Treatment variation by cancer alliance

This section describes an analysis of variation in treatment between the 19 Cancer Alliances defined for England in 2018. Cancer Alliances are geographic areas that bring together clinicians and managers from different hospital trusts and other health and social care organisations with the aim of coordinating the diagnosis and treatment of cancer patients in the local area. A map of these Cancer Alliances is shown in Figure 5 below.

Figure 5 Map of Cancer Alliances in England, as defined by NHS England in 2018. Image source NHS England.

1. North East and Cumbria
2. Lancashire and South Cumbria
3. West Yorkshire and Harrogate
4. Humber, Coast and Vale
5. Greater Manchester
6. Cheshire and Merseyside
7. South Yorkshire, Bassetlaw and North Derbyshire, Hardwick
8. West Midlands
9. East Midlands
10. East of England
11. North West and South West London (RM Partners)
12. North Central and North East London (UCLH Cancer Collaborative)
13. South East London
14. Somerset, Wiltshire, Avon and Gloucestershire
15. Thames Valley
16. Peninsula
17. Wessex
18. Surrey and Sussex
19. Kent and Medway

The analysis of geographic variation in treatment selected only stage 2-4 and stage unknown cancers. Stage 1 tumours were excluded as, unlike in stage 2-4 tumours, trial evidence has not demonstrated a major survival benefit of chemotherapy for low-grade stage 1 tumours. As 96.3% (n=3,091) of stage 1 tumours were treated with primary surgery only or surgery with adjuvant chemotherapy, this suggests minimal variation in treatment pathways (Figure 1). Owing to how few were present in the cohort, tumours with a non-specific site morphology were also removed (n=62; 0.4%). This left a cohort for analysis of 13,889 tumours. Patient demographics and tumour characteristics for the analytical sample are provided in Appendix 2.

Findings are presented below as both funnel plots and results tables. Each point on a funnel plot represents a geographical area (in this case, a Cancer Alliance). The standard error is shown on the horizontal axis and provides an indication of the number of tumours diagnosed within the Cancer Alliance. Estimates from Cancer Alliances with a greater number of tumours are more precise, appearing further to the right-hand side of the plot and represented by bigger red markers than Cancer Alliances with fewer tumour diagnoses. The percentage difference in the probability of a treatment or treatment combination is shown on the vertical axis relative to the population average (all tumours combined). A Cancer Alliance with an estimate above the average (indicated by a solid black horizonal line) suggests that tumours within the geography were more likely to receive treatment than the population average, with estimates below the line indicating a lower probability.

Two pairs of dashed lines are included on each funnel plot that represent the bounds of statistical confidence around the average value. The inner set of dashed lines represents two standard deviations (SD) from the population average and the outer set represents three SD, being approximately equivalent to 95.0% and 99.7% confidence intervals, respectively. Any observation plotted outside of these dashed lines will have a confidence interval that does not include the average value, and may therefore indicate a systematic deviation in clinical practice that warrants further investigation. However, some random variation in the probability of treatment is expected between regions such that some points will sit outside the dashed lines through chance alone. This should be taken into consideration when interpreting funnel plots (for example, five out of every 100 observations are likely to lie outside the two SD funnel).

Within each accompanying table, Cancer Alliances highlighted in blue had treatment probabilities that were significantly higher (p=<0.05) than the average, and those highlighted in red had significantly lower probabilities. These represent Cancer Alliances that fall outside the innermost pair of dashed lines in the corresponding funnel plot (two SDs).

Given the variation in treatment across patient and tumour characteristics, as shown in Figures 1-4, three different models were developed for each of the treatment scenarios reported through this section:

• Model 1 represents an unadjusted analysis, which compares crude treatment probabilities for tumours diagnosed by each Cancer Alliance.  
• Model 2 adds adjustment for differences between Cancer Alliances in the distribution of patient age, tumour morphology and tumour stage.
• Model 3 adjusts for the same factors as Model 2, plus area income deprivation and Charlson comorbidity score.

While funnel plots are shown for the minimally (Model 1) and maximally (Model 3) adjusted models only, findings from all three models are presented in table form to allow comparisons according to differing levels of covariate adjustment.

Ovarian cancer is an aggressive disease and long-term survival relies on access to treatment. This initial analysis looks at differences between Cancer Alliances in the proportions of diagnosed tumours that received any treatment, defined here as surgery or chemotherapy either alone or in combination.

The weighted average probability of a stage 2-4 and unknown stage ovarian cancer receiving any treatment was 73.8% (Table 1). Despite adjustment for a range of factors associated with differences in the treatment pathway, funnel plot B in Figure 6 shows that geographic variation in treatment delivery remained. Moreover, the probability of any treatment fell more than two SDs below the population average for four Cancer Alliances, while five Alliances had probabilities more than two SDs greater than the average. These results indicate not just that there is marked variation in overall treatment between Cancer Alliances in England, but that some of these differences may not be explained by chance alone and warrant further investigation. A complete table of coefficients is reported in Appendix 3. 

