10.1 Introduction

The West Midlands Cancer Intelligence Unit (WMCIU) was asked to investigate the incidence and mortality for primary care trusts (PCTs) within the region, with the aim of highlighting areas of possible concern.  ‘League tables’ are a common technique used for comparing performance of health care providers.  Ranking in this way does not allow for variation within an expected range, and even with the addition of confidence intervals spurious impressions can be created when interpreting these tables focus is naturally drawn to position.  An alternative method making use of Funnel Plots has been suggested1 to avoid ranking of PCTs whilst still highlighting outliers.

10.1.1 Method

Funnel plots have been produced using 2003-2005 incidence data from the WMCIU’s Cancer Registration Database and Office of National Statistics population and mortality data.  The cancer sites analysed in this chapter are:

Most non-melanoma skin cancers are detected early and are rarely thought to be life-threatening.  They are often diagnosed and treated on an outpatient basis and or in primary care, which leads to concerns about the completeness of registration.  For these reasons non-melanoma skin cancers are often excluded from comparative analyses of cancer data.

Using the numbers diagnosed and the directly age standardised rate, the age adjusted population for each PCT was calculated (for new PCT boundaries as defined in 2006 see Map 1.1 in Chapter One).  Directly European age standardised rates, on the y-axis, were plotted against the adjusted population on the x-axis.  The average rates were calculated by the mean directly age standardised rate of the PCTs.  Control limits were then calculated, assuming a Poisson distribution, at 2 and 3 standard deviations from the mean giving 95% and 99.8% control limits respectively.  This method was based on a template available from Eastern Region Public Health Observatory²;.

10.1.2 Interpreting the Results

The figures show the incidence and mortality rates plotted against the population after adjustment for age, with the West Midlands average rate (solid centre line), and its 95% control limits (thin dotted lines) and 99.8% control limits (thick dotted lines).  Each PCT in the West Midlands is represented by a coloured dot:

Very High - when the PCT lies above the upper 99.8% control limit, and the PCT has a rate significantly higher than the regional rate High – when the PCT lies between the upper 95% and upper 99.8% control limits, the PCT may have a rate that is significantly higher than the regional average. Within expected range – when the PCT lies within the 95% control limits then the PCT has a rate that is statistically consistent with the regional rate. Low – when the PCT lies between the lower 95% and lower 99.8% control limits, the PCT may have a rate that is significantly lower than the regional average. Very Low - when the PCT lies below the lower 99.8% control limit, then the PCT has a rate significantly lower than the regional rate

10.2 Results

10.2.1 All Malignant Cancers (excluding non-melanoma skin cancer) (ICD10 C00-C97 excl. C44)

Figures 10.1 and 10.2 show funnel plots comparing directly age standardised incidence and mortality rates for PCTs in the West Midlands for men diagnosed with / deaths from cancer in 2003-2005.

The incidence rate of all malignant cancers in men was significantly very high in South Birmingham PCT (459.7 per 100,000, exceeding the 99.8% control limit of the regional average (408.8 per 100,000).  Birmingham East & North PCT’s incidence rate (445.4 per 100,000) was also high, exceeding the 95% control limit.

The incidence rates of all malignant cancers in men were significantly very low for residents of Herefordshire and Warwickshire PCTs (340.6 and 374.8 per 100,000 respectively) in comparison to the West Midlands average.  The incidence rate in North Staffordshire PCT (369.0 per 100,000) was also low, falling below the 95% control limit.

The mortality rate in Stoke on Trent PCT (281.3 per 100,000), exceeded the 95% control limit and therefore was significantly higher than the regional average.  The mortality rate in this PCT was not consistent with the incidence rate, which was not statistically different from the West Midlands average.

A significantly very low mortality rate was found for all malignant cancers in men in Worcestershire PCT (196.8 per 100,000), below the lower 99.8% control limit of the regional average (222.2 per 100,000).  Four PCTs; Herefordshire (187.7 per 100,000), Shropshire County (201.9 per 100,000), Solihull (191.5 per 100,000) and Warwickshire (206.7 per 100,000) PCTs, had mortality rates below the lower 95% control limit of the West Midlands average.

For the other PCTs incidence and mortality rates were within the control limits, showing they were within the expected variation around the regional averages.

