1

Title: Delays in referral from primary care are associated with a worse survival in patients with oesophagogastric cancer

Authors: Chanpreet S Arhi clinical research fellow1 MRCS BSc, Sheraz Markar clinical research fellow1

MRCS MSc MA, Elaine M Burns clinical lecturer1 FRCS PhD, George Bouras consultant surgeon1 FRCS PhD, Alex Bottle reader in medical statisitcs2 PhD, George Hanna professor of surgery1 FRCS PhD, Paul Aylin professor in epidemiology and public health2 MBChB FFPH FRCPE, Paul Ziprin consultant surgeon1 FRCS, Ara Darzi professor of surgery1 OM FRS FRCS FREng FMEdSci

Addresses: 1. Imperial College London, Department of Surgery and Cancer, St Mary’s Hospital Campus, Praed Street, W2 1NY 2. Imperial College London, School of Public Health, 3 Dorset Rise, EC4Y 8EN

Corresponding author: Chanpreet Arhi, c.arhi@imperial.ac.uk. Address as above. 07949762763

Acknowledgements: The Sowerby Foundation who kindly provided funding for data acquisition from the Clinical Practice Research Datalink, protocol 13_078.

Running title: Delays in oesophagogastric cancer diagnosis

1

2

Abstract

Introduction: NICE referral guidelines for suspected cancer were introduced to improve prognosis by

reducing referral delays. However, over 20% of patients with oesophagogastric cancer experience

three or more consultations before referral. In this retrospective cohort study, we hypothesise that

such a delay is associated with a worse survival compared with patients referred earlier.

Method: By utilising CPRD, a national primary care linked database, the first presentation, referral

date, number of consultations before referral and stage for oesophagogastric cancer patients was

determined. The risk of a referral after one or two consultations compared with three or more

consultations was calculated for age and the presence of symptom fulfilling the NICE criteria. The risk

of death according to the number of consultations before referral was determined, while accounting

for stage and surgical management.

Results: 1307 patients were included. Patients referred after one (HR 0.80 95% CI 0.68 – 0.93 p =

0.005) or two consultations (HR 0.81 95% CI 0.67 – 0.98 p = 0.034) demonstrated significantly

improved prognosis compared with those referred later. The risk of death was also lower for patients

who underwent a resection, were younger or had an earlier stage at diagnosis. Those presenting with

a symptom fulfilling the NICE criteria (OR 0.27 95% CI 0.21 to 0.35 p <0.0001) were more likely to be

referred earlier.

Conclusion: This is the first study to demonstrate an association between a delay in referral and worse

prognosis in oesophagogastric patients. These findings should prompt further research to reduce

primary care delays.

Key Words: Referral delays, primary care, oesophagogastric cancer, survival

Abstract count: 249Article word count: 3501

2

3

Main article

Introduction:

The EUROCARE-4 study data demonstrated over 600 deaths in oesophagogastric cancer patients per

year could have been avoided if cancer survival in Great Britain was similar to the European mean (1).

Delays in diagnosis have been suggested as a reason for this difference. Although the NICE (the

National Institute for Health and Care Excellence) criteria in 2005, and the update in 2015 (2), for

urgent referral have reduced the referral interval from primary care(3), further studies are required to

determine if they have improved prognosis.

The number of consultations before referral can be used as a marker of delay within the primary care

setting. In the National Cancer Patient Experience Survey (2010)(4), 24.9% of oesophageal and 36.0%

of stomach cancer patients experienced three or more consultations before referral. Such results

reflect the complexity of cancer diagnosis. Previous studies have not investigated an association

between the number of consultations and prognosis.

Our hypothesis states three or more consultations before referral to secondary care is associated with

a delay in diagnosis and therefore a worse survival for patients with oesophagogastric cancer. By using

retrospective data from primary care, linked to hospital inpatient episodes, cancer stage and mortality

data, we consider potential confounders to survival such as the symptom at first presentation, stage at

diagnosis and surgical resection. A secondary objective of this study is to identify patient

characteristics that are associated with delays in referral.