Figure 6 Geographic variation in the probability of receiving any treatment versus no treatment, excluding stage 1 disease, 2016 to 2018 (Source: CAS AV2018, HES, SACT)

Table 1 - Probability of receiving any treatment versus no treatment, excluding stage 1 disease - n=13,889 for all 3 models

Notes:

1. The cohort contains cases of ovarian, tubal and primary peritoneal cancer diagnosed between January 2016 & December 2018 inclusive. Stage 1, borderline and non-specific site tumours are excluded, along with cancers diagnosed via death certificate.
2. Treatment data are compiled from the cancer registry, Hospital Episode Statistics (HES) admitted patient care dataset, and the Systemic Anti-Cancer Therapy (SACT) dataset. Data were captured during the primary course of treatment (the period up to nine months following diagnosis). Treatments dated outside of this window are not considered.
3. In a small minority of cases, tumours were documented as receiving both systemic anti-cancer therapy and major surgical resection on the same day. These cases are coded to the neoadjuvant anti-cancer therapy treatment group:

• Model 1 includes Cancer Alliance only
• Model 2 as Model 1, plus adjustment for patient age, tumour morphology and tumour stage
• Model 3 as Model 2, plus area income deprivation and Charlson comorbidity score

The nature of ovarian cancer means that surgery is essential in the large majority of cases to remove (debulk) the tumour. While the results above provide an indication of variation in women receiving any treatment, the analysis below looks specifically at the probability of surgery either alone or in combination with other therapies.

The average probability of Stage 2-4 and unknown stage ovarian cancer being treated with any surgery was 51.0% (Table 2). The funnel plots in Figure 7 indicate large geographic variation in the delivery of surgery between Cancer Alliances, even after adjustment for other factors, where six Cancer Alliances had surgery probabilities two SDs above the average and five had probabilities two SDs below the average (Figure 7b). A full table of coefficients is provided in Appendix 4.The probability of treatment with surgery was particularly low in one Cancer Alliance and warrants investigation to determine whether the figure represents true regional variation in clinical practice or is the product of either poor treatment reporting or unadjusted differences in the type of tumour or patient being treated.

Figure 7 Geographic variation in the probability of receiving any surgery versus no surgery, excluding stage 1 disease, 2016 to 2018 (Source: CAS AV2018, HES, SACT)

Table 2 - Probability of receiving any surgery versus no surgery, excluding stage 1 disease (n=13,889)

Notes:

1. The cohort contains cases of ovarian, tubal and primary peritoneal cancer diagnosed between January 2016 and December 2018 inclusive. Stage 1, borderline and non-specific site tumours are excluded, along with cancers diagnosed via death certificate.
2. Treatment data are compiled from the cancer registry, Hospital Episode Statistics (HES) admitted patient care dataset, and the Systemic Anti-Cancer Therapy (SACT) dataset. Data were captured during the primary course of treatment (the period up to nine months following diagnosis). Treatments dated outside of this window are not considered. In a small minority of cases, tumours were documented as receiving both systemic anti-cancer therapy and major surgical resection on the same day. These cases are coded to the neoadjuvant anti-cancer therapy treatment group.
3. Model 1 includes Cancer Alliance only; Model 2 as Model 1, plus adjustment for patient age, tumour morphology and tumour stage; Model 3 as Model 2, plus area income deprivation and Charlson comorbidity score.

All women with an ovarian cancer diagnosis should receive chemotherapy to treat and help manage the disease, with the exception of stage 1 cancers (for which primary surgery only is typical), certain less common tumour types, circumstances where the risks from chemotherapy may outweigh the benefits (such as due to comorbidity), and patient choice.

Of tumours in the stage 2-4 and unknown stage cohort, the average probability of treatment with chemotherapy was 66.5% (Table 3). Following adjustment for a broad range of factors, including patient age and tumour morphology, funnel plot B in Figure 8b shows five Cancer Alliances with chemotherapy rates two SDs above the average (one notably so) and three with chemotherapy rates two SDs below the average. A full table of coefficients is provided in Appendix 5.

