Registry, are limited to treatment given within the first four months following diagnosis; since Registry reporting forms are often completed in the early period of treatment planning, they may not reflect actual treatment. If complications stemming from comorbidity lead to changes in treatment, this could result in residual confounding.

We used the validated Charlson Comorbidity Index as a measure of comorbidity. When applied to administrative data information on comorbidity is based on ICD-codes. The comorbid diseases may be coded with different accuracy in the different administrative registries and misclassifications occur in most registries [31]. The Charlson index has been shown to have a high specificity [4], but a more variable sensitivity when compared with diagnoses abstracted from the medical charts [32]. It is thus possible that some patients with comorbid conditions may have been classified erroneously as having Charlson score 0. Similarly, patients with severe comorbidity may have been classified erroneously as having Charlson score 1–2. However, because comorbidity was independently recorded before the cancer diagnosis, any misclassification of comorbid conditions was probably unrelated to ovarian cancer stage.

Patients with comorbidities may experience a delay in diagnosis or they may actually be diagnosed earlier because they have a close relationship with the health care system. We found that the presence of severe comorbidity was associated with an advanced stage of ovarian cancer. If ovarian cancer progresses from Figo-stage I to IV this could suggest a delay in diagnosis. It has, however, been suggested that stage I and stage III may be different forms of the disease [33].

Our findings disagree with a Dutch population-based study by Mass et al. [6], which was restricted to approximately 500 patients with FIGO-stage II and III ovarian cancer. Using a slightly modified Charlson Comorbidity Index and adjusting for treatment, age, stage, and period of diagnosis, they concluded that comorbidity did not influence prognosis. Other cohort studies have had similar findings [8, 10, 11]. A Norwegian population-based cohort study (N = 571) examining the impact of several possible prognostic factors on survival found that comorbidity was a prognostic factor in univariate but not multivariate analyses [11]. One reason this study did not find an association may be its adjustment for residual tumour. Presence of a residual tumour is related to the aggressiveness of surgery and if comorbidity results in less aggressive surgery, residual tumour may be an intermediate in the causal pathway from comorbidity to death. In this situation, adjustment for residual tumour would be inappropriate. The effect of comorbidity on mortality may be mediated to a large degree by higher volume of residual tumour. A Dutch population-based study (N = 1,116) that adjusted for age, stage and treatment also did not find an independent effect of comorbidity on prognosis [10]. Similarly, an American hospital-based study reported an age-, stage- and symptom stage-adjusted mortality rate ratio of 1.04 in ovarian cancer patients with comorbidity compared to those with no comorbidity [8]. Its study population consisted of 137 ovarian cancer patients recruited during a period of almost 6 years, which could have introduced selection bias [8].

In accordance with our study, a negative impact of comorbidity on ovarian cancer mortality was found in an American population-based and in a German cohort study [7, 9]. In these studies the mortality rate ratios were adjusted for stage, but the impact of stage on survival was not reported. The current study corroborates the findings in our recent study [12] on this topic. Our observation of the impact of comorbidity on mortality also confirm and extend findings for other groups of cancer patients, including breast cancer, prostate cancer, colon cancer, and lung cancer patients [34].

Since stage did not entirely account for the differences in mortality in our study, other factors may explain the role of comorbidity as a negative prognostic factor. The presence of comorbidity in a cancer patient may influence treatment choices, which in turn affect prognosis and survival [10, 35]. The optimal treatment of patients with ovarian cancer is surgery and chemotherapy, with regimens depending on stage [36]. In advanced stages it is important to optimally debulk the tumour, and while this extensive surgery can be performed safely in patients with comorbid conditions [3], not all such patients receive this treatment [6]. It also is possible that following established treatment guidelines is not the best strategy for patients with multiple comorbidities [37], either because they cannot tolerate the adjuvant chemotherapy necessary after surgery or because the drugs used to treat their comorbid diseases may interact with those in chemotherapy regimens. As well, the toxicity of chemotherapy may be exacerbated by the side effects of the drugs that are used to treat comorbidities [38]. While we were able to adjust for treatment using data from the Danish Cancer Registry, information was lacking on the aggressiveness of the surgery performed or use of modified chemotherapy regimens. Adjustment for treatment in the analyses did not diminish the impact of severe comorbidity on the one-year MRR. However, despite adjustment for treatment, aggressiveness of treatment could vary in the comorbidity groups, as indicated a Norwegian study [11].

Because our study addressed all-cause rather than cause-specific mortality, patients could have died from their comorbidity or other causes not related to ovarian cancer. This may also explain some of the higher mortality in patients with comorbidity. However, it is difficult to distinguish between the contributions to mortality from the ovarian cancer itself and that from cancer complications or comorbidities. An example is death from heart-disease vs. death due to chemotherapy-related aggravation of pre-existing cardiac problems.

There is a need for clinicians to be aware of the possible presence of comorbidity in ovarian cancer patients in order to improve their treatment. Patients with regional spread/FIGO stage II and III ovarian cancer are a particularly important subgroup in this context.

In this population-based study of ovarian cancer patients, we found that the presence of severe comorbidity (Charlson score 3+) was associated with an advanced stage of cancer. Mortality was also higher among patients with comorbidities, but could not be entirely explained by the cancer stage.