The distribution of baseline characteristics of the men in the three cohort studies are presented in . As expected, the Whitehall cohort has a much higher proportion of individuals in social classes I and II (72.7%) compared to the Collaborative study (30.4%) and Renfrew & Paisley study (19.3%). The high proportion of people in social class IIIM, IV and V in the Collaborative (52.8%) and Renfrew & Paisley (69%) cohorts is unusual for a study of this type, but reflects the social class make-up of the communities from which the cohorts are drawn. Cholesterol levels were similar in the Collaborative and Renfrew & Paisley studies, and much lower in the Whitehall study. Blood pressure was similar in the Whitehall and Collaborative studies, and higher in the Renfrew & Paisley study. BMI was highest in the Renfrew & Paisley study and lowest in the Collaborative study. Men in the Whitehall study were the tallest and had the best FEV1 and men in the Renfrew & Paisley study the shortest with the worst FEV1. The largest baseline differences in the questionnaire measures relate to smoking (where there is a higher proportion of ex-smokers in the Whitehall study and lower proportion of current smokers) and self-reported morbidity (for cardiorespiratory symptoms and previous diagnoses).
Proportions and means+ for established risk factors by study.
gives the number of deaths and the age standardised mortality rates in each cohort by cause of death. The largest number of deaths was due to coronary heart disease (CHD), followed by respiratory causes, stroke and lung cancer. The mortality rates increase across the follow-up periods since the men are older in the later periods of follow-up and are consistently higher in the Renfrew & Paisley than the Collaborative cohort and the Whitehall cohort (at 41.6, 38.7 and 35.8 deaths per 1,000 person years respectively). The relative risk of mortality after age-adjustment was higher in both Scottish cohorts than in the Whitehall study for each specific cause and for all causes (). All-cause mortality was 25% and 41% higher, CHD mortality 32% and 41% higher, stroke mortality 55% and 73% higher, respiratory mortality 5% and 17% higher, lung cancer mortality 65% and 98% higher, mortality from accidents and suicide was 77% and 100% higher, and alcohol-related mortality 73% and 128% higher in the Collaborative and Renfrew & Paisley cohorts respectively as compared to the Whitehall cohort. The higher mortality rates were substantially attenuated with the addition of socio-economic position to the model, but remained higher for all-causes (8% and 17%), CHD (17% and 22%), stroke (45% and 60%), lung cancer (16% and 30%), accidents and suicide (56% and 70%), and alcohol-related causes (47% and 85%). The addition of single biological or behavioural factors in addition to age in the model (including smoking, FEV1, cardio-respiratory symptoms or history, height, blood pressure and cholesterol) were unable to explain as much of the higher all-cause (or any of the specific causes) mortality in the Scottish cohorts as socio-economic position (with the exceptions of FEV1 for CHD mortality and blood pressure for stroke mortality, in the Renfrew & Paisley study).
Age adjusted mortality rates by study and time of follow-up.
Hazard ratios for mortality in the Scottish cohorts compared to the Whitehall Study+ (reference) with successive adjustment for potential explanatory factors.
After adjusting for all risk factors except social class, all-cause mortality remained 9% and 7% higher, CHD mortality remained 16% and 1% higher, stroke mortality 44% and 36% higher, lung cancer mortality 24% and 38%, mortality due to accidents and suicide 57% and 78% higher and alcohol-related mortality 56% and 88% higher in the Collaborative and Renfrew & Paisley cohorts respectively compared to the Whitehall cohort. Respiratory mortality adjusts to be lower in the Scottish cohorts (), despite it being higher before adjustment for the earliest time period after baseline data collection ().
The fully adjusted model (including social class and all other explanatory factors together) explained almost all of the mortality excess in the Scottish cohorts for all-cause mortality but there remained some unexplained excess for the specific causes. Stroke mortality remained 45% and 37% higher, mortality from accidents and suicide 51% and 70% higher, and alcohol-related mortality 46% and 73% higher in the Collaborative and Renfrew & Paisley cohorts respectively compared with the Whitehall cohort. CHD mortality remained 11% higher in the Collabarative study and lung cancer remained 16% higher in the Renfrew & Paisley study in the fully adjusted model as compared to the Whitehall study. As before, respiratory mortality appeared to be lower in the Scottish cohorts with addition of all of the explanatory factors ().
Given the markedly different social class composition of the three cohorts, the baseline characteristics and hazard ratios adjusted for the biological and behavioural risk factors are also presented stratified by social class in and . The differences in baseline risk factors were less marked after stratification. Cholesterol levels were lower in the Whitehall study compared to the Scottish cohorts within each social class strata. Systolic and diastolic blood pressure were lower in each strata in the Whitehall cohort than in the Renfrew & Paisley cohort, but was higher in each strata than the Collaborative cohort with the exception of systolic blood pressure amongst those in strata IIIM, IV & V. Mean height in social classes I & II and IIINM were highest in the Collaborative study and lowest in the Renfrew & Paisley study (but highest in the Whitehall study in social classes IIIM, IV & V). Cigarette smoking was more prevalent in social class I & II in the Scottish cohorts compared to Whitehall, but not in the other social class strata. However, amongst those smokers, there were a greater mean number of cigarettes smoked per person in the Scottish cohorts for each social class strata. FEV1 was highest, and infective phlegm least common, in the Whitehall study except amongst those in social class IIINM where the mean FEV1 was highest and prevalence of infective phlegm lower, in the Collaborative study. The proportion with angina was higher in the Scottish cohorts but there was no consistent pattern for previous MI or for breathlessness. The prevalence of chronic bronchitis was lowest in the Collaborative study in all social classes.
Proportions and means+ for established risk factors in the Whitehall, Collaborative and Renfrew & Paisley studies by social class.
Hazard ratios for all-cause mortality in the Scottish cohorts compared to the Whitehall Study+ (reference) separately by social class with successive adjustment for potential explanatory factors.
shows that there is an excess in all-cause mortality in social class I & II of 9% in the Collaborative and 22% in the Renfrew & Paisley study compared to Whitehall after age adjustment, but the excesses estimated in the other social class strata are not so large or precise (3% and 11% for social class IIINM, and 6% and 13% in social classes IIIM, IV and V, for the Collaborative and Renfrew & Paisley cohorts respectively) and those for the Collaborative study may be due to chance. Adding the other baseline characteristics to the model completely removes the excess mortality in the Scottish cohorts within each social class strata.
shows the differences in mortality between the cohorts stratified by the follow-up time. It shows that the hazard ratios in the Scottish cohorts in comparison with Whitehall declined over time. The excesses in the 10–19 yr period are slightly less well explained whereas the smaller excesses in the ≥20 yr period are completed explained (from 1.10 and 1.01 in the first 10 years of follow-up to 0.96 and 0.95 after >20 years of follow-up in the fully adjusted model in the Collaborative and Renfrew & Paisley cohorts respectively).
Hazard ratios for all-cause mortality in the Scottish cohorts compared to the Whitehall Study+ (reference) by time of follow-up.
The sensitivity analyses using a smaller height correction for measurement in shoes in the Whitehall cohort (of 1.27 cm as opposed to 2.54 cm) are shown in tables S1
. The impact on the hazard ratios is small and does not change the overall findings.