Figure 8 Geographic variation in the probability of receiving any chemotherapy versus no chemotherapy, excluding stage 1 disease, 2016 to 2018 (Source: CAS AV2018, HES, SACT)

Table 3 - Probability of receiving any chemotherapy versus no chemotherapy, excluding stage 1 disease (n=13,889)

Notes:

1. The cohort contains cases of ovarian, tubal and primary peritoneal cancer diagnosed between January 2016 and December 2018 inclusive. Stage 1, borderline and non-specific site tumours are excluded, along with cancers diagnosed via death certificate.
2. Treatment data are compiled from the cancer registry, Hospital Episode Statistics (HES) admitted patient care dataset, and the Systemic Anti-Cancer Therapy (SACT) dataset. Data were captured during the primary course of treatment (the period up to nine months following diagnosis). Treatments dated outside of this window are not considered. In a small minority of cases, tumours were documented as receiving both systemic anti-cancer therapy and major surgical resection on the same day. These cases are coded to the neoadjuvant anti-cancer therapy treatment group.
3. Model 1 includes Cancer Alliance only; Model 2 as Model 1, plus adjustment for patient age, tumour morphology and tumour stage; Model 3 as Model 2, plus area income deprivation and Charlson comorbidity score.

Chemotherapy is increasingly used prior to surgery (neoadjuvant).  This approach is often used in two cases which are specific to advanced disease. Firstly, if a patient is too unwell to undergo surgery, chemotherapy can start to treat the tumour while waiting for the patient to recover. Secondly, if the multidisciplinary team (MDT) considers it unlikely that complete tumour resection (removal) will be feasible during primary surgery, neoadjuvant chemotherapy may be used to make the tumour more operable, reducing the risk of surgical complications and morbidity².

The following analysis explores geographic variation in the probability of receiving primary surgery with adjuvant chemotherapy versus neoadjuvant chemotherapy with interval debulking surgery. Accordingly, the analysis 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 (Table 4). Figure 9b shows regional variation after adjustment for patient demographics and tumour characteristics associated with treatment, with three Cancer Alliances falling two SD above the average and five regions falling two SDs below the average. The maximally-adjusted probability of treatment with primary surgery and adjuvant chemotherapy of one Cancer Alliance was markedly higher than others and requires further investigation. A full table of coefficients from the underlying models can be viewed in Appendix 6.

Figure 9 Geographic variation in the probability of receiving primary surgery and adjuvant chemotherapy versus neoadjuvant chemotherapy with interval debulking surgery, excluding stage 1 disease, 2016 to 2018 (Source: CAS AV2018, HES, SACT)

Table 4 - Probability of receiving primary surgery with adjuvant chemotherapy versus neoadjuvant chemotherapy and interval debulking surgery, excluding stage 1 disease (n=6,065)

Notes:

1. The cohort contains cases of ovarian, tubal and primary peritoneal cancer diagnosed between January 2016 and December 2018 inclusive. Stage 1, borderline and non-specific site tumours are excluded, along with cancers diagnosed via death certificate.
2. Treatment data are compiled from the cancer registry, Hospital Episode Statistics (HES) admitted patient care dataset, and the Systemic Anti-Cancer Therapy (SACT) dataset. Data were captured during the primary course of treatment (the period up to nine months following diagnosis). Treatments dated outside of this window are not considered. In a small minority of cases, tumours were documented as receiving both systemic anti-cancer therapy and major surgical resection on the same day. These cases are coded to the neoadjuvant anti-cancer therapy treatment group.
3. Model 1 includes Cancer Alliance only; Model 2 as Model 1, plus adjustment for patient age, tumour morphology and tumour stage; Model 3 as Model 2, plus area income deprivation and Charlson comorbidity score.

Table 5 shows the maximally-adjusted results from Table 1 (any treatment), Table 2 (any surgery) and Table 3 (any chemotherapy). Collectively, they indicate pronounced geographic variation in treatment delivery after accounting for differences in the regional distribution of various patient demographics and tumour characteristics.

Table 5 - Summary of maximally-adjusted geographic variation in any treatment, surgery and chemotherapy (Model 3; n=13,889) 

Notes:

1. The cohort contains cases of ovarian, tubal and primary peritoneal cancer diagnosed between January 2016 and December 2018 inclusive. Stage 1, borderline and non-specific site tumours are excluded, along with cancers diagnosed via death certificate.
2. Treatment data are compiled from the cancer registry, Hospital Episode Statistics (HES) admitted patient care dataset, and the Systemic Anti-Cancer Therapy (SACT) dataset. Data were captured during the primary course of treatment (the period up to nine months following diagnosis). Treatments dated outside of this window are not considered. In a small minority of cases, tumours were documented as receiving both systemic anti-cancer therapy and major surgical resection on the same day. These cases are coded to the neoadjuvant anti-cancer therapy treatment group.
3. Model 1 includes Cancer Alliance only; Model 2 as Model 1, plus adjustment for patient age, tumour morphology and tumour stage; Model 3 as Model 2, plus area income deprivation and Charlson comorbidity score.