Table 4 Analyses to convey clinical trials results in non-trial settings

Standardized effect size (e.g., Cohen’s d)

• For a comparison of group means, Cohen’s d large effect size: ≥0.8; medium effect size 0.5–0.8; small effect size: 0.2–0.5

• Cohen’s d can be applied to any continuous measure, including those used in preclinical AD and MCI due to AD [42, 64]

• Effect sizes can provide an interpretable index of the direction and magnitude of the effect of an intervention [65–67]

• Because effect size estimates enable some control of variability, they also allow for some level of comparison across similar studies [65–68]

Relative risk (RR)/odds ratio (OR)

• Complement standard effect sizes such as Cohen’s d [65–67]

• Useful for estimating effect sizes from categorical measures, such as “improved” versus “not improved” or “converted to MCI” or “did not convert to MCI” [66, 69]

• Both RR and OR are ways in which clinicians often generally consider treatment effects [66]

Number needed to treat (NNT)/number needed to harm (NNH)

• A high NNT indicates a less effective treatment [70]: interventions with an NNT in the single or low double digits are generally considered effective, although an NNT in the lower hundreds may also be considered useful, depending on the significance of the outcome, such as preventing death [71]

• NNH provides similar index for the occurrence of one or more specified adverse events

• NNT, which is related to absolute risk reduction, may be the effect size estimate that best reflects clinical significance for binary outcomes such as success or failure [66]

Time-to-event (TTE)

• Versatile analytical method also known as survival analysis [17]

• Refers to a set of methods for analyzing the length of time until the occurrence of a well-defined endpoint of interest [17]

• In preclinical and early AD trials, an outcome of great interest is conversion from one stage (e.g., preclinical AD) to the next (MCI due to AD) on the AD continuum [17]

• Determining what is considered a meaningful event can be challenging, especially in early AD. In addition, operationalizing transitions may be burdensome given the subtle differences between AD stages; nonetheless, TTE analyses may provide useful information as part of a broader evaluation of effect

• Recognized as endpoint option by both FDA and EMA [10]

Meaningful change or difference threshold

• Reflects the level of score change(s) on a COA that is perceived to be meaningful in the target patient population

• There are two main approaches:

o Clinically meaningful change thresholds for individual patients (within-patient approach, recommended by the FDA) [53]

o Clinically important difference thresholds applied at the group level (between-groups approach) [72]

• The aim is to ensure that the observed treatment benefit as measured by the COA is meaningful to a patient

• Responder analyses convey the proportion of patients who meaningfully respond to treatment (i.e., achieve or exceed the within-patient meaningful change threshold)

• In the context of a progressive disease like AD, a progressor analysis may be more appropriate (i.e., define meaningful progression on the COA), to demonstrate the proportion of patients who meaningfully progress on treatment vs placebo

• Exceeding a threshold for MCID between groups can support the meaningfulness of a statistically significant treatment effect at the group level