Method

Data source

The Clinical Practice Research Datalink (CPRD) was used as the source of patients for this study. This

database covers about 4% of the UK population after linkage on an individual patient basis to Hospital

3

4

Episodes Statistics (HES), cause of death from Office for National Statistics (ONS) and the cancer

registry from the National Cancer Intelligence Network (NCIN)(5). The ‘Read’ system is used to record

symptoms, investigations and referrals. These codes are translated into numerical ‘medcodes’, and are

provided to researchers, together with the date of the consultation.

Patient selection

Patients were included if they had an ICD10 code for an oesophageal (C15) or gastric (C16) cancer in

the NCIN cancer dataset between January 2000 and December 2010, were 18 years of age or older,

had at least one consultation with a relevant symptom (see table 1) recorded in CPRD in the year

leading up to the NCIN cancer diagnosis date and an entry indicating a referral to secondary care

(Appendix D).

Patient and tumour variables

The date of diagnosis was taken as the date recorded in the NCIN dataset. As we considered surgery as

a covariate (see appendix A for OPCS codes), patients were excluded if their tumour was of squamous

cell morphology (see appendix B for morphology codes) as resection is not commonly the primary

treatment for such cancers. Stage data was split into I, II, II and IV with a fifth group representing

patients with missing stage data.

Patents who were diagnosed as an emergency (defined as an emergency admission with an

oesophagogastric cancer diagnosis in the six months leading up to the NCIN date of diagnosis) were

excluded as a proportion would have had no primary care involvement before diagnosis. Charlson

score was calculated from three years of linked HES data preceding the cancer registry date of

diagnosis.

Relevant consultations

4

5

The first presentation was the first recorded symptom, as listed in table 1, in the year leading up to the

diagnosis date. The symptoms were separated into red flag and non-specific symptoms based on the

NICE guidance that corresponds to the study period(2) (Appendix C). These symptoms relate to

referral guidelines which were introduced to aid the primary care physician in the UK to identify

patients with possible cancer. These are a set of symptom and haemoglobin based-criteria, which, if

met, should stimulate referral to secondary care requesting an urgent consultation within two weeks

of receipt of the referral. It was not possible to ascertain the duration of symptoms before

presentation from CPRD. A combination of symptoms was not considered necessary for inclusion.

A referral after one consultation was described as a referral date within two days of the first

presentation. For the remaining patients the number of relevant consultations (i.e with symptoms as

set out in table 1) was identified between the first presentation and the referral date. Consultations

with duplicate dates were counted once. Only the first referral after the first presentation was

considered.

Statistical Analysis

The Chi square test was used to identify significant differences between patients referred after one,

two or three or more consultations in terms of age, gender, smoking history, year of diagnosis,

Charlson score, GP location, presence of a red flag symptom and surgical management. A sensitivity

analysis was carried out to determine differences between patients with and without a referral

recorded.

The number of consultations before referral was considered the dependent variable in an ordinal

logistic regression model.

5

6

Survival was calculated from the date of diagnosis to the date of death noted in the ONS dataset.

Patients without an ONS entry were considered censored, with their survival calculated up to the end

of ONS coverage (1st January 2012).

A Cox proportional hazard model described the hazards ratio of death for referral after one or two

consultations compared with referral after three or more consultations while adjusting for the factors

mentioned above. An interaction between time and stage was included in the survival model to

ensure the proportional hazards assumption was not violated.

Significance level was taken at 0.05%. Robust standard errors were calculated for multivariate analysis

to account for heteroscedasticity within the data. All analysis was carried out in SPSS (IBM, v22).