Figure 10.1: Incidence of malignant cancers (excluding non-melanoma skin cancer) in men (2003-2005)

Incidence of malignant cancers (excluding non-melanoma skin cancer) in men (2003-2005)

Figure 10.2: Mortality from malignant cancers (excluding non-melanoma skin cancer) in men (2003-2005)

Mortality from malignant cancers (excluding non-melanoma skin cancer) in men (2003-2005)

10.2.1b Females

Similarly, figures 10.3 and 10.4 show funnel plots comparing the incidence and mortality rates for PCTs in the West Midlands for women diagnosed with / deaths from cancer in 2003-2005.

In women, none of the incidence and mortality rates for all malignant cancers exceeded the 99.8% control limits of the regional averages. However, in some instances rates did exceed the 95% control limits. 

Stoke on Trent PCT had a significantly higher incidence rate (380.4 per 100,000) in comparison to the West Midlands average (344.0 per 100,000), whilst Heart of Birmingham Teaching PCT had a lower than expected incidence rate (303.9 per 100,000). 

The mortality rate in Shropshire County PCT (138.0 per 100,000) was significantly lower than the regional average (153.1 per 100,000).  The mortality rate in Stoke on Trent PCT (179.6 per 100,000), consistent with its incidence rate, was higher than the regional average, exceeding the upper 95% control limit.

Figure 10.3: Incidence of malignant cancers (excluding non-melanoma skin cancer) in women (2003-2005)

Incidence of malignant cancers (excluding non-melanoma skin cancer) in women (2003-2005)

Figure 10.4: Mortality from malignant cancers (excluding non-melanoma skin cancer) in women (2003-2005)

Mortality from malignant cancers (excluding non-melanoma skin cancer) in women (2003-2005)

10.2.2 Breast Cancer in Women (ICD10 C50)

Figures 10.5 and 10.6 show funnel plots comparing directly age standardised incidence and mortality rates for women diagnosed with / deaths from invasive breast cancer in 2003-2005.

Whilst Heart of Birmingham Teaching and Sandwell PCTs had lower incidence rates (93.7 and 104.7 per 100,000) at the 95% level than the West Midlands average (121.6 per 100,000), the mortality rates of these PCTs remained within the expected variation of the regional average.

None of the breast cancer mortality rates showed any statistical variation from the West Midlands average (28.5 per 100,000).

Figure 10.5: Incidence of invasive breast cancer in women (2003-2005)

Incidence of invasive breast cancer in women (2003-2005)

Figure 10.6: Mortality from invasive breast cancer in women (2003-2005)

Mortality from invasive breast cancer in women (2003-2005)

10.2.3 Colorectal Cancer (ICD10 C18-C20)

Figures 10.7 and 10.8 show funnel plots comparing directly age standardised incidence and mortality rates for PCTs in the West Midlands for men diagnosed with / deaths from colorectal cancer in 2003-2005.

For all PCTs both the incidence and mortality rates for colorectal cancer in men remained within the expected variation of the West Midlands averages (incidence: 58.3 per 100,000; mortality: 24.1 per 100,000).

Figure 10.7: Incidence of colorectal cancer in men (2003-2005)

Incidence of colorectal cancer in men (2003-2005)

Figure 10.8: Mortality from colorectal cancer in men (2003-2005)

Mortality from colorectal cancer in men (2003-2005)

10.2.3b Females

Corresponding figures, for females diagnosed with / deaths from colorectal cancer in 2003-2005, are shown in Figures 10.9 and 10.10.

Whilst the incidence rate of female colorectal cancer in Stoke on Trent PCT was not statistically different from the regional average, the mortality rate (21.4 per 100,000) was found to be higher than the regional average (13.7 per 100,000), exceeding the 95% control limit. 

Both Heart of Birmingham Teaching and Wolverhampton City PCTs had significantly lower incidence rates of colorectal cancer (23.2 and 25.2 per 100,000 respectively); in comparison to the West Midlands average (32.9 per 100,000).  However, the mortality rates for these PCTs were both within the variation expected of the regional average

All of the other PCTs were within the control limits for both incidence and mortality rates, indicating these were not statistically significantly different from the West Midlands averages.

Figure 10.9: Incidence of colorectal cancer in women (2003-2005)

Incidence of colorectal cancer in women (2003-2005)

Figure 10.10: Mortality from colorectal cancer in women (2003-2005)

Mortality from colorectal cancer in women (2003-2005)

10.2.4 Lung (ICD10 C33-C34)

10.2.4a Males

Figures 10.11 and 10.12 show funnel plots comparing directly age standardised incidence and mortality rates for PCTs in the West Midlands for men diagnosed with / deaths from lung cancer in 2003-2005.

