Appendix 1: Critical appraisal on the internal validity of the CMCS7 using GATE11

Internal study validity - potential for bias
Evaluation criteria   How well were the criteria addressed?       Quality*
+ ~ x nr
Participants   Eligibles: inclusion and exclusion criteria:
-Sufficient detail?		  
Participants were:
1) Chinese men and women
2) Age group 35 to 64
16 centres in 11 provinces of China were selected non-randomly in 1992 and 1993, inclusion criteria of the centres include:
1) Having taken part in the Sino-MONICA project
2) Being able to conduct the study
There was over-representation in the urban areas (80.3% of participants)
For the addition of 3118 participants from Beijing in 1996 and 1999, the inclusion criteria were not stated.
Exclusion criteria was persons with history of myocardial infarction or angina pectoris (504 excluded).
  ~
    - Applied consistently (e.g. if multiple cohorts or multi-centre study)?   Yes   +
    - Were eligibles at a common point in the course of their condition?   Yes, all with no prior CHD.   +
    - If not, were differences addressed in analyses/interpretation
(e.g. stratified analyses)?		   Not done   nr
   
Participant selection from eligibles:
- Sufficient detail on selection process/strategy?
  At each of the 16 centres, stratified random sampling for each sex and 10-year age group was performed, but randomization process was not described in detail.   ~
    - Selection process appropriate to study objectives
(e.g. representative sample / consecutive cases)?		   Yes, the selection process by stratified random sampling for each sex and 10-year age group was appropriate.   +
    % eligibles selected who participated?   82%   +
Exposures and comparisons   Exposures (e.g. predictive / prognostic / risk / descriptive factors) and Comparison definitions:
-Sufficient detail? 		  Yes, considerable details given according to the WHO-MONICA protocol for risk factor surveys.   +
    - If appropriate, was there a comparison group?   Not applicable   nr
    - Objective and valid measurements?   Yes, well defined for all risk factors. The stratification of blood pressure, diagnostic criteria of diabetes were deliberately defined according to the Framingham model to ensure comparability.   +
    - Applied consistently to all participants?   No clear description on who did the measurements or quality assurance to ensure consistent application of measurements to all participants. Potential bias.   ~
    Were exposure group (EG) and comparison groups (CG) similar at baseline?   Comparison group not applicable here.   nr
    If groups not similar, were differences addressed in analyses / interpretation (e.g. multivariate analyses)?   Not applicable because no comparison group.   nr
    Were there likely to be residual differences between EG and CG that could have important effects on outcomes (i.e. confounding)?   Not applicable because no comparison group.   nr
    Was exposure / comparison re-measured during follow-up?   No, only a baseline survey on the risk factor was conducted.   nr
    Did any of CG become exposed during follow-up (contamination)? What %?   Not applicable because no comparison group.   nr
    Did any of EG become unexposed or receive the comparison during follow-up (contamination)? %?   No, the risk factor profile of a participant was defined at the beginning of the study.   +
    Was there sufficient contamination to cause important bias?
Was this addressed?		   Not applicable   nr
    Aside from Exp/Comp were EGs and CG treated equally during follow-up?   Presumably yes   +
    Was there sufficient co-intervention to cause important bias?   No   +
             
Outcomes and time  
Outcome definitions:
- Sufficient detail (could be replicated)?
  Yes, diagnosis of "hard" CHD events was according to the WHO-MONICA project. Trained physicians were sent to visit the patient or relatives, review the hospital records and complete a standard event form. The form was sent to the collaborating centre and reviewed by a group of investigators.   +
    - Objective and valid measurements?   Physicians making the diagnosis were trained before the study started and tested every 2 years. 20% of cases were verified by investigators from the collaborating centre.   +
    - Assessed blind to Exp / Comp classification or to their predictive/prognostic significance?   Physicians in the provincial collaborating centre were not blinded. Investigators in the Beijing collaborating centre who reviewed the standard event forms were also not blinded to the risk factors.   X
    - Applied consistently (e.g. if multi-centre study)?   Presumably yes   +
    Was follow-up long enough to detect important effects?   Long enough original 27003 participants (10 years) not for the additional 3118 participants followed up from 1996 to 1999.   +/~
Analyses   Were all participants initially allocated, accounted for at study conclusion?   No   ~
    What was the % lost to follow-up?   For original cohort (27003 participants) 1992-1995: 6% lost. At each stage of study? From 1996 onward: 38.7% lost because of ceasing of 6 centres on completion of national research project. 1996-2002: remaining 16552 participants, 14% lost.   ~
        3118 Beijing participants from 1996: 14% lost.    
        65.3% of the total cohort remained at the end of the study.    
    Was loss to follow-up sufficient to cause important bias?   A separate model was created to assess the effect of drop-out. The total person-years of follow-up and CHD events were reduced and the 95% CIs for some risk factor categories were wider after the exclusion of the participants who dropped out.   +
        The bias was properly assessed, and the RRs, 10-year CHD rates and prediction capabilities did not differ from the current cohort. There could be bias but the effect should be slight.   +
    Were all participants analyzed in groups they were initially allocated to (intention-to-follow)?   Presumably yes   +
Summary Quality Score for validity: how well did the study minimize bias?(+ or ~ or x)       +/~
*Criteria quality scores: += good, ~ = okay, x = poor, nr = not reported