Short-term mortality in ovarian, fallopian tube and primary peritoneal carcinomas across England

A disproportionately high number of women with ovarian cancer die within a year of diagnosis. To investigate the reasons behind this, a report was produced that explored the characteristics and prognostic factors of these women compared to those who survived longer than 12 months after diagnosis. Geographical variation in short-term mortality was also investigated.

Data from the NCRD were used to identify women diagnosed with ovarian cancers (excluding borderline malignancies) in England between 2013 and 2018.

To examine how mortality rates changed over the first year after diagnosis, patients were assigned to four groups:

• patients who died within 0 to 2 months from diagnosis
• patients who died within 2 to 6 months from diagnosis
• patients who died within 6 to 12 months from diagnosis
• patients who survived more than 12 months following diagnosis

Crude mortality rates were calculated to examine the distribution of patient demographics and tumour characteristics across these short-term mortality groups. Factors were assessed for statistically significant differences in mortality via chi-squared tests. Mixed effects logistic regression models were fitted for each of the short-term mortality groups to estimate the association of statistically significant factors with short-term mortality in ovarian cancer patients. To examine geographical variation, crude and case-mix adjusted mortality rates for each Cancer Alliance were calculated.

Data completeness rates for most data fields were generally good (>95% completeness). However, as for the treatment report cohort, the capture of performance status and stage at diagnosis was low, with data missing for 63.4% and 16.9% of patients respectively.

The derived Charlson comorbidity score correlated with short-term mortality but appeared to underestimate the true comorbidity burden of patients, with 84.3% of cases having been assigned a comorbidity score of zero, which is inconsistent with the age profile of the patient cohort.

Results from these analyses show that the short-term mortality rate for ovarian cancer patients remains high: 14% of women died within two months of diagnosis, and 30% died within the first year (Figure 7 below).

Figure 7 Short-tern mortality in patients diagnosed with ovary, fallopian tube or primary peritoneal carcinomas, 2013 to 2018.

This compares to a previous analysis available on the NCIN website of 2006-2008 cases, published in 2013, which shows a 15% two-month and 31% 12-month mortality rate. However, the 2013 analysis included patients with borderline tumours (who typically exhibit excellent long-term prognosis) and excluded primary peritoneal cancer patients (who have a relatively high short-term mortality rate), suggesting a modest overall improvement in short-term mortality rates from ovarian cancer in England over the last decade.

Mortality rates were impacted by several factors, with the greatest variation seen for women who died within two months following their diagnosis. For example:

1. Older age: 15.7% of women aged 70-79 years and 34.8% of women aged ≥80 years died within two months of diagnosis. Adjusting for confounding factors, women aged ≥80 years at diagnosis were 40% more likely to die within two months than those aged 70-79 years. 
2. Women diagnosed at a later or unknown stage of disease: 21.0% of women with stage 4 disease and 30.6% of those with unknown died within two months of diagnosis, versus 9.6% of those diagnosed with stage 3 disease. Compared to patients diagnosed at stage 1-3, women with unknown stage were 9.5 times more likely to die within two months of diagnosis following adjustment for confounding factors.
3. Burden of comorbidity: 34.9% of women with a score >2 died within two months of diagnosis, and 11.9% of those assigned a score of 0. Logistic regression models reported odds of mortality within two months that were 60% higher for women with a greater burden of comorbidities (34.9% of woman with a Charlson score of more score >2 than women with a score of 0.
4. Women living in areas with more deprived socioeconomic status: 16.2% of women in the most deprived quintile died within two months, versus 12.0% in the least deprived quintile, or 50% more likely to die within two months following adjustment for confounders.

Alongside these was the impact of route to diagnosis. As shown in Figure 8 below, mortality rates were highest in ovarian cancer patients who were diagnosed via an emergency presentation, with 30.8% having died within 2 months of diagnosis. By contrast, women diagnosed via the Two Week Wait referral system showed the lowest mortality rate within two months of diagnosis (4.6%). Patients diagnosed via a non-urgent GP referral generally had higher mortality as these patients may experience longer delays in diagnosis, allowing the disease more time to progress, leading to poorer survival. Differences remained following adjustment for confounding factors, whereby patients diagnosed via emergency presentation were 4.3 times more likely to die within the first two months following diagnosis than women diagnosed via a Two Week Wait route.

Figure 8 Short-tern mortality in patients diagnosed with ovary, fallopian tube or primary peritoneal carcinomas by route to diagnosis, 2013 to 2018.

There was also a trend observed of higher short-term mortality rates in patients diagnosed in an NHS secondary care trust that did not house a specialist gynaecological cancer centre compared to trusts that did (Figure 9). Following adjustment, differences were statistically significant only for the odds of short-term mortality between two and six months following diagnosis.

Figure 9 Short-tern mortality in patients diagnosed with ovary, fallopian tube or primary peritoneal carcinomas by whether the trust at diagnosis had a specialist gynaecological cancer centre, 2013 to 2018.

Crude and case-mix adjusted results indicate variation in short-term mortality at a Cancer Alliance level, but to a lesser extent than the variation seen in survival and treatment analyses. The short-term mortality analyses corrected for route to diagnosis, which was not included in the models for the survival or treatment analyses. This may explain the diminution of Cancer Alliance variation.  As with the survival and treatment analyses, it is likely that variation present at a more local level has been lost by aggregation of data at a Cancer Alliance level.