Primary care characteristics and stage of cancer at diagnosis using data from the national cancer registration service, quality outcomes framework and general practice information

There were 363,991 tumours diagnosed in 2012 (all cancers excluding non-melanoma skin
cancers, ICD-10 C00 to C97 excluding C44). Of these there were 42,572 female breast
cancers, 36,822 prostate cancers, 34,458 colorectal cancer and 38,652 lung cancers,
accounting for 42 % of all cancers diagnosed in 2012. From these 34,119 female breast
cancers (5,666 stage 3 or 4, 16.6 %), 27,880 prostate cancers (10,756 stage 3 or 4,
38.6 %), 27,079 colorectal cancers (14,793 stage 3 or 4, 54.6 %) and 28,479 lung cancers
(21,520 stage 3 or 4, 75.6 %) were included in the analyses (see Fig. 1 for details of inclusion/exclusion of tumours). These were from patients at 7,786
GP practices across England.

(For details of the number of tumours of each cancer type by patient and GP variable
see the Additional file 1: Table Sb).

At an individual level we found that various exposures could be important confounders
for presenting with advanced female breast cancer (see Table 1). Non-white vs. white women and women living in more deprived areas were more likely
to be diagnosed at a more advanced stage (RD 6.0 % (95 % CI 3.3 % to 8.6 %) p??0.001;
Q5 vs. Q1 RD 3.9 % (95 % CI 2.5 % to 5.3 %), p-value for trend 0.001).

Table 1. Univariate and adjusted risk differences for female breast cancer and prostate cancer

Women aged 15–44 years were more likely to be diagnosed at a more advanced stage than
women aged 65 years and over whereas women aged 45–64 years were less likely to be
diagnosed at a more advanced stage (15-44years vs. 65+ RD 2.1 % (95 % CI 0.6 % to
3.6 %) p?=?0.01; 45–64 years vs. 65+ RD ?3.2 % (95 % CI ?4.1 % to ?2.4 %) p??0.001).

A variety of GP exposures were associated with stage at presentation but after adjustment
for age and deprivation the following predicted lower proportion with advanced stage
female breast cancer: having a GP in a town/fringe area compared to urban area (RD
?1.5 % (95 % CI ?2.5 % to ?0.4 %) p?=?0.01), ), practices with higher two week wait
(TWW) referral rate and a higher TWW detection rate (Q5 vs. Q1 RD ?1.5 % (95 % CI
?2.8 % to ?0.2 %) p value for trend?=?0.01; Q5 vs. Q1 RD ?1.3 % (95 % CI ?2.6 % to
0.0 %) p value for trend?=?0.01) and practices that had a higher emergency admission
rate (Q5 vs. Q1 RD ?2.0 % (95 % CI ?3.3 % to ?0.8 %) p value for trend?=?0.03). In
contrast having only female general practitioners at the practice and being at a practice
where people thought it was less easy to book an appointment was associated with a
higher percentage diagnosed at a more advanced stage (all female GPs: RD 4.0 % (95 %
CI 0.6 % to 7.4 %) p?=?0.02; 80 % thought easy to book appointment compared to 90 %
RD 1.7 % (95 % CI 0.1 % to 3.3 %) p?=?0.04.

At the individual level we found that various exposures could be important confounders
for presenting with advanced prostate cancer (see table 1). Men living in more deprived areas were more likely to be diagnosed at a more advanced
stage than those living in less deprived areas (Q5 vs. Q1 RD 4.7 % (95 % CI 2.7 %
to 6.8 %), p-value for trend 0.001). Non-white vs. white men and younger men were
less likely to be diagnosed at a more advanced stage (RD ?6.0 % (95 % CI ?10.3 % to
?1.7 %) p?=?0.01; 45-64 years vs. 65+ RD ?8.1 % (95 % CI ?9.4 % to ?6.8 %) p??0.001,
15-44 years vs. 65+ RD ?19.0 % (95 % CI ?29.5 % to ?8.5 %) p??0.001).

After adjustment for age and deprivation GP practice deprivation and practices with
higher rates of colonoscopy, sigmoidoscopy and endoscopy were associated with a higher
percentage diagnosed at a more advanced stage (Q5 vs. Q1 RD 1.8 % (95 % CI ?0.6 %
to 4.2 %) p-value for trend 0.04; tertile 3 vs. tertile 1 RD 2.4 % (95 % CI 0.9 %
to 3.9 %) p value for trend?=?0.002).