Results

3210 patients were initially identified as having an elective oesophagogastric adenocarcinoma

diagnosis between 2000 and 2010 inclusive. 2021 (62.9%) had a recording of a relevant symptom in

CPRD in the year before diagnosis as per table 1. Of these 1307 (64.6%) patients had a referral

recorded before diagnosis and so were included in further analysis. Sensitivity analysis demonstrated

patients diagnosed before 2006 were more likely to have a missing symptom or referral, as were

patients aged less than 55.

Of these 1307 patients, 672 (51.4%) had an oesophageal and 635 (48.6%) had a stomach cancer

diagnosis. Patients referred after one consultation had a median of zero days to referral (interquartile

range 0 to 1 days), after two consultations had a median of 15 days (interquartile range 7 to 37) and

those referred after three or more consultations had a median of 57 days (interquartile range 27 to

163 days).

6

7

There was no significant difference in the number of consultations before referral in terms of age,

gender, Charlson score, smoking history, tumour stage or year of diagnosis (table 2). Of those patients

initially presenting with a red-flag symptom 47.1% (n = 433) were referred at first consultation, with

19% (n = 175) and 33.8% (n = 311) referred after the two or three or more consultations respectively.

Patients initially presenting with non-specific symptoms experienced significantly more consultations

before referral, with 17.0% (n = 66) referred after the first consultation, compared with 62.4% (n =

242) referred after 3 or more consultations.

Delays in referral

Male patients were more likely to be referred after fewer consultations (OR 0.66, 95% CI 0.46 to 0.96,

p = 0.031), as were patients who initially presented with a red flag symptom (OR 0.27, 95% CI 0.21 to

0.35, p < 0.0001) (table 3).

Impact on mortality

Patients who were referred after one (0.81, 95% CI 0.68 to 0.93, p = 0.005) or two consultations (0.81,

95% CI 0.67 to 0.98, p = 0.034) had significantly reduced hazards of death compared with patients who

were referred after three or more consultations (Table 4). Patients who presented with a red flag

symptom at first consultation did not have a significant difference in the hazard of death compared

with those that presented with a non-specific symptom. Compared with tumour stage I, a higher stage

at diagnosis increased the hazards ratio of death from 1.41 for stage II to 2.32 for stage IV.

At 5 years, patients who were referred after one or two consultations demonstrated a survival benefit

of over 30% compared with those referred after 3 or more consultations (figure 1).

Discussion

Summary of findings

7

8

This study demonstrates an association between referral after three or more consultations and

decreased prognosis for patients with oesophagogastric cancer. Male patients and those presenting

with a red-flag symptom were less likely to experience a delay in referral. However, presenting with a

red-flag symptom did not significantly alter survival.

Comparison with previous studies

Previous studies have concentrated on describing the cause and duration of delays in primary care,

without correlating this with prognosis(6,7). Two systematic reviews identified initial misdiagnosis and

inappropriate tests as factors that could lead to a delay in referral (8,9). By comparing survival of

oesophagogastric cancer patients diagnosed via the two-week-wait pathway or routinely, Sharpe et al.

(10) found those referred via the latter had a better prognosis. Although they described the interval

from referral to diagnosis, the interval from first presentation to referral was not provided. Shawihdi

et al.(11) demonstrated oesophagogastric patients managed in GP practices with a higher rate of

referral for gastroscopy had a better prognosis, suggesting a lower threshold for investigations

identifies patients earlier in the progression of their disease. However, stage data and the interval

between first presentation and referral was not included in this study. As our study aimed to

determine the impact of referral delays by GPs, we excluded emergency cases as some of these

patients would have bypassed primary care due to their acute and severe symptoms.

Our study demonstrated the same median interval from first presentation to referral for patients

referred after one consultation as the National Audit of Cancer Diagnosis in Primary care(12). For

patients referred after two consultations, the audit described a median of 20 days for oesophageal

and 21 days for stomach, with interquartile ranges of 8 - 37 and 9 - 42 respectively. The ranges were

similar to our study, with only a five-day difference in medians. The median interval for three or more

consultations was not provided in the audit.

