15581-m-nanninga

CHAPTER 4 70 follow-up in the case of cohort studies. Subsequently, based on the score, studies were classified into three categories: weak ( N= 22), moderate ( N= 102), or strong quality ( N= 22) (Figure 2). Interrater Cohen’s weighted kappas were 0.66 (M.N./J.T.), with p pos =0.50 and p neg =0.95, and 0.44 (M.N./D.E.M.C.J.), with p pos =0.55 and p neg = 0.88 [21, 22]. ‘Strong’ quality meant that no major flaws threatened the internal validity of the study, i.e. regarding any of the five dimensions, implying minor chances of selection bias, information bias, and uncontrolled confounding. Quality was scored as ‘moderate’ when at least in one of the dimensions a flaw occurred that challenged the confidence that could be attached to the results. Examples were: no adjustment for confounders, or a low response rate with selective non-response or lack of non-response information. In case major and/or severe flaws occurred, the quality was scored ‘weak’. We only included strong studies in order to obtain unbiased and meaningful evidence [25].