In comparing gender ratios between case and control groups across four studies, a notable variation emerges in the observed associations. Swanson et al.’s study reveals a significant disparity, with individuals in the study group having substantially lower odds of being in the case group compared to controls (p < 0.0001). Conversely, Van Dyck et al., McDade et al., and Logovinsky et al. report less pronounced differences, with p-values of 0.179, 0.688, and 0.488, respectively, failing to reach statistical significance (Table 3).

Table 3 Interpreting gender differences in included studiesTreatment effectLecanemab 10 mg/kg vs. placebo effect on ADCOMS change

The findings from our meta-analysis revealed a statistically significant decrease (p-value < 0.0001) in ADCOMS scores when lecanemab is administered at a 10 mg/kg dosage compared to a placebo. Both the common effect model and the random effects model estimate a mean difference of − 0.0508, and the narrow 95% confidence intervals suggest a consistent effect size across the studies.

Furthermore, the absence of significant heterogeneity (I2 = 0.0%) among the studies implies a consensus regarding the impact of lecanemab on ADCOMS scores. The test of heterogeneity (Q statistic) yields a p-value of 0.9545, indicating a lack of significant heterogeneity, and suggesting a high degree of consistency in the findings across the studies (Fig. 2A). For further details, please refer to Sect. 1 in the supplementary file.

Fig. 2

A–C Forest plots depicting lecanemab 10 mg/kg efficacy against AD parameters (Fig A: ADCOMS; Fig 2: CDR-SB; Fig 3: ADAS-cog14) 

Lecanemab 10 mg/kg vs. placebo effect on CDR-SB change

The results of the meta-analysis suggest that the administration of lecanemab at a dosage of 10 mg/kg biweekly is linked to a statistically significant reduction (p < 0.0001) in CDR-SB scores compared to a placebo. Both the common effect model and the random effects model estimate a mean difference of approximately − 0.4264, with narrow 95% confidence intervals, indicating consistency in the effect size across the studies.

The absence of significant heterogeneity (I2 = 0.0%) among the studies indicates a consensus in their findings, further supporting the strength of the results. The test of heterogeneity (Q statistic) yields a p-value of 0.9339, signifying a lack of significant heterogeneity among the studies and suggesting relative consistency in their findings (Fig. 2B). For further details, please refer to Sect. 2 in the supplementary file.

Lecanemab 10 mg/kg vs. placebo effect on ADAS-cog14 change

The results of the meta-analysis indicate that administering lecanemab at a 10 mg/kg dosage biweekly leads to a statistically significant reduction (p < 0.0001) in ADAS-cog14 scores compared to a placebo in patients diagnosed with Alzheimer’s disease. Both the common effect model and the random effects model estimate a mean difference of approximately − 1.3416, with narrow 95% confidence intervals, highlighting a consistent effect size across studies.

The absence of significant heterogeneity (I2 = 0.0%) among the studies suggests agreement in their findings, further supporting the reliability of the results. The heterogeneity test (Q statistic) yields a p-value of 0.4550, indicating a lack of significant heterogeneity among the studies and suggesting a relatively consistent pattern in their findings (Fig. 2C). For further details, please refer to Sect. 3 in the supplementary file.

Safety concernsLecanemab 10 mg/kg vs. placebo effect on any TEAE outcome

The meta-analysis results concerning the occurrence of TEAE when comparing lecanemab at a dosage of 10 mg/kg administered biweekly to placebo shows that the random effects model estimates a pooled relative risk (RR) of 0.6647. However, the 95% CI is wide indicating no effect, and the p-value is 0.3034, signifying a lack of statistical significance.

Additionally, there is significant heterogeneity among the studies, with an I2 of 96.8%. This suggests diverse findings regarding TEAE outcomes across the studies (Fig. 3A). For further details, please refer to Sect. 4 in the supplementary file.

Fig. 3

A–C The safety assessment of lecanemab at a dosage of 10 mg/kg in AD patients (Fig A: TEAE; Fig B: ARIA-E; Fig C: ARIA-H)

Lecanemab 10 mg/kg vs. placebo effect on ARIA-E outcome

The meta-analysis findings for the occurrence of ARIA-E when comparing lecanemab at a dosage of 10 mg/kg administered biweekly to placebo are outlined as shows that the random effects model estimates a pooled RR of 7.9613, with a 95% CI of [4.8358; 13.1068], and a very low p-value (< 0.0001), indicating a statistically significant increased risk of ARIA-E associated with lecanemab treatment.

Furthermore, there is no significant heterogeneity among the studies, as indicated by the p-value of 0.5430 and an I2 of 0.0%. This implies that the studies are consistent in their findings regarding the risk of ARIA-E (Fig. 3B). For further details, please refer to Sect. 5 in the supplementary file.

Lecanemab 10 mg/kg vs. placebo effect on ARIA-H outcome

The meta-analysis results for the occurrence of ARIA-H when comparing lecanemab at a dosage of 10 mg/kg administered biweekly to placebo reveal that the pooled RR estimated by the random effects model is 1.7533, with a 95% CI of [1.3488; 2.2790], and a very low p-value (< 0.0001). This indicates a statistically significant increased risk of ARIA-H associated with lecanemab treatment.