The incidence rate of lung cancers in men in Stoke on Trent PCT (85.4 per 100,000) was significantly very high in comparison to the regional average (62.4 per 100,000), exceeding the 99.8% control limit.  Incidence in Sandwell PCT (78.493.7 per 100,000) was also high, exceeding the 95% control limit.  Consistent with these lung cancer incidence rates, mortality rates in men were found to be higher than the West Midlands average (54.7 per 100,000) in Sandwell PCT, with a rate of 69.0 per 100,000 and Stoke on Trent PCT with a rate of 73.8 per 100,000, both exceeding the upper 95% control limit.

Birmingham East & North PCT had a high incidence rate exceeding the 95% control limit, although the mortality rate of this PCT was within the expected variation of the regional average.

In Herefordshire, Warwickshire and Worcestershire PCTs, the lung cancer incidence rates (34.2, 48.3 and 50.4 per 100,000 respectively) in men were significantly lower than the West Midlands average.  Mortality rates in both Herefordshire and Worcestershire PCTs (33.7 and 43.5 per 100,000 respectively) were very low and Warwickshire PCTs rate (45.1 per 100,000) was also low.

Shropshire County and Solihull PCTs had incidence rates below the 95% control limit (48.3 and 49.6 per 100,000 respectively).  As might be expected from their incidence rates, mortality rates for both of these PCTs (43.1 and 40.9 per 100,000 respectively) were also significantly lower than the regional average.

Whilst the incidence rate for South Staffordshire PCT was not significantly different to the regional average, the mortality rate of 47.5 per 100,000 was below the 95% control limit of the West Midlands average.

Figure 10.11: Incidence of lung cancer in men (2003-2005)

Incidence of lung cancer in men (2003-2005)

Figure 10.12: Mortality from lung cancer in men (2003-2005)

Mortality from lung cancer in men (2003-2005)

10.2.4b Females

Figures 10.13 and 10.14 show funnel plots comparing directly age standardised incidence and mortality rates for PCTs in the West Midlands for women diagnosed with / deaths from lung cancer in 2003-2005.

The mortality rate in Birmingham East & North PCT (33.2 per 100,000) was significantly higher than the regional average (26.4 per 100,000), exceeding the 95% control limit.  This corresponded with the incidence rate observed in this PCT (39.9 per 100,000) also being above the 95% control limit of the West Midlands average (31.7 per 100,000). 

The incidence rates of 40.7 per 100,000 in South Birmingham PCT and 41.0 per 100,000 in Stoke on Trent PCT also exceeded the 95% control limit; however the mortality rates for these PCTs were not statistically significantly different from the regional average.

The incidence rate of lung cancer in women was significantly very low for residents of Herefordshire PCT (20.4 per 100,000).  In contrast, this PCT’s mortality rate (19.5 per 100,000) was within the expected range of the regional average. 

Shropshire County PCTs incidence rate (22.5 per 100,000) was below the lower 95% control limit, and the mortality rate (17.8 per 100,000) was significantly very low in comparison to the West Midlands average.  Dudley and Worcestershire PCTs both had low incidence (24.7 and 26.3 per 100,000 respectively) and mortality rates (19.1 and 20.6 per 100,000 respectively).

Figure 10.13: Incidence of lung cancer in women (2003-2005)

Incidence of lung cancer in women (2003-2005)

Figure 10.14: Mortality from lung cancer in women (2003-2005)

Mortality from lung cancer in women (2003-2005)

10.2.5 Prostate (ICD10 C61)

Figures 10.15 and 10.16 show incidence and mortality rates for PCTs in the West Midlands for men diagnosed with / deaths from prostate cancer in 2003-2005.

Although mortality rates for prostate cancer did not show any statistically significant variation from the West Midlands average (26.5 per 100,000), there was a wide variation observed for the incidence rates.

Prostate cancer incidence rates were found to be higher than the West Midlands average (100.8 per 100,000) in four PCTs; Birmingham East & North, Solihull, South Birmingham and Worcestershire with rates of 120.6, 125.2, 123.1 and 117.9 per 100,000 respectively, all of which exceeded the upper 95% control limit. 

Incidence rates were significantly very low in North Staffordshire and Stoke on Trent PCTs (70.1 and 68.5 per 100,000 respectively).  Two further PCTs were found to have low incidence rates, below the 95% control limit, namely Herefordshire and Sandwell PCTs (79.2 and 83.0 per 100,000 respectively).