For colorectal cancer, at the individual level, we found that various exposures could
be important confounders for presenting later (see Table 2). Non-white vs. white people and younger people were more likely to be diagnosed
at a more advanced stage (RD 6.7 % (95 % CI 2.7 % to 10.7 %) p?=?0.001; 15-44 years
vs. 65+ RD 10.3 % (95 % CI 7.1 % to 13.4 %) p??0.001, 45-64 years vs. 65+ RD 6.0 %
(95 % CI 4.6 % to 7.3 %) p??0.001). After adjustment for age, sex and deprivation
the only GP exposure which was associated with stage at presentation was the average
colonoscopy, sigmoidoscopy and endoscopy rate. We found that a higher average colonoscopy,
sigmoidoscopy and endoscopy rate was associated with a lower percentage of people
diagnosed at a more advanced stage (tertile 3 vs. tertile 1 RD ?2.0 % (95%CI ?3.5 %
to ?0.5 %) p value for trend?=?0.01).

Table 2. Univariate and adjusted risk differences for colorectal cancer and lung cancer

Age and gender were important confounders for presenting with advanced lung cancer
(see Table 2). Women were less likely to be diagnosed at a more advanced stage than men (RD ?3.3 %
(95 % CI-4.3 % to ?2.3 %) p??0.001). People aged 45–64 years were more likely to
be diagnosed at a more advanced stage than people aged 65 and over (RD 3.3 % (95 %
CI 1.9 % to 4.6 %) p??0.001) but there was no difference between people aged 15–44
years and people 65 and over (RD 4.5 % (95 % CI ?1.2 % to 10.2 %) p?=?0.12).

After adjustment for age, sex and deprivation, being at a practice with a higher TWW
referral rate, having no GPs aged 50 and over and having all female GPs was associated
with a lower percentage diagnosed with more advanced stage lung cancer (Q5 vs. Q1
RD-3.3 % (95 % CI ?4.9 % to ?1.7 %) p-value for trend 0.001; none vs. some RD-2.5 %
(95%CI ?4.3 % to ?0.7 %) p?=?0.01; all vs some. RD-4.6 % (95%CI ?8.4 % to ?0.7 %)
p?=?0.02). In contrast being at a practice which had more patients per GP, being at
a practice with a higher TWW conversion rate and being at a practice that had a higher
emergency admission rate was associated with a higher percentage diagnosed at a more
advanced stage (Q5 vs. Q1 RD 1.8 % (95 % CI0.2 % to 3.4 %), p-value for trend 0.01;
Q5 vs. Q1 RD 4.0 % (2.4 % to 5.5 %) p-value for trend 0.001; Q5 vs. Q1 RD 1.6 % (95%CI
0.0 % to 3.2 %) p-value fpr trend 0.04). There is a weak negative correlation between
TWW referral and TWW conversion and this may explain some of the association between
higher TWW conversion and more advanced stage at diagnosis.

Missing stage data and multiple imputation

There was no systematic pattern of missing stage data between patient age and sex
across the four common cancers. For female breast, prostate and lung cancer people
who were more deprived were less likely to have missing stage data. Comparison of
risk difference with and without the use of stage imputation shows very small alterations
to risk differences which did not alter trends or interpretation for exposure variables.

Sensitivity analysis

For cancers with stage data ethnicity was missing for 36.1 % of patients with female
breast cancer, 47.9 % of prostate cancer, 33.1 % of colorectal cancer and 30.7 % of
lung cancer. To assess the impact of adjusting for ethnicity, results for patients
with complete ethnicity data adjusted for age, sex, deprivation and ethnicity were
compared to an analysis excluding ethnicity. There were only very small changes in
risk differences between these analyses with no change to the trends or conclusions
drawn from the results. This is probably due to the distribution of ethnicity with
96 % of those with staged female breast, colorectal, lung and prostate cancer being
white.

The main analysis used all relevant stage data from NCRS (see Additional file 1: appendix 1 for description of collection of stage data). If only the data from the
NCRS ‘Stage best’ field was used 32,590 (81.0 %) of female breast cancers had staging
data, 26,847 (78.4 %) of prostate cancers, 25,362 (80.7 %) of colorectal cancer and
27,134 (82.2 %) of lung cancers. Analysis to assess the impact of using all relevant
stage data compared to using the ‘Stage Best’ field alone showed very small changes
to the risk differences for female breast, colorectal and lung cancer. There was no
change to the trends or conclusions of the results. For prostate cancer there were
some slightly greater changes to the risk differences.

Due to the large proportion of lung cancers diagnosed at stage 3 or 4 we conducted
an analysis to compare stage 4 with stage 1, 2 or 3. The trends for number of patients
per GP, TWW referral rate and TWW conversion rate did not alter. However the relationship
between GP demographics (age and gender) and emergency admissions were attenduated.