Diagnostic accuracy systematic review of rectal bleeding in combination with other symptoms, signs and tests in relation to colorectal cancer Olde Bekkink M, McCowan C, Falk G, Teljeur C, Van de Laar F, Fahey T. Marleen Olde Bekkink Erasmus student, clinical research fellow Dept of General Practice Radboud University Nijmegen Medical Centre, the Netherlands 8th July 2009
2 Background 2 *1. Fijten GH et al *Mant A. et al Incidence of CRC in patients with rectal bleeding General population: < 1 per people General practice: 2-11 per patients Secondary care: 36 per referred patients 1 st selection by patient 2 nd selection by GP Olde Bekkink M, McCowan C, Falk G, Teljeur C, Van de Laar F, Fahey T.
Methods - Study protocol Inclusion criteria Population * Patients recruited from primary care population * Rectal bleeding and associated symptoms Study design * Observational cohort studies Reference standard and follow up *Colonoscopy *Flexible/ rigid sigmoidoscopy (and barium enema) *Questionnaire/ follow up only Outcome measures * Data must allow construction of 2x2 tables 3 Olde Bekkink M, McCowan C, Falk G, Teljeur C, Van de Laar F, Fahey T.
4 Methods – Inclusion of articles Olde Bekkink M, McCowan C, Falk G, Teljeur C, Van de Laar F, Fahey T. Potentially relevant citations (n=1534) Excluded after scanning of citation or abstract (n=1470) Full text retrieved (n=64) Excluded (n=56) Reason for exclusion: Setting not in general practice (n = 23) Screening study (n = 5) Case control study (n = 5) Data not extractable from 2x2 table (n = 15) Other (n=8) Included studies (n=8)
55 Methods – Characteristics of the included studies No of patients included: 2323 Prior CRC Weighted average: 7.0 % Median : 8.1 % Eight included studies ( ) Olde Bekkink M, McCowan C, Falk G, Teljeur C, Van de Laar F, Fahey T. Included articles: GP setting NoStudy IDPrior 1du Troit % (15/265) 2Ellis % (11/319) 3Fijten % (9/269) 4Heintze % (23/476) 5Mant % (16/145) 6Metcalf % (8/99) 7Norrelund % (32/208) 14.1% (22/156) 8Wauters % (27/386)
66 Methods – Quality assessment Olde Bekkink M, McCowan C, Falk G, Teljeur C, Van de Laar F, Fahey T.
77 Methods- Data extraction Extraction in 2x2 tables to calculate LR+ Metcalf 1996 prior 8/99 Olde Bekkink M, McCowan C, Falk G, Teljeur C, Van de Laar F, Fahey T. Sign/ symptom/ characteristicsNo of patients% populationNo cancers Dark red blood loss3131%3 Weight loss1515%2 Blood mixed with stool4646%5
8 Results- Pooled positive likelihood ratios 8 Olde Bekkink M, McCowan C, Falk G, Teljeur C, Van de Laar F, Fahey T. Patient characteristicsNo of studies No of patientsPooled PLR (95% CI) Male ( ) Age <40y ( ) Age ≥ 60y ( ) Family history colorectal cancer ( ) Signs/ symptoms No of studies No of patientsPooled PLR (95% CI) Dark red blood ( ) Weight loss ( ) Abdominal pain ( ) Blood mixed with the stool ( ) Changed bowel habit ( ) Perianal symptoms- pain on defecation ( ) Diagnostic testsNo of studies No of patients Pooled PLR (95% CI) Anaemia (Hb ♀ < g/dL ♂ <13.3 g/dL) ( ) Rectal palpation-haemorrhoid ( )
99 Discussion Limitations of sub studies - Reference standard ≠ gold standard Limitations of the review - Internal validity Reference categories - External validity Incidence of colorectal cancer Edited image ( MO ) Original: Mayo Foundation for Medical Education and Research Olde Bekkink M, McCowan C, Falk G, Teljeur C, Van de Laar F, Fahey T.
10 Conclusion Value of the review Implications for clinical practice guidelines Predictive value of rectal bleeding low-moderate Alarm symptoms yield only moderate likelihood ratios Further studies required calculating independent values using a multivariable analysis Olde Bekkink M, McCowan C, Falk G, Teljeur C, Van de Laar F, Fahey T.
11 Questions Thank you!