8

9

In our study 38.4% of stomach and 48.3% of oesophageal patients were referred after three or more

consultations, compared with 32% of stomach cancer patients and 22.6% of oesophageal cancer

patients in this same audit (12). The difference is likely due to alternative methods of data collection.

The national audit requested the number of consultations that were related to the cancer diagnosis,

but did not specify the criteria of what should be considered relevant. GPs may not have considered

the list of vague symptoms as appropriate as they do not fulfil the NICE criteria. Our list of Read codes

is similar to other studies using CPRD(3). In addition, the national audits can be influenced by recall

bias.

Relevance to health policy

Reducing patient, practitioner and hospital delays have all been highlighted as potential areas of

improvement to narrow the gap in survival compared with the better performing European

countries(13,14). The International Cancer Benchmarking Partnership(15) and the National Awareness

and Early Diagnosis Initiative(14) are aiming to identify the cause of these delays and what effect they

have on survival. Our study provides a rationale to this important area of focused quality healthcare

improvement.

Although research into the impact of a delayed referral is lacking, there have been investigation into

associations between the diagnostic interval and survival(16). However, due to the heterogeneity in

the definition and the methods used to obtain data, there is no consensus that a delay in diagnosis

leads to a worse prognosis for either oesophageal or gastric cancer(16). For example Windham et

al(17). defined the start of this interval as the onset of symptoms, while Dregan et al (6) defined it as

the first presentation to primary care.

9

10

Our data demonstrates patients who were referred after three or more consultations experienced a

median of 57 and 42 days longer than those referred after the first or second consultation

respectively. During this time, it is conceivable that the cancer could have progressed to a later stage

of cancer, which would explain the difference in outcome. However our data failed to show a

significant difference in stage, likely due to the level of missing data (51.4% overall). it was not possible

to determine the proportion of different stages amalgamated within the missing group. Of note

patients who experienced a delay of three or more consultations had the lowest proportion of

resections, which may partly explain our findings. In an alternative analysis (data not shown), the

interval from presentation to referral was considered in days and split according to the median

interval for the number of consultations. Patients referred within 15 days had a better outcome than

those referred later. An advantage of presenting the delay as number of consultations is that the

findings can be related to clinical practice and published national audits.

Due to the low frequency of individual symptoms, there was inadequate power in this study to

demonstrate an association between a specific symptom and prognosis. However, we did not find a

significant association between red flag symptoms as a group and worse survival. Dregan et al (18)

found no association between the presence of dysphagia and survival. This is an area that requires

further study, as it may help stratify prognosis at the time the patient first presents to their GP.

Reasons for a delay

There are a number of reasons why patients may experience a delayed referral. It is difficult to

separate by symptomology the oesophgaogastric cancer patients from the majority of patients with an

underlying benign condition. Our study demonstrates 29.7% (n = 388) presented with symptoms

outside the NICE criteria. The GP may have adopted a ‘watch and wait’ policy for these patients, or

may have ordered investigations such as an ultrasound scan for a presumed benign condition(8).

10

11

52.9% (n = 486) of patients who initially presented with a red-flag symptom waited at least two

consultations perhaps they did not initially fulfil the referral criteria.

Strengths of our study

A strength of this study is that it includes covariates that are known to influence survival - stage at

diagnosis and surgical management. As would be expected in clinical practice our model has shown

cancer resection, earlier stage at the time of diagnosis and younger age are associated with a better

prognosis. The data included have been routinely collected by GPs, and therefore are not influenced

by recall bias. At the time of the consultation the GP would be unaware of the eventual diagnosis.

Although CPRD contains data for only 8 – 10% of the UK population, it has been shown to be

representative of the population as a whole(19)

Limitations

The findings of this study should be taken into context of the limitations of CPRD. As demonstrated in

the process of patient selection, only 40.8% of patients with an OG cancer diagnosis were included in

analysis due to missing data, which is less common in the latter years covered by the database.