Moreover, there is no significant heterogeneity among the studies, as indicated by the p-value of 0.4088 and an I2 of 0.0%. This suggests that the studies are consistent in their findings regarding the risk of ARIA-H (Fig. 3C). Retrieved data related to adverse events are depicted in Table 4. For further details, please refer to Sect. 6 in the supplementary file.

Table 4 List for adverse eventsPublication bias testing and sensitivity analyses

Funnel plots and Egger’s tests were carried out to evaluate the publication bias (Fig. 4A, B, C). The results indicated no evidence of significant publication bias, as demonstrated by the Egger’s test with a p-value greater than 0.05.

Fig. 4

A–C Funnel plots and Egger’s tests evaluating the publication bias (Fig A: ADCOMS; Fig 2: CDR-SB; Fig 3: ADAS-cog14) 

The results of the regression test for funnel plot asymmetry for ADCOMS

The results of the regression test for funnel plot asymmetry suggest that there is no statistically significant evidence of publication bias or funnel plot asymmetry in the meta-analysis. The p-value of 0.9564 indicates that the relationship between study precision SE and effect size is not significantly different from what would be expected by chance. In other words, the distribution of studies in the funnel plot appears to be symmetric and not skewed due to publication bias.

The limit estimate as SE approaches zero gives an estimate of the effect size under ideal conditions, if the studies had perfect precision. The point estimate of − 0.0491 suggests that, under these ideal conditions, the mean difference or effect size is approximately − 0.0491. However, the wide confidence interval (− 0.1117 to 0.0136) indicates substantial uncertainty in this estimate.

The results of the regression test for funnel plot asymmetry for CDR-SB

The results of the regression test for funnel plot asymmetry suggest that there is no statistically significant evidence of publication bias or funnel plot asymmetry in the meta-analysis. The p-value of 0.7808 indicates that the relationship between study precision SE and effect size is not significantly different from what would be expected by chance. In other words, the distribution of studies in the funnel plot appears to be symmetric and not skewed due to publication bias.

The limit estimate as SE approaches zero gives an estimate of the effect size under ideal conditions, assuming that the studies had perfect precision. The point estimate of − 0.5044 suggests that, under these ideal conditions, the mean difference or effect size is approximately − 0.5044. However, the wide confidence interval (− 1.0807 to 0.0719) indicates substantial uncertainty in this estimate.

The results of the regression test for funnel plot asymmetry for ADAS-cog14

The results of the regression test for funnel plot asymmetry suggest that there is no statistically significant evidence of publication bias or funnel plot asymmetry in the meta-analysis. The p-value of 0.8812 indicates that the relationship between study precision SE and effect size is not significantly different from what would be expected by chance. In other words, the distribution of studies in the funnel plot appears to be symmetric and not skewed due to publication bias.

The limit estimate as SE approaches zero gives an estimate of the effect size under ideal conditions, assuming that the studies had perfect precision. The point estimate of − 1.1020 suggests that, under these ideal conditions, the mean difference or effect size is approximately − 1.1020. However, the wide confidence interval (− 4.2541 to 2.0502) indicates substantial uncertainty in this estimate.

Variability and heterogeneity

The heterogeneity p-value is a statistical measure that helps assess whether there is significant variability in the effect sizes observed across different studies included in a meta-analysis. A low p-value (typically less than 0.05) suggests the presence of significant heterogeneity, indicating that the variation in effect sizes is not likely due to chance alone.

The Egger’s test p-value > 0.05 generally indicates no evidence of significant publication bias. The I2 value represents the proportion of total variability across studies due to heterogeneity.

These results suggest a lack of significant heterogeneity in the analyses related to ADCOMS Change, CDR-SB Change, and ADAS-cog14 Change, with I2 values of 0.0%. The Egger’s test for publication bias yielded non-significant results across all analyses, indicating a relatively consistent pattern in the findings without substantial publication bias. However, the high heterogeneity (I2 = 96.8%) and statistically significant p-value (< 0.0001) for the “Any TEAE Outcome” indicate substantial variability in the TEAE outcomes across the included studies this variability may be due to the huge difference in sample size between studies. This level of heterogeneity suggests that there might be differences in study populations, interventions, or methodologies that contribute to the observed variation in TEAE outcomes (Table 5).

Table 5 Variability and heterogeneityQuality assessment

The quality of research design and reporting was assessed in several studies by Swanson et al., Logovinsky et al., Van Dyck et al., and Mcade et al. The analyzed studies present varying levels of bias across different domains. Swanson et al. and Logovinsky et al. have low risk of bias in multiple domains, particularly in terms of selection and performance. Van Dyck et al. demonstrate consistently low risk across all domains. On the other hand, Mcade et al. exhibit high risk in the selection and performance domains. Overall, domains such as selection and performance show lower bias across the studies, while attention to minimizing bias is warranted in domains such as detection and attrition, particularly in the study by Mcade et al. These findings indicate variations in the methodological rigor and potential bias in these studies (Fig. 5) (Table 6). For further details, please refer to Sect. 7 in the supplementary file.

Fig. 5

Variations in the methodological rigor and potential bias in included studies via the Cochrane risk of bias assessment tool

Table 6 Cochrane risk of bias assessment tool for RCTs