Intra-abdominal cancer risk with abdominal pain: a prospective cohort primary care study

Background Quantifying cancer risk in primary care patients with abdominal pain informs diagnostic strategies. Aim To quantify oesophagogastric, colorectal, liver, pancreatic, ovarian, uterine, kidney, and bladder cancer risks associated with newly reported abdominal pain with or without other symptoms, signs, or abnormal blood tests (that is, features) indicative of possible cancer. Design and setting This was an observational prospective cohort study using Clinical Practice Research Datalink records with English cancer registry linkage. Method The authors studied 125 793 patients aged ≥40 years with newly reported abdominal pain in primary care between 1 January 2009 and 31 December 2013. The 1-year cumulative incidence of cancer, and the composite 1-year cumulative incidence of cancers with shared additional features, stratified by age and sex are reported. Results With abdominal pain, overall risk was greater in men and increased with age, reaching 3.4% (95% confidence interval [CI] = 3.0 to 3.7, predominantly colorectal cancer 1.9%, 95% CI = 1.6 to 2.1) in men ≥70 years, compared with their expected incidence of 0.88% (95% CI = 0.87 to 0.89). Additional features increased cancer risk; for example, for men, colorectal or pancreatic cancer risk with abdominal pain plus diarrhoea at 60–69 years of age was 3.1% (95% CI = 1.9 to 4.9) predominantly colorectal cancer (2.2%, 95% CI = 1.2 to 3.8). Conclusion Abdominal pain increases intra-abdominal cancer risk nearly fourfold in men aged ≥70 years, exceeding the 3% threshold warranting investigation. This threshold is surpassed for the >60 years age group only with additional features. These results will help direct appropriate referral and testing strategies for patients based on their demographic profile and reporting features. The authors suggest non-invasive strategies first, such as faecal immunochemical testing, with safety-netting in a shared decision-making framework.


Method
The authors studied 125 793 patients aged ≥40 years with newly reported abdominal pain in primary care between 1 January 2009 and 31 December 2013.The 1-year cumulative incidence of cancer, and the composite 1-year cumulative incidence of cancers with shared additional features, stratified by age and sex are reported.

Conclusion
Abdominal pain increases intra-abdominal cancer risk nearly fourfold in men aged ≥70 years, exceeding the 3% threshold warranting investigation.This threshold is surpassed for the >60 years age group only with additional features.These results will help direct appropriate referral and testing strategies for patients based on their demographic profile and reporting features.The authors suggest non-invasive strategies first, such as faecal immunochemical testing, with safetynetting in a shared decision-making framework.
whose individual cancer risks are >3%.The collective risk of a number of cancers may also be >3% in patients with abdominal pain, particularly when they have additional features common to those cancers.Knowing the hierarchy of risk within that group may help inform diagnostic strategies where clinicians consider that investigation is warranted based on overall cancer risk.
Finally, it would be useful to know the collective risk of cancers that share a common diagnostic pathway (for example, bladder and kidney).This study aimed to quantify the risk of cancers in the abdominal cavity or cancers that may present with abdominal pain (termed 'intra-abdominal') in patients aged ≥40 years in the year after newly reported abdominal pain, with or without other features of intraabdominal cancer.

METHOD Study design and setting
This prospective cohort study was set in English primary care, and used Clinical Practice Research Datalink (CPRD GOLD) data with partial linkage to data from England's cancer registry (National Cancer Registration and Analysis Service [NCRAS]).The CPRD contains prospectively collected and anonymised electronic medical records of patient demographics, symptoms, signs, tests, diagnoses, and treatments. 21Participants were followedup for intra-abdominal cancers in the year after abdominal pain newly reported during 2009-2013.

Sample selection criteria and study size
Study participants were aged ≥40 years, with an abdominal pain code (Supplementary Table S1) recorded between 1 January 2009 and 31 December 2013, but no such code in the previous year.Participants had continuous CPRD records meeting up-to-date standards from at least 1 year before their first abdominal pain code ('index date') to the end of the 1-year follow-up.Age was identified from the CPRD year of birth, assigning a birthday of 1 July.
Participants who were <40 years or with a cancer diagnostic code recorded before their first abdominal pain code were excluded.

