Geographic variation in ovarian, fallopian tube and primary peritoneal cancer treatment in England (November 2020)

One possible reason for regional disparities in age-standardised one- and five-year survival was variation in the local clinical management of disease. To explore this hypothesis, a report was produced that describes geographic variation in treatment between Cancer Alliances in England, and the extent to which these might be explained by regional differences in tumour and patient characteristics.

Results were also produced for circulation within gynaecological cancer multidisciplinary teams, at a provider level and made available to NHS staff via the CancerStats2 website - (this opens in a new page). Please note that this platform requires an N3/HSCN secure network connection. To ensure the best user experience, we encourage the use of modern web browsers such as Google Chrome, Mozilla Firefox or Microsoft Edge to access the platform. A small number of platform users have reported issues when opening reports using Internet Explorer.

Support was provided through data liaison teams to provide help with retrospectively updating and prospectively improving treatment data quality and completeness.

A cohort of ovarian cancers was created using data from the NCRD. Diagnoses in England were selected for the period between 2016 and 2018.

Where data were available, each tumour was linked to information describing the delivery of systemic anti-cancer therapy (hereafter ‘chemotherapy’) or major surgical resection during the primary (i.e., first) course of treatment, defined as treatment initiated during the nine months following diagnosis. These treatment data were obtained from the NCRD in the first instance and supplemented with additional routine data available through the Systemic Anti-Cancer Therapy (SACT) and Hospital Episode Statistics (HES) datasets.

Based on the type and ordering of treatment received, four binary treatment categories were created:

1. Any treatment versus no treatment
2. Surgery versus no surgery
3. Chemotherapy versus no chemotherapy
4. Primary surgery with adjuvant chemotherapy (i.e., surgery followed by chemotherapy) versus neoadjuvant chemotherapy with interval debulking surgery (i.e., chemotherapy first followed by surgery).

Each binary treatment category was included as an outcome in a linear probability model, adjusted for a range of patient factors associated with clinician and patient decision making. Results are interpretable as the percentage-point difference in the probability of treatment relative to the cohort average.

Stage 1 tumours were omitted from these analyses owing to a lack of variability in treatment for early-stage cancers (96.3% were treated with either primary surgery only, or primary surgery and adjuvant chemotherapy).

Three different levels of adjusted were applied:

• Model 1: an unadjusted analysis, comparing crude treatment probabilities for tumours diagnosed within each Cancer Alliance.
• Model 2: adjusted for differences between Cancer Alliances in the distribution of patient age, tumour morphology and tumour stage.
• Model 3: adjusted for the same factors as Model 2, plus area income deprivation and Charlson comorbidity score.

Linear probability models were adjusted to estimate geographic variation in treatment independent of differences in patient factors likely to differ between regions and influence the treatment pathway. Despite this, any observed geographic differences in treatment may be attributable to residual confounding rather than real disparities in clinical practice, such as differential routes to diagnosis (data unavailable at the time of analysis) or geographic differences in patient frailty (data not captured).

There were challenges with both the Charlson comorbidity index and patient performance status at diagnosis. The index scored a pre-defined subset of chronic comorbid conditions based on diagnosis codes extracted from HES Admitted Patient Care (APC) data. Although the index is known to correlate with patient treatment and survive, it does not capture the full burden of comorbid illness. For example, of the 17,155 tumours selected in the cohort, 82.8% (n=14,196) were assigned a comorbidity score of zero. Performance status, a key prognostic indicator, could not be included in the models described above due to high levels of missing data (57.9%). This was despite repeated outreach to the clinical community to stress the importance of routine capture of this data item by multidisciplinary teams.

Reported analyses do not consider treatments provided in private healthcare settings. Due to the absence of private healthcare data, tumours registered by multidisciplinary teams and subsequently treated in private institutions will have been incorrectly assigned to the ‘no major surgical resection or chemotherapy’ category. Accordingly, the true proportion of tumours that received treatment will be higher than reported, and with the possibility of differences in private treatment access between Cancer Alliances.

The most striking finding of this study was that within the overall cohort of 17,155 tumours, 21.9% (n=3,751) received no surgery nor chemotherapy.

Regarding patient demographics and tumour characteristics:

1. Tumours with missing stage data or stage 4 disease were much less likely to receive any treatment, with 28.2% and 60.7% respectively having received neither chemotherapy nor surgery.
2. Women with tumours classed as miscellaneous and unspecified were less likely to receive any treatment, with 89.1% receiving neither surgery nor chemotherapy.
3. Women with underlying medical conditions, identified by the Charlson comorbidity index, were less likely to receive surgery. Of women with a score >2, 55.8% received neither surgery nor chemotherapy.
4. Tumours in women aged >79 years were the least likely to receive any treatment, with 60.1% receiving neither chemotherapy nor surgery.

In terms of geographic variation, the weighted average probability of any treatment for stage 2-4 and unknown stage tumours was 73.8%. As shown in the funnel plot below, wide variations in the rates of any treatment were observed across Cancer Alliances even after adjustment for a range of factors associated with the treatment pathway.

Figure 4 Geographic variation in the probability of receiving any treatment versus no treatment, excluding stage 1 disease, 2016 to 2018.

Figure 5 Geographic variation in the probability of receiving surgery versus no surgery, excluding stage 1 disease, 2016 to 2018.

There was less variability with regard to chemotherapy treatment rates between the Cancer Alliances, indicating that variation in surgery rates was the main driver of the variability in overall treatment seen in Figure 4.

The analysis of primary surgery with adjuvant chemotherapy versus neoadjuvant chemotherapy with interval debulking surgery was restricted to the 6,065 tumours within the cohort of stage 2-4 and unknown stage cancers that were assigned to one of these two treatment groups. Of these tumours, the probability of primary surgery and adjuvant chemotherapy was 49.4% on average. Figure 6 shows the maximally adjusted probability of treatment with primary surgery and adjuvant chemotherapy.

Figure 6 Geographic variation in the probability of receiving primary surgery with adjuvant chemotherapy versus neoadjuvant chemotherapy with interval debulking surgery, excluding stage 1 disease, 2016 to 2018.

The variation in any treatment, any surgery and any chemotherapy rates between Cancer Alliances is presented in Table 1. Cancer Alliances with rates two SDs above the national average are denoted in blue, and those with rates two SDs below are denoted in pink. This table provided clinical teams with ready comparison of their own Cancer Alliance performance and proved a powerful tool supporting engagement with the findings of the OCAFP.

Table 1 - Summary of maximally adjusted geographic variation in any treatment, surgery and chemotherapy (Model 3; n=13,889), excluding stage 1 disease, 2016 to 2018.