Symptoms reported by the patient may have been recorded in the free text rather than coded, which

is no longer available to researchers (20). As such patients presenting with non-specific symptoms are

likely to be under-reported in our analysis, which may explain the higher rate of missing data for those

aged under 55.

With CPRD it is not possible to determine the duration or severity of symptoms before presentation,

as the READ coding does not contain sufficient detail. This interval could potentially have altered both

the urgency of referral and by extension, the stage at diagnosis. Although this extra information may

have influenced our findings, two systematic reviews(8)(9) into factors that determine patient

11

12

presentation concluded that type of symptom and co-morbidities had the greatest sway, while age

and gender had little or no impact. In this study these possible confounders were included.

A higher proportion of the cancer patients underwent a resection, with the national upper

gastrointestinal cancer audit suggesting the figure should be closer to 30%(21). This reflects our

inclusion criteria of only adenocarcinomas. The presentation pattern is not expected to be different

for patients with squamous cell cancer. Although over 50% of patients had missing stage data the

odds of survival decreased with increasing stage, which provides credence to the quality of data

collected.

Conclusion

This study demonstrates an association between the number of consultations and prognosis for

oesophagogastric cancer patients, and should encourage further research into primary care delays and

the effect on survival. Detecting cancer in primary care is a difficult task, and relying only on the type

of symptom may be delaying the referral for specialist input.

12

13

The authors have no conflicting interests to declare.

Author Contribution

Chanpreet Arhi: Conception, design, acquisition, analysis, drafting, approval of final version and accountable for study

Sheraz Markar: Conception, design, analysis, drafting, approval of final version and accountable for study

Elaine Burns: Conception, design, analysis, drafting, approval of final version and accountable for study

George Bouras: Design, analysis, drafting, approval of final version and accountable for study

Alex Bottle: Design, analysis, drafting, approval of final version and accountable for study

George Hanna: Design, analysis, drafting, approval of final version and accountable for study

Paul Aylin: Design, analysis, drafting, approval of final version and accountable for study

Paul Ziprin: Design, analysis, drafting, approval of final version and accountable for study

Ara Darzi: Design, analysis, drafting, approval of final version and accountable for study

13

14

References

1. Abdel-Rahman M, Stockton D, Rachet B, Hakulinen T, Coleman MP. What if cancer survival in Britain were the same as in Europe: how many deaths are avoidable? Br J Cancer [Internet]. 2009 Dec 3 [cited 2015 Jun 3];101 Suppl:S115-24. Available from: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2790713&tool=pmcentrez&rendertype=abstract

2. National Initiative for Health and Care Excellence. NICE guidlelines for suspected cancer. 2015;(June). Available from: https://www.nice.org.uk/guidance/ng12/

3. Neal RD, Din NU, Hamilton W, Ukoumunne OC, Carter B, Stapley S, et al. Comparison of cancer diagnostic intervals before and after implementation of NICE guidelines: analysis of data from the UK General Practice Research Database. Br J Cancer [Internet]. England: Nature Publishing Group; 2014 Feb 4;110(3):584–92. Available from: http://www.ncbi.nlm.nih.gov/pubmed/24366304

4. Lyratzopoulos G, Neal RD, Barbiere JM, Rubin GP, Abel GA. Variation in number of general practitioner consultations before hospital referral for cancer: findings from the 2010 National Cancer Patient Experience Survey in England. Lancet Oncol [Internet]. 2012 Apr [cited 2016 Aug 15];13(4):353–65. Available from: http://www.ncbi.nlm.nih.gov/pubmed/22365494

5. Herrett E, Thomas SL, Schoonen WM, Smeeth L, Hall AJ. Validation and validity of diagnoses in the General Practice Research Database: A systematic review. Br J Clin Pharmacol. 2010;69:4–14.