How this fits in
Abdominal pain is a non-specific symptom, which may portend serious disease, including intra-abdominal cancers.There is no unified pathway for investigation.This paper reports the 1-year cumulative incidence risk of intra-abdominal cancer with or without concurrent clinical features for men and women aged 40-59, 60-69 and ≥70 years.Results show that patient demographics and type of concurrent feature effects the cancer risk.These results will inform appropriate testing strategies and specialist referral.
Code lists for each feature were collated. 26ausea and/or vomiting were combined because of overlapping codes.Participants with concurrent additional features were identified by searching the CPRD records in a 6-month window centred on the index date.

Outcomes
The outcomes were incident cancers, individually and collectively (composite outcomes consisting of any one of the cancer sites sharing a feature, see above).

Analyses
Analyses were stratified by age group and sex.The 1-year cumulative incidence of individual cancer sites in participants with or without an additional feature is reported.For context, the expected incidence of each cancer was estimated, based on 2011 data for cancer incidence and population size. 27,28or composite outcomes, the 1-year cumulative incidence for participants with abdominal pain plus an additional feature are reported.Estimates are reported with binomial exact 95% confidence intervals (CIs).Data analysis was conducted using Stata (version 16).

Missing data and bias
All code lists are available on request.Code absence was interpreted as non-occurrence of the clinical event. 21Confounding by sex and age were controlled by stratified analyses.

Numbers of participants with additional features
Additional features were more common in women and with increasing age.At least one additional feature occurred in 12.9%, 19.2%, and 34.6% of men, and in 21.0%, 25.7%, and 39.7% of women aged 40-59, 60-69, and ≥70 years, respectively (see Table 2 for the numbers of participants with additional features, by age group and sex, and by composite cancer grouping), (see Supplementary Table S2 for breakdown by cancer site).

Cancer incidence with abdominal pain
In this sample with abdominal pain, the 1-year cumulative incidence of intra-abdominal cancer was higher for men than women, and increased with age (Supplementary Table S3), reaching 3.4% (95% CI = 3.0 to 3.7) and 2.3% (95% CI = 2.1 to 2.5) for men and women, respectively, aged ≥70 years.For all age groups, participants were at greatest risk of colorectal cancer, followed by pancreatic and oesophagogastric cancers (and ovarian cancer for women).
Bladder, kidney, and liver cancers had the lowest incidence estimates.The 1-year cumulative incidence values in participants with abdominal pain were consistently higher than the population estimates, apart from liver cancer in women aged 40-59 years (Supplementary Table S3).

Composite cancer risk in participants with additional features
Additional features increased cancer risk over that with abdominal pain alone.At 40-59 years (Supplementary Table S4), having an abdominal mass increased colorectal, ovarian, oesophagogastric, or liver cancer risk in women to 7% (95% CI = 2.0 to 17).Weight loss increased colorectal, ovarian, pancreatic, or oesophagogastric cancer risk to 4% (95% CI = 1 to 9) in women and to 4% (95% CI = 2 to 9) in men.Constipation increased colorectal or pancreatic cancer risk to 2.6% (95% CI = 1.5 to 4.2) in men.
Additional weight loss increased colorectal, ovarian, pancreatic, or oesophagogastric cancer risk to >3% in both sexes (Tables 3 and 4), with pancreas the most likely site at 60-69 years (Table 3).Additional nausea and/or vomiting increased pancreatic or oesophagogastric cancer risk >3% in men aged 60-69 and ≥70 years, with pancreatic more likely.
Additional constipation or diarrhoea increased colorectal or pancreatic cancer risk >3% in men, with the more likely sites being colorectal at ≥70 years, but pancreatic cancer for constipation and colorectal cancer for diarrhoea at 60-69 years.Additional change in bowel habit increased colorectal or ovarian cancer risk in women >3% at 60-69 and ≥70 years, with colorectal more likely than ovarian (Table 3).

Haematuria, urinary tract infection and abnormal blood test results
Bladder/kidney cancer and uterine/ oesophagogastric cancer risks were similarly low in participants with abdominal pain alone, or plus urinary tract infection and abnormal blood tests, respectively (Supplementary Table S5).Bladder or kidney cancer risk in women with abdominal pain and haematuria was 3.0% (95% CI = 0.6 to 8.5) at age 40-59 years and 8% (95% CI = 4 to 14) in those ≥70 years, with bladder the more likely site.

