Skip to main content

Table 1 Summary of mortality analyses of randomised controlled cancer-screening trials

From: UKCTOCS update: applying insights of delayed effects in cancer screening trials to the long-term follow-up mortality analysis

Trial name Disease area Country Number of participants Recruitment period Number of screens Screening period Censorship date Median FU from randomisation Original analysis LTFU analysis No of years from randomisation to mortality reduction*
Statistical analysis methodology Final mortality reduction (95%CI) Statistical analysis methodology (if different) Final mortality reduction (95%CI)
Two county Breast Sweden 162,981 1977 4 1977–1984 end 1984 5.93 years (mean) (29 years? LTFU) “Mantel-Haenszel” techniques - stratified by county and age RR = 0.69 (no CI reported; p = 0.013) Negative binomial regression, robust SEs for cluster randomisation RR = 0.69 (95% CI 0.56–0.84; p = 0.0001) ~ 4 years (Figure 1) [14]
Malmo Breast Sweden 42,283 1976–1978 5 1976–1986 end 1987 8.8 years (mean) Relative risk (RR), test based CI RR = 1.29 (95% CI 0.74–2.25)    No screening effect (no figure in analysis time) [32]
Gothenburg Breast Sweden 51,611 1982–1984 4–5 1982–1991 end 1996 11.8 years (mean) (~ 14 years LTFU) RR, Poisson regression. Test based on Likelihood ratio RR = 0.56 (95% CI 0.31–0.99; p = 0.046) RR, Poisson regression adjusted for birth cohort RR = 0.79 (95% CI 0.58–1.08; p = 0.14) ~ 0 years (Figure 1) [5]
Edinburgh Breast UK 54,654 1978–1985 2–4 (depending on cohort) 1978–1988 1992 ~ 9 years? 10 years max (12.8 years (mean) LTFU) Logistic regression modified for cluster randomisation and stratified by age. ITT RR = 0.82 (95% CI 0.61–1.11) [RR with logistic regression?] Same RR = 0.87 (95% CI 0.70–1.06) ~ 6 years (Figure 2) [33]
UK Age Trial Breast UK 160,921 1991–1997 7 1991–2004? end 2004 10.7 years (17.7 years LTFU) RRs, no detail. ITT RR = 0.83 (95% CI 0.66–1.04; p = 0.11) Poisson regression (presumably as before). RR = 0.88 (95% CI 0.74–1.04; no p-value) ~ 3 years (Figure 2) [34]
ERSPC Prostate Europe (7 countries) 162,387 (in the core age group) 1991–2003 up to 3? 1991–2003 end 2006 9.0 years (13 years LTFU) Poisson regression to estimate mortality ratio (RR), stratified by centre and age group. ITT RR = 0.80 (95% CI
0.65–0.98; p = 0.04).
Same RR = 0·79 (95% CI 0·69- 0·91; p = 0·001) ~ 7 years (Figure 2) [31]
SCORE Colorectal Italy 34,292 1995–1999 1 1995–1999? 2006? 11.4 years RRs based on average mortality rates (Poisson distribution). ITT RR = 0.78 (95% CI 0.56 to 1.08)    ~ 5–6 years (Figure 2c) [35]
NORCCAP Colorectal Norway 98,792 1999–2001 1 1999–2001 end 2011 10.9 years (14.2 years LTFU (mean)) HRs from Cox model, adjusted for age. ITT HR = 0.73 (95% CI, 0.56–0.94; p = 0.02) Same, except primary analysis now has separate estimates for men and women Men HR = 0.63 (95% CI 0.47-0.83) Women HR = 1.01 (95% CI 0.77-1.33) ~ 5–9 years (~ 3 years for men) (Figure 2c) [21]
PLCO Prostate USA 76,693 1993–2001 4–6 1993–2005? 2008 11.5 years (14.8 years LTFU) RRs assuming Poisson distribution. ITT. No mention of WLR test and no p value given subsequently RR = 1.13 (95% CI 0.75 to 1.70) Same RR = 1.04 (95% CI 0.87-1.24) no screening effect (Figure 1) [26]
PLCO Lung USA 154,901 1993–2001 4 1993–2005? end 2009 11.9 years RRs assuming Poisson distribution. Adjusted p for sequential analyses (interim). No mention of how p calculated RR = 0.99 (95% CI, 0.87–1.22; p = 0.48)    no screening effect (no figure) [20]
PLCO Colorectal USA 154,900 1993–2001 2 1993–2004 end 2009 11.9 years (15.8 years) Weighted (0,1) LR test with RRs assuming Poisson distribution. Adjusted p for sequential analyses (interim) RR = 0.74 (95% CI 0.63 to 0.87; p < 0.001) Same for RRs though notably no test/p-value RR = 0.75 (95% CI 0.66–0.85) ~ 3 years (Figure 2a) [36]
PLCO Ovarian USA 78,216 1993–2001 4–6 1993–2005? 28 Feb 2010 12.4 years (14.8 years LTFU) Weighted (0,1) LR test (one-sided) with RRs assuming Poisson distribution. Adjusted p-value for sequential analyses (interim) RR = 1.18 (95% CI, 0.82–1.71) CI sequentially adjusted. No p value reported possibly because test was 1-sided? Same for RRs though notably no test/p-value (also added a Cox model) RR = 1.04 (95% CI 0.87–1.24) no screening effect (Figure 1) [37]
NLST Lung USA 53,454 2002–2004 3 2002–2007 end 2009 5.4 years (mean) RRs assuming Poisson distribution. Adjusted p for sequential analyses. Weighted RR = 0.80 (95% CI 0.73–0.93; P = 0.004)    ~ 1.5 years (Figure 1B) [24]
UK Flexible Sigmoidoscopy Screening Trial (UKFSST) Colorectal UK 170 034 1994–1999 1 1994–1999 31 Dec 2014 17·1 years HRs from Cox model. ITT HR = 0.69 (95% CI 0.59–0.82; p<0.0001). Same HR = 0.59 (95% CI 0.49–0.70) ~ 3 years (Figure 1G) [17]
Canadian National Breast Screening Study (CNBSS) Breast Canada 50,430 1980–1985 5 1980–1985 end 1991 8.5 years (mean) (25 years LTFU) T test on difference of proportions RR = 1.36 (95% CI 0.84–2.21) Cox PHs model HR = 0.99 (95% CI 0.88-1.12; p = 0.87) no screening effect (Figure 3) [11]
  1. *Estimate of mortality curve separation comes from visual inspection of appropriate published mortality plot if provided. The Figure number and paper reference are given to allow the reader to make their own judgement
  2. FU follow-up, LTFU long-term follow-up, RR rate ratio, HR hazard ratio, ITT intention to treat analysis, LR log-rank