6. Dregan A, Møller H, Charlton J, Gulliford MC. Are alarm symptoms predictive of cancer survival?: population-based cohort study. Br J Gen Pract [Internet]. Royal College of General Practitioners; 2013 Dec [cited 2016 Aug 15];63(617):e807-12. Available from: http://www.ncbi.nlm.nih.gov/pubmed/24351496

7. Stapley S, Peters TJ, Neal RD, Rose PW, Walter FM, Hamilton W. The risk of oesophago-gastric cancer in symptomatic patients in primary care : a large case – control study using electronic records. Br J Cancer [Internet]. Nature Publishing Group; 2012;108(1):25–31. Available from: http://dx.doi.org/10.1038/bjc.2012.551

8. Macleod U, Mitchell ED, Burgess C, Macdonald S, Ramirez a J. Risk factors for delayed presentation and referral of symptomatic cancer: evidence for common cancers. Br J Cancer

14

15

[Internet]. Nature Publishing Group; 2009;101 Suppl(S2):S92–101. Available from: http://dx.doi.org/10.1038/sj.bjc.6605398

9. Macdonald S, Macleod U, Campbell NC, Weller D, Mitchell E. Systematic review of factors influencing patient and practitioner delay in diagnosis of upper gastrointestinal cancer. Br J Cancer. 2006;94(April):1272–80.

10. Sharpe D, Williams RN, Ubhi SS, Sutton CD, Bowrey DJ. The &quot;two-week wait&quot; referral pathway allows prompt treatment but does not improve outcome for patients with oesophago-gastric cancer. Eur J Surg Oncol [Internet]. 2010 Oct [cited 2016 Aug 24];36(10):977–81. Available from: http://www.ncbi.nlm.nih.gov/pubmed/20702059

11. Shawihdi M, Thompson E, Kapoor N, Powell G, Sturgess RP, Stern N, et al. Variation in gastroscopy rate in English general practice and outcome for oesophagogastric cancer: retrospective analysis of Hospital Episode Statistics. Gut [Internet]. England; 2014 Feb;63(2):250–61. Available from: http://www.ncbi.nlm.nih.gov/pubmed/23426895

12. Lyratzopoulos G, Saunders CL, Abel GA, McPhail S, Neal RD, Wardle J, et al. The relative length of the patient and the primary care interval in patients with 28 common and rarer cancers. Br J Cancer [Internet]. Nature Publishing Group; 2015;112 Suppl(s1):S35-40. Available from: http://www.nature.com/bjc/journal/v112/n1s/full/bjc201540a.html#content%5Cnhttp://dx.doi.org/10.1038/bjc.2015.40

13. Foot C, Harrison T. How to improve cancer survival Explaining England’s relatively poor rates. Kings Fund. 2011;(June):1–32.

14. Hiom SC. Diagnosing cancer earlier: reviewing the evidence for improving cancer survival. Br J Cancer [Internet]. Nature Publishing Group; 2015;(March):1–5. Available from: http://www.nature.com/doifinder/10.1038/bjc.2015.23

15. Rose PW, Hamilton W, Aldersey K, Barisic A, Dawes M, Foot C, et al. Development of a survey instrument to investigate the primary care factors related to differences in cancer diagnosis between international jurisdictions. BMC Fam Pract. England; 2014;15:122.

16. Neal RD, Tharmanathan P, France B, Din NU, Cotton S, Fallon-Ferguson J, et al. Is increased time to diagnosis and treatment in symptomatic cancer associated with poorer outcomes? Systematic review. Br J Cancer [Internet]. 2015 Mar 3;112(March):S92–107. Available from: http://www.nature.com/doifinder/10.1038/bjc.2015.48

17. Windham TC, Termuhlen PM, Ajani JA, Mansfield PF. Adenocarcinoma of the stomach in patients age 35 years and younger: No impact of early diagnosis on survival outcome. J Surg Oncol. 2002;81(3):118–24.

