Skip to main content

Advertisement

Table 1 Summary of quality assessment and study design for the 18 included papers

From: Risk prediction models for colorectal cancer in people with symptoms: a systematic review

Author, date, country, setting Qualityf Outcome Data collection Selection of variables Identification of study population Identification of outcome cases Exclusions Study population
Model development; Case-control studies        
   Cases Controls    Cases Controls
ᅟHamilton, 2005, UK, primary care [22] M CRC Primary care records from 21 practices Occurring in at least 2.5 % of cases or controls >40 years with primary CRC 5 controls per case matched for sex, general practice and age and alive at point of case diagnosis Cancer registry at one hospital Unobtainable records, no consultations in 2 years before diagnosis, previous CRC, residence outside Exeter at time of diagnosis. 349 1,744
ᅟHamilton, 2009, UK, primary care [23] M CRC THIN database Literature review > 30 years with CRC Up to 7 controls without CRC matched for practice, sex and age Diagnosis of CRC within study period < 2 years of full electronic records before date of case diagnosis. 5,477 38,314
Model development and external validation; Case-control study        
   Cases Controls    Cases Controls
ᅟMarshall, 2011, UK, primary care [24] H CRC BB equation development and CAPER Score external validation
    See Hamilton, 2009 [23] As in Hamilton, 2009 plus patients with severe anaemia (Hb < 10 g/dl), rectal bleeding, abnormal rectal examination or positive FOBT, or without any of abdominal pain, weight loss, diarrhoea or constipation 117 433
   CRC CAPER Score development and BB equation external validation
    See Hamilton, 2005 [22]
Model development and random split-sample internal validation; Cohort studies       
      Included Cases (% of included)
ᅟHippisley-Cox, 2012, UK, primary care ᅟ(QCancer® (colon)) [25] H CRC QResearch database 'Established predictor variables' and red flag symptoms 30–84 year-old patients registered with practices between 1/1/2000 and 30/09/2010 and without CRC Incident cancer diagnosis in the 2 years after cohort entry recorded in GP records or ONS cause-of-death record History of CRC, recorded red flag symptomf in the 12 months preceding the study date, or missing Townsend deprivation score. Development
F: 1,172,670 F:4,798 (0.2 %)
M:1,178,382 M:4,798 (0.2 %)
Internal validation
F: 616,361 F: 2603 (0.2 %)
M: 620,240 M:2603 (0.2 %)
ᅟHippisley-Cox, 2013 (female), UK, primary care ᅟ(QCancer® (combined)) [26] H CRC and 11 other cancersa QResearch database Previous study, and literature review 25–89 year-old patients registered with practices between 1/1/2000 and 1/04/2012 and without CRC Incident cancer diagnosis in the 2 years after cohort entry recorded in GP records or ONS cause-of-death record Recorded red flag symptomf in the 12 months before the study entry date, or missing Townsend deprivation score. Development
1,240,864 2607 (0.18 %)
Internal validation
679,174 1725 (0.25 %)
ᅟHippisley-Cox, 2013 (male), UK, primary care ᅟ(QCancer® (combined)) [27] H CRC and 9 other cancersb QResearch database Previous study, and literature review 25–89 year-old patients registered with practices between 1/1/2000 and 1/04/2012 and without CRC Incident cancer diagnosis in the 2 years after cohort entry recorded in GP records or ONS cause-of-death record Recorded red flag symptomf in the 12 months before the study entry date, or missing Townsend deprivation score. Development
1,263,071 3250 (0.26 %)
Internal validation
667,603 1356 (0.2 %)
Model development; Cross-sectional studies       
      Included Cases (% of included)
ᅟAdelstein, 2010, Australia, secondary care [32] H CRC Self-administered questionnaire Not reported Patients > 18 years old scheduled for colonoscopy at hospitals Complete colonoscopy and histology Completion of questionnaire > 6 months before colonoscopy, advanced adenomac, incomplete colon evaluation 7,736 159 (2.1 %)
ᅟAdelstein, 2011, Australia, secondary care [31] H CRC See Adelstein 2010 [32] Completion of questionnaire > 6 months before colonoscopy, adenomad, incomplete colon evaluation 6943 159 (2.3 %)
ᅟFijten, 1995, Netherlands, primary care [21] L CRC Patient and doctor questionnaires, and blood sample Literature review Patients presenting to 83 GP practices with overt rectal bleeding or a history of visible rectal blood loss in previous 3 months. Medical record review coded using the International Classification of Primary Care for diagnostic classification Patients aged <18 or >75, pregnancy, urgent admission to hospital or follow-up not available. 290 9 (3.4 %)