Summary
This study examined a common diagnostic problem -abdominal pain -and quantified intra-abdominal cancer risk in the subsequent year.Overall, the risk with abdominal pain per se was lowest in women aged 40-59 (0.49%), and highest in men aged ≥70 (3.39%) (see Supplementary Table S3).The higher risk in men likely reflects sex differences in colorectal, oesophagogastric, pancreatic, bladder, and kidney cancer incidence.Abdominal pain increased intra-abdominal cancer risk over that in the general population.For example, men aged ≥70 years have a general risk of 0.88% (95% CI = 0.87% to 0.89%) compared with 3.39% with newly reported abdominal pain (see Supplementary Table S3).Having additional features increased cancer risk further, more so with age.
This study has identified which cancers were more likely within the cancers that share features.For example, a man aged 60-69 years with abdominal pain and weight loss has a cumulative intra-abdominal cancer risk of 9%, made up of pancreatic (6%), colorectal, and oesophagogastric (each approximately 2%) cancers.

Strengths and limitations
22,24 The healthcare setting is important, as most patients with abdominal pain present to primary care.Robust methods were used to identify cancer features and diagnoses. 26PRD cancer recording is >90%, and was supplemented by cancer registry linkage. 29ome symptom data will be missing: patients may not mention abdominal pain or other cancer features, and doctors may not record them, or only record them in text. 30Text-only abdominal pain records may have reduced the pool of possible patients; nevertheless, the current study was sufficiently powered. 30tratified analyses ensured that the results would not be skewed by the varying incidences of abdominal pain and cancers by age and sex.The decision in this study to seek additional cancer features within a 6-month window centred on the index date was pragmatic.The authors acknowledge omission of cancer features recorded outside this timeframe.