18. Dregan A, Møller H, Charlton J, Gulliford MC. Are alarm symptoms predictive of cancer survival? Br J Gen Pract [Internet]. England; 2013 Dec 1;63(617):807–12. Available from: http://bjgp.org/cgi/doi/10.3399/bjgp13X675197

19. Boggon R, van Staa TP, Chapman M, Gallagher AM, Hammad TA, Richards MA. Cancer recording and mortality in the General Practice Research Database and linked cancer registries. Pharmacoepidemiol Drug Saf [Internet]. 2013 Feb [cited 2015 Apr 22];22(2):168–75. Available from: http://www.ncbi.nlm.nih.gov/pubmed/23239282

20. Herrett E, Shah AD, Boggon R, Denaxas S, Smeeth L, van Staa T, et al. Completeness and diagnostic validity of recording acute myocardial infarction events in primary care, hospital care, disease registry, and national mortality records: cohort study. BMJ [Internet]. England; 2013 May 20;346:f2350. Available from: http://www.ncbi.nlm.nih.gov/pubmed/23692896

21. National Oesophago-gastric Cancer Audit | AUGIS [Internet]. Available from: http://www.augis.org/national-oesophago-gastric-cancer-audit/

22. Price S, Stapley E, Shepherd E, Barraclough E, William H. Is omission of free text records a possible source of data loss and bias in Clinical Practice Research Datalink studies? A case–control study BMJ Open 2016;6:e011664

15

16

Table 1 - Symptoms of interest. Read codes are provided in Appendix C

Red flag Non-specific

Dyspepsia over the age of 55 Dyspepsia under age of 55

Dysphagia Anorexia

Vomiting Fatigue

Anaemia Reflux/gastritis

Weight loss Other abdominal pain

Upper abdominal pain

16

17

Abdominal mass

GI bleeding

Table 2 – Summary statistics of patients referred after the one, two or three or more presentations

Number of consultations before referral

All One Two Three or more

pN = 1307 (100%) n = 499 (38.2%) N = 255 (19.5%) n = 553 (42.3%)

n % n % n % n %Site Oesophagus 672 51.4 287 42.7 135 20.1 250 37.2 < 0.005

Stomach 635 48.6 212 33.4 120 18.9 303 47.8Age Less than 55 123 9.4 40 32.5 24 19.5 59 48.0 0.520

55 to 64 288 22.0 119 41.3 53 18.4 116 40.365 to 75 427 32.7 152 35.6 89 20.8 186 43.6Over 75 469 35.9 188 40.1 89 19.0 192 40.9Median 71 71 71

Gender Male 942 72.1 373 39.6 170 18.0 399 42.4 0.065Female 365 27.9 126 34.5 85 23.3 154 42.2

Charlson score

Zero 1083 82.9 414 38.2 214 19.8 455 42.0 0.845One or more 224 17.1 85 37.9 41 18.3 98 43.8

Diagnosis year

2000 to 2005 599 45.8 224 37.4 121 20.2 254 42.4 0.799

2006 to 2010 708 54.2 275 38.8 134 18.9 299 42.2

Smoker Yes 1103 84.4 426 38.6 217 19.7 460 41.7 0.585No 204 15.6 73 35.8 38 18.6 93 45.6

Location Yorkshire & North East 120 9.2 51 42.5 26 21.7 43 35.8 0.023

North West 251 19.2 117 46.6 36 14.3 98 39.0

17

18

Midlands 229 17.5 85 37.1 39 17.0 105 45.9East of England 171 13.1 61 35.7 38 22.2 72 42.1

South West 278 21.3 87 31.3 67 24.1 124 44.6London & South East 268 20.5 98 36.6 59 22.0 111 41.4

First presenting symptom

Red flag 919 70.3 433 47.1 175 19.0 311 33.8 < 0.005

Vague 388 29.7 66 17.0 80 20.6 242 62.4Stage I 161 12.3 54 33.5 34 21.1 73 45.3 0.832

