TABLE 3.

Should ketogenic therapies compared with control be used for people with mild cognitive impairment?

GRADE domain	Judgement	Concerns about certainty domains
Methodological limitation of the studies	1/5 trials (49) had low risk of bias across the 3 assessed domains and had the largest sample size (52 participants). Another study (57) did not clearly report on allocation concealment or sequence generation, which lowers confidence in the certainty of the evidence. The remaining 3 studies (55, 58, 61) were judged to have some or high risk of bias across methodological items assessed. Therefore, we have judged the trials to have very serious methodological limitations	Very serious
Indirectness	The patients, intervention, and comparators within included studies all provide direct evidence to the clinical question at hand. All studies included participants with a diagnosis of MCI. 4/6 studies measured cognition as the primary outcome, whilst 2 other individual studies measured brain ketone metabolism (49) and CSF AD biomarkers (58). Further, 4/6 studies implemented a dietary intervention (51, 57, 58, 61) whilst the remaining 2 trials (49, 55) used MCTs. The predominant primary outcome measure in this sample (cognition) was assessed using varying neuropsychological testing methods in different trials. We also note variation in the modified ketogenic diets adapted across the sample. We judged the trials to have no serious indirectness	Not serious
Imprecision	The total number of participants included in all trials was 130, which is concerning. Some studies reported small improvements in cognition using varying neuropsychological assessments (55, 61) and other trials reported “nonsignificant results” (51, 57) likely because of enrolling a small number of participants. We judged the evidence to have very serious imprecision	Very serious
Inconsistency	Overall, mixed results were consistently seen across studies which examined cognition as the primary outcome measure (4/6 studies) (51, 55, 57, 61). The direction and magnitude of effect varied across trials. One study reported an improvement in verbal learning memory (medium effect-size) (61) whilst another reported some improvements in specific testing outcomes within the ADAS-Cog, although this study had an extremely small sample and was deemed to be at high risk of bias (55). The remaining studies showed no significant effects on cognition, although effects were trending in the right direction. We judged the evidence to have serious inconsistency	Serious
Publication bias	We did not suspect publication bias, due to the presence of both positive and negative trials in the evidence base. Furthermore, the systematic search strategy was comprehensive and covered an extensive number of databases	Not suspected

AD, Alzheimer's disease; ADAS-Cog, Alzheimer's Disease Assessment Scale-Cognitive Subscale; CSF, cerebrospinal fluid; GRADE, Grading of Recommendations, Assessment, Development and Evaluation; MCI, mild cognitive impairment; MCT, medium-chain triglycerides.