Socio-demographic variation in stage at diagnosis and its population impact: How many cancer patients would be diagnosed at an earlier stage if socio-demographic inequalities were eliminated?


Session type:

Matthew Barclay1,Gary Abel2,Brian Rous3,David Greenberg3,Georgios Lyratzopoulos4
1University of Cambridge,2University of Exeter,3Public Health England,4University College London



Stage at diagnosis is an important predictor of cancer outcomes. Understanding related socio-demographic inequalities could guide interventions for earlier diagnosis. 


We analysed data on English residents aged 30-99 diagnosed in 2015 with any of 10 common cancers (melanoma, breast, endometrial, bladder, prostate, renal, colon, rectal, ovarian, and lung cancer). Stage information was available on 183,406/202,001 (91%) tumours. Multiple imputation was used to account for missing stage in main analysis. We investigated socio-demographic differences in stage at diagnosis (parameterised as ‘advanced’ (TNM stages 3-4) and ‘non-advanced’ (1-2)) using logistic regression models, initially including (and adjusting for) cancer site, and subsequently stratifying by cancer site. Modelling strategies took into account tumour type and screening detection status, in addition to age, sex, deprivation status. We estimated site-specific inequalities and the number of advanced stage cases attributable to age and deprivation inequalities.


43% of all cases were diagnosed in stages III/IV. Advanced stage at diagnosis was more likely in men, older, more deprived, and non-white patients, with marked variations in these inequalities by cancer site. Men had higher adjusted odds of advanced stage for melanoma, lung and renal cancer, and women for bladder cancer. The probability of advanced stage diagnosis generally increased in those aged 65+ across sites, particularly for prostate and endometrial cancer and least so for lung cancer. Screening detection influenced age-related patterns for breast, colon and rectal cancer. Deprivation gradients were relatively small for all sites except lung cancer. About 6,500 tumours (3.2% of this cohort) would be diagnosed at non-advanced stage if older age and deprivation inequalities were eradicated.


Substantial socio-demographic inequalities in stage at diagnosis exist for most cancers; these may be amenable to public health interventions targeted according to population impact. However, removing variation in stage at diagnosis is not a sufficient strategy for eliminating known survival inequalities.