II 161 12.3 61 37.9 36 22.4 64 39.8

III 205 15.7 74 36.1 39 19.0 92 44.9

IV 108 8.3 41 38.0 19 17.6 48 44.4

Unknown 672 51.4 269 40.0 127 18.9 276 41.1Resection Yes 586 44.8 223 44.7 126 49.4 237 42.9 0.219

No 721 55.2 276 55.3 129 50.6 316 58.1

Table 3 – Ordinal logistic regression of referral after one, two or 3 or more consultations, including Interaction of gender and surgical resection, gender and first presenting symptom, smoking history

and Charlson score, surgical resection and Charlson score

Odds ratio p valueAge Over 75 ref

65 to 75 1.16 (0.90 to 1.51) 0.23155 to 64 0.96 (0.72 to 1.30) 0.806Under 55 0.96 (0.64 to 1.44) 0.849

SmokerYes refNo 0.80 (0.59 to 1.07) 0.120

GenderFemale refMale 0.66 (0.46 to 0.96) 0.031

Charlson score

One or more refZero 0.91 (0.69 to 1.21) 0.513

First presenting symptom

Non-specific ref

Red flag 0.27 (0.21 to 0.35) <0.0001

LocationLondon & South East ref

Yorkshire & North East 0.75 (0.50 to 1.14) 0.151

18

19

North West 0.73 (0.53 to 1.02) 0.080

Midlands 1.13 (0.80 to 1.59) 0.485East of England 1.07 (0.74 to 1.56) 0.724

South West 1.22 (0.89 to 1.69) 0.221

Surgical resection

Yes refNo 0.69 (0.45 to 1.04) 0.076

Year of diagnosis

2000 to 2005 1.09 (0.88 to 1.35) 0.4282006 to 2010 ref ref

Table 4 – Adjusted hazards ratio of death for oesophagogastric cancer patients, including an interaction of stage and surgical management

Hazard ratio pNumber of consultations before referral

One 0.80 (0.68 to 0.93) 0.005

Two 0.81 (0.67 to 0.98) 0.034

3 or more ref

Age

Under 55 0.71 (0.56 to 0.96) 0.026

55 to 64 0.68 (0.57 to 0.82) <0.000165 to 75 0.74 (0.62 to 0.86) <0.0001Over 75 ref

SmokerNo refYes 0.82 (0.66 to 1.02) 0.069

GenderFemale refMale 1.12 (0.95 to 1.32) 0.168

Charlson score

Zero refOne or more 0.83 (0.67 to 1.03) 0.065

First presenting symptom

Vague ref

Red flag 1.01 (0.85 to 1.19) 0.936

LocationLondon & South East ref

Yorkshire & North East 0.92 (0.67 to 1.27) 0.620

19

20

North West 1.11 (0.89 to 1.39) 0.349

Midlands 1.18 (0.93 to 1.50) 0.174

East of England 0.92 (0.71 to 1.19) 0.576

South West 1.06 (0.86 to 1.31) 0.463

OutcomeNo resection refResection 0.31 (0.27 to 0.37) < 0.0001

Stage

I ref

II 1.41 (0.84 to 2.37) 0.193

III 1.73 (1.13 to 2.65) 0.013

IV 2.32 (1.53 to 3.51) < 0.0001

Unknown 1.72 (1.18 to 2.50) 0.005

Year of diagnosis

2006 to 2010 ref2000 to 2005 1.32 (1.15 to 1.51) < 0.0001

Figure 1 - Long term survival stratified according to the number of consultations before referral

20

21

Overall survival (months)0 12 24 36 48 60 72 84 96 108 120

Number of consultations before referral

All patients 1307 746 434 303 226 166 123 90 59 37 23One 499 301 171 120 88 64 47 35 21 12 8Two 255 157 94 66 49 39 31 25 16 13 7Three or more 553 288 169 117 89 63 45 30 22 12 8

21