Figure 10.15: Incidence of prostate cancer (2003-2005)

Incidence of prostate cancer (2003-2005)

Figure 10.16: Mortality from prostate cancer (2003-2005)

Mortality from prostate cancer (2003-2005)

10.3 Summary

During 2003-2005, a total of 76,229 malignant tumours (excluding non-melanoma skin cancer) were diagnosed in residents of the PCTs in the West Midlands, 38,817 in men and 37,412 in women.  In this same time period 21,572 men and 19,128 women died from malignant cancer (excluding non-melanoma skin cancer).  A summary of the variation in incidence and mortality rates as indicated by the funnel plots are shown in the Tables 10.1 and 10.2, for men and women respectively.

Table 10.1: Male Funnel Plot Summary Table (2003-2005)

Table 10.1: Male Funnel Plot Summary Table (2003-2005)

Table 10.2: Female Funnel Plot Summary Table (2003-2005)

Table 10.2: Female Funnel Plot Summary Table (2003-2005)

Overall for men, South Birmingham PCT had incidence rates which were statistically significantly higher than the West Midlands average, seemingly being driven by the high prostate cancer rate.  However, the male mortality rate of South Birmingham PCT did not reflect its very high incidence rate.  This is possibly due to South Birmingham being relatively affluent, thereby more likely to detect good prognosis prostate cancers via higher use of prostate specific antigen (PSA) testing.

Men in Stoke on Trent PCT had a higher mortality rate than the regional average, not indicated by this PCT’s overall incidence.  However, Stoke on Trent PCT was found to have significantly increased rates of lung cancer in men, the second most common cancer in men in the West Midlands3, with the incidence rate being very high in this PCT in comparison to the regional average; likewise the mortality for lung cancer was also high.  This effect is likely to be due to the higher levels of deprivation in Stoke, leading to higher incidence and mortality rates from lung cancer, related to higher smoking in the area.

For females, in 2003-2005, the incidence and mortality rates of all malignant cancers (excluding non-melanoma skin cancer) were not statistically significantly different, at the 99.8% level, from the West Midlands average in any of the PCTs in the region. However, Stoke-on-Trent PCT did fall into the area between the 95% and 99.8% control limits, indicating higher rates than the average for the region.  This is likely to be due to high incidence of lung cancer and high mortality of colorectal cancer in comparison to the regional averages.

As well as South Birmingham already being mentioned above, three other PCTs (Birmingham East & North, Solihull and Worcestershire) had a high incidence of prostate cancer, the most common cancer in men in the West Midlands3, compared with the regional average. None of them, however, had mortality rates outside the control limits. 

For men, areas with reduced risk may include Herefordshire and Warwickshire PCTs for incidence and Worcestershire PCT for mortality of all malignant cancers (excluding non-melanoma skin cancer), very significantly below the regional averages.  All of these areas are relatively affluent when compared with the rest of the West Midlands.  These reduced risks are consistent with these PCTs having significantly low / very low incidence and mortality of lung cancer in comparison to the regional average.  In contrast the deprived areas, Stoke on Trent and Sandwell PCTs had significantly very low and low incidence of prostate cancer, respectively.  High lung cancer and high prostate cancer rates in Birmingham East & North PCT seem counter-intuitive to explanation by deprivation levels.  However, this PCT was formed in 2006 from two previous PCTs, the deprived Eastern Birmingham PCT and the more affluent North Birmingham PCT.

For both men and women, Shropshire County PCT was found to have significantly low or very low incidence of and mortality from lung cancers; and similarly to males, Herefordshire PCT was found to have a significantly very low incidence rate of lung cancer. 

10.4 Conclusions

Most of these effects, seen in both males and females, are likely to be associated with known deprivation patterns.  However, it is possible that the mortality effects could reflect differing access to treatment or having different levels of co-morbidity, affecting the treatment options available; or both the incidence/mortality effects could reflect differing presentation patterns (leading to different severity of disease). These all need further examination and investigation across the region.

References:

1. Spiegelhalter, D. Funnel Plots for institutional comparison. Qual Saf Health Care 2002;11:390-391


2. Eastern Region Public Health Observatory. Template for producing funnel plots for directly age standardised rates (2005). http://www.erpho.org.uk/viewResource.aspx?id=12476. Last accessed 05/01/2007


3. West Midlands Cancer Intelligence Unit. The Top Ten Cancers in the West Midlands (2006). http://www.wmpho.org.uk/wmciu/PublicationDetail.asp?PublicationID=147. Last accessed 05/01/2007