aIs assigned if there are a considerable number of studies including prospective observational studies and, wherever possible, randomised controlled intervention studies of sufficient size, duration and quality with consistent results
bIs assigned if epidemiological studies show fairly consistent relations between factor and disease, but there are noticeable weaknesses regarding the evidence or there is evidence of an opposite relation, which does not allow a definite judgement
cIs assigned if the results on an association between exposure and target disease are mainly based upon case–control studies and cross-sectional studies. There are only insufficiently performed controlled intervention studies, observational studies, or non-controlled clinical trials
dIs assigned if there are a few study results that indicate an association between a factor and a disease, but they are not sufficient to establish the relation. There is only limited or no evidence from randomised intervention studies
Based on the number of available studies and their classifications of level of evidence, the judgement of the strength of the evidence was performed (Table 2). In total, four categories of the strength of the evidence were used [3], termed as convincing, probable, possible, and insufficient. Table 2 shows the connection between the levels of evidence of the studies and the overall strength of the evidence. In addition to the relation between the level of evidence of the studies and the strength of the evidence, there were also further specifications that determined the strength of the evidence:
Convincing evidence regarding a preventive effect or a lack of an association
The strength of evidence was judged as “convincing” if at least 2 studies of highest quality (level of evidence I) showed consistent results. If the studies showed methodological weakness or were only cohort studies, the minimum number of intervention studies was raised to 5. However, for this strength of the evidence, it was required that the question has been extensively investigated and that there were a lot of results from different study populations including comprehensive data on consumption. Results from cohort studies should have been confirmed by intervention studies with intermediary markers regarding causality. Ideally, a meta-analysis of the present studies is available that did neither indicate heterogeneous study results nor include a high percentage of study results with opposite effects.
Probable evidence regarding a preventive effect or a lack of an association
The strength of the evidence was judged as “probable” if epidemiological studies showed consistent relations between factor and disease, but also showed weaknesses regarding the causal argumentation. This may be the case, for example, (1) if compared with the “positive” studies, there are a considerable number of studies without risk relation, (2) if there is a lack of study results or inconsistent results from intervention studies with intermediary markers, or (3) if meta-analyses gave heterogeneous results. The number of studies required to classify the strength of the evidence as “probable” remains at not