ᅟHurst, 2007, UK, secondary care [28] M CRC or pre-malignant adenomas Proforma-based history, examination and blood sample Not reported All adult patients referred to a specialist colorectal clinic Patients tracked until a definitive diagnosis was reached Patients not further investigated after initial consultation or who did not attend follow up 300 95 (31.7 %)
ᅟLam, 2002, Hong Kong, secondary care [20] L CRC or significant neoplasiae Questionnaire conducted by non-medically trained interviewers Not reported New patients attending surgical department for rectal bleeding Rigid sigmoidoscopy and proctoscopy, followed by barium enema or colonoscopy at the physician's discretion Refusal for colonoscopy or barium enema 174 29 (16.7 %)
ᅟMahadavan, 2011, UK, secondary care [29] M CRC Self-administered questionnaire, history, faecal, blood and rectal samples Not reported All patients >40 years referred to a surgical clinic via the 2wwg system for colorectal cancer Incident diagnosis of CRC within 6 months of study entry from primary care or hospital records confirmed histologically Previous confirmed IBD, GI cancer, investigation of the bowel within the last 6 months or absent rectal sampling result 714 72 (10.1 %)
Model development and external validation; Cross-sectional study       
      Included Cases (% of included)
ᅟSelvachandran, 2002, UK, secondary care (WNS) [30] h H CRC Self-administered questionnaire Not reported Patients referred by GPs with symptoms suggestive of distal colonic or anorectal disease Not reported (all patient's received endoscopy) Not reported 2,268 95 (4.2 %)
Model external validation; Cohort study       
   Model(s) validated        Included Cases (% of included)
ᅟCollins, 2012, UK, primary care [33] H QCancer® (colon) (female and male) [25] THIN database N/A 30–84 year-old patients registered with practices between 1/1/2000 and 30/09/2010 and without CRC Incident cancer diagnosis of CRC in the 2 years after cohort entry Patients with a history of CRC, a recorded red flag symptomf in the 12 months preceding the study date, registered <12 months with practice or with invalid dates Female: 1,075,775 Female:1,676 (0.15 %)
          Male: 1,059,765 Male: 2,036 (0.19 %)
Model external validation; Cross-sectional studies       
   Model(s) Validated        Included (% of eligible) Cases (% of included)
ᅟBallal, 2009, UK, secondary care [35] H WNS [30] Self-administered questionnaire N/A Patients with colorectal symptoms referred by GPs A combination of rigid sigmoidoscopy, flexible sigmoidoscopy, colonoscopy or barium enema Patients thought (on the basis of the referral) most likely to have right-sided CRC, or but did not attend for investigation 3,457 186 (5.4 %)
ᅟHodder, 2005, UK, secondary care [34] H WNS [30], Fijten 1995 [21] Self-administered questionnaire N/A Patients referred from primary care with colorectal symptoms Secondary care investigations - minimum flexible sigmoidoscopy Not reported 3,302 156 (4.7 %)
ᅟRai, 2008, UK, secondary care [11] H WNS [30] Self-administered questionnaire N/A GP referral with any of: lower bowel-related symptoms, unexplained iron deficiency anaemia, positive FOBT, or palpable rectal/abdominal mass Follow up during course of hospital investigations until a final diagnosis made Patients admitted hospital as an emergency and subsequently diagnosed with CRC 1,422 83 (5.84 %)
Model utility; cohort study       
   Model used      Outcome measures   Included Interviews
ᅟHamilton, 2013, UK, primary care [19] L Hamilton 2005 [22] GP usage and outcomes from practices and local trusts; qualitative interviews Not reported Risk assessment tools (RATs) supplied to 614 GPs at 164 practices for 6 months; interviews with GP cancer network leads and sample of GP users from practices with differing patient demographics. Number of 2WW referrals and colonoscopies for patients >40; symptoms used in RATs; qualitative interview data. RATs performed on patients <40; RATs that did not identify the reported symptoms. 1433 23 GP responders
  1. CRC colorectal cancer, ONS office of national statistics, FOBT faecal occult blood test, IBD Inflammatory bowel disease, Hb haemoglobin, WNS Weighted Numerical Score developed by Selvachandran 2002 [30], RAT risk assessment tool
  2. aLung, gastro-oesophageal, pancreatic, renal tract, haematological, breast, ovarian, uterine, cervical and other cancer
  3. bLung, gastro-oesophageal, pancreatic, renal tract, haematological, prostate, testicular, and other cancer
  4. cAdenoma with significant (> 25 %) villous features, or high grade dysplasia, including carcinoma-in-situ, or size 10 mm or larger
  5. dAdenoma of any size or histology
  6. ePolyp 10 mm or larger, or a polyp of any size with a villous histology
  7. frectal bleeding, weight loss, abdominal pain, loss of appetite
  8. g2WW - Two week wait
  9. hThe method of developing the WNS is copyrighted and incompletely reported