Comparison with existing literature
Most existing analyses were not stratified, complicating direct comparisons with the results of the current study.Holtedahl et al followed-up 6264 adults attending European primary care with abdominal symptoms. 31he PPVs of upper and lower abdominal pain, respectively, for any abdominal cancer were 1.5% (95% CI = 1.0 to 2.1) and 1.0% (95% CI = 0.6 to 1.5).Their collective value of 2.5% is of similar magnitude to the estimates in the current study for the ≥70 years age group.Additional constipation, diarrhoea, or weight loss increased the hazard of new abdominal cancer, 31 consistent with the findings of the current study.Of the 511 cancers diagnosed, 94 were in the colon or rectum (mean age 71 years). 31ower abdominal pain was a common prediagnostic symptom, with a PPV of 0.7% (95% CI = 0.4 to 1.1). 31This is lower than the estimated colorectal cancer risk in the ≥70 years age group in the current study, possibly because of differences in age distribution and abdominal pain location. 32For each feature, the total risk and that of contributing cancers is reported.Estimates are reported to the precision afforded by the standard errors, which varies with the cancer-feature combination.b Cancer risk ≥1% and <2%.
c Cancer risk ≥2% and <3%.d Cancer risk ≥3%.e There are no data for men for change in bowel habit, as the composite outcome is not meaningful for male sex, who are not diagnosed with ovarian cancer.CI = confidence interval.n/a = not applicable.
Herbert et al 20 followed-up adults aged ≥30 years with abdominal symptoms in primary care for 1 year.Similar to the findings in the current study, colorectal cancer was the most common intra-abdominal cancer diagnosed, with PPVs of 0.6% and 0.3% for men and women, respectively.Hippisley-Cox and Coupland 33,34 followedup adults aged 25-89 years attending primary care for 2 years, reporting the PPVs of abdominal pain for any incident cancer of 4.0% for men and 2.8% for women.Discrepancies with estimates in the current study probably relate to the restriction of diagnoses to intra-abdominal cancers, differences in age/sex profile, and follow-up period.The increased risks with additional features in the current study are similar to those reported elsewhere.For example, nausea and/or vomiting has been found to increase the PPVs of abdominal pain from 0.3% for both pancreatic and oesophagogastric cancers to 2.2% (95% CI = 1.1 to 4.6) for pancreatic cancer and 0.7% (95% CI = 0.5 to 0.9) for oesophagogastric cancer. 22,24plications for clinical practice It is important to remember that non-malignant abdominal pain causes were not sought.For many of the clinical profiles studied, clinicians may be able to diagnose a non-malignant disease, without considering cancer in the differential.This, supplemented by treatment response, means that patient groups with profiles suggesting a cancer risk ≥3% may be categorised further: a lower-risk group not requiring initial cancer investigation, and a (much) higher-risk group warranting cancer investigation.This selection process is too subtle for observational studies to elucidate entirely.However, managing older patients with cancer features requires assessing the risks and benefits of possibly invasive investigations, such as colonoscopy, and the results of this study aid that.Older patients wish to be involved in decision making, but this is difficult for individuals who are cognitively impaired or frail in standard settings. 35Decisions to investigate are more likely to be deferred in older patients (that is ≥65 years), who tend to have longer diagnostic intervals than younger patients. 35his suggests a real risk of harm to older patients from diagnostic delays, reinforcing the need for rigorous safety-netting in this patient group. 36ot all patients will need referral: where colorectal cancer is the likeliest, faecal immunochemical testing may be used before invasive colonoscopy.In some healthcare systems, primary care clinicians may order computed tomography for possible pancreatic cancer and intra-abdominal lymphomas.Even so, some patients with negative primary care testing will still harbour cancer, and may need specialist referral perhaps to an RDC.This selection is more than totting up estimated risks; intuition and experience may play a part, 37 and in the UK, NICE supports GPs using these to make referrals.The results of the current study may guide clinicians in RDCs as to the optimum investigation, or sequence of investigations, to facilitate early diagnosis.
In conclusion, abdominal pain may indicate cancer, the chance being higher in men and with increasing age.Additional features in the history may indicate specific cancers, allowing targeted investigation.This is relevant to primary care, and to facilities for investigating non-specific features of possible cancer.Abdominal pain alone increases baseline cancer risk nearly fourfold in men aged ≥70 years, to over the threshold warranting investigation.The authors suggest starting with non-invasive testing strategies, such as faecal immunochemical testing, with robust safety-netting in a shared decision-making framework.
The 3% threshold is surpassed for participants >60 years of age only when additional features are present.These results help direct appropriate referral and testing strategies for patients based on their demographic profile and the features they report.

Table 1 . Participant characteristics
a Additional feature = the signs, symptoms or abnormal test results, in addition to abdominal pain, that are listed in NG12 as possible presenting features of intra-abdominal cancers.Figure 1.

Flow chart of individuals included in the study with application of exclusion criteria. CPRD = Clinical Practice Research Datalink.
e363 British Journal of General Practice, May 2022

Table 2 . Numbers of participants with additional features, by age group and sex, and by feature-cancer combination
n/a = not applicable: included because people born of male sex cannot be diagnosed with uterine cancer or with ovarian cancer.British Journal of General Practice, May 2022 e364

Table 4 . The 1-year cumulative incidence of cancers in participants aged ≥70 years with abdominal pain plus another feature of possible cancer, stratified by sex a
a For each feature, the total risk and that of individual cancers is reported.b Cancer risk ≥1% and <2%.c Cancer risk ≥2% and <3%.d Cancer risk ≥3%.CI = confidence interval.n/a = not applicable.

:
This study was supported by Department of Health and Social Care Grant 'Unexplained abdominal pain in primary care: establishing a clinical profile and diagnostic pathway for identifying cancer', under the Cancer Awareness, Screening and Early Diagnosis Policy Research Unit.Sarah Price is funded by the National Institute for Health Research (NIHR) Policy Research Programme, conducted through the Policy Research Unit in Cancer Awareness, Screening and Early Diagnosis, PR-PRU-1217-21601.Elizabeth Shephard is funded through the ERICA trial (https://www.theericatrial.co.uk/), which is funded by the Dennis and Mirelle Gillings Foundation, and receives support from CRUK and Macmillan charities.This current research is also linked to the CanTest Collaborative (of which Willie Hamilton is Co-Director), which is funded by CRUK (grant reference number: C8640/ A23385).The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.