Skip to main content

Table 7 The effect of considering inconsistency when judging quality of evidence

From: Judging the quality of evidence in reviews of prognostic factor research: adapting the GRADE framework

Factor Example
Inconsistency Our narrative systematic review included three studies examining chronological age as a prognostic factor for persistence of headache [2628]. While Larsson and Sund [27] found in a 1-year cohort study with 2,465 schoolchildren aged between 12 and 15 years that the risk of persistence of frequent headaches increases with age (odds ratio = 1.88, 95% confidence interval = 1.65 to 2.15), Kienbacher and colleagues [26] and Wang and colleagues [28] did not reach the same conclusion when running longer term follow-up studies with children and adolescents with a headache diagnosis. Kienbacher and colleagues [26] performed a follow-up study on 227 children and adolescents with migraine or tension-type headache. The average follow-up period was 6.6 years, with the age ranging from 5 to 8 years. Wang and colleagues [28] performed an 8-year follow-up study with adolescents with migraine. In both cases, age was reported not to be associated with persistence of headache. There is inconsistency of results across these three studies; however, neither of the two longer term cohort studies reported either means and a measure of variance or results from the statistical analyses to judge the level of overlap of the confidence intervals between the three studies. Duration of the follow-up, outcome and headache population were very different across studies. As far as the duration of the follow-up is concerned, despite the fact that the cohort study with short-term follow-up reported a significant risk estimate for an advanced age which was not found in cohort studies with longer term follow-ups, we did not find a plausible explanation for such potential influence of the duration of follow-up on the association between age and persistence of headaches. However, the outcome and the headache population could explain some of this heterogeneity [29]. Consequently, we did not consider the inconsistency across findings examining chronological age as a prognostic factor for headache persistence to be very serious and rated this item as serious inconsistency.