Objective To assess the efficacy and safety of Aβ-targeting agents for mild to moderate Alzheimer’s disease.
Methods The MEDLINE, Embase, Cochrane Central Register of Controlled Trials, PsycINFO, ClinicalTrials.gov and the WHO’s International Clinical Trials Registry Platform search portal were searched from their inception to April 2020. We generated pooled estimates using random effects meta-analyses.
Results Nineteen randomised controlled trials, of which 17 had a low risk of bias, included 12 903 participants. The meta-analysis showed no difference in the cognitive subscale of Alzheimer’s Disease Assessment Scale (ADAS-Cog) between anti-Aβ drugs and placebo (mean difference (MD): 0.20, 95% CI −0.40 to 0.81; I 2=99.8%; minimal important difference 3.1–3.8 points, moderate-certainty evidence). For ADAS-Cog, results suggested that one drug that increases Aβ clearance may differ in effect (MD: −0.96, 95% CI −0.99 to −0.92) from drugs that reduce Aβ production (MD: 0.78, 95% CI 0.25 to 1.32) (interaction p<0.000001); this difference also existed in the outcome of MMSE and CDR-SOB. Compared with placebo, anti-Aβ drug-related adverse events were as follows: anxiety, depression, diarrhoea, fatigue, rash, syncope and vomit.
Discussion From current evidence, anti-Aβ interventions are unlikely to have an important impact on slowing cognitive or functional decline. Although the subgroup analysis suggested possible benefits from Aβ clearance drugs, the analysis has limited credibility, and a benefit from drugs that increase clearance, if real, is very small.
Trial registration number PROSPERO registration number CRD42019126272.
Statistics from Altmetric.com
Alzheimer's disease (AD) occurs decades prior to the onset of clinical symptoms, which are accompanied by deposition of amyloid-beta (Aβ) peptide plaques that accumulate in the cortex and hippocampus.1 2 Considering the early pathophysiological characteristics that precede clinically evident findings, halting or reversing the pathophysiological process in the early stage is a preferable strategy to the late stage. For the last 20 years, great efforts have been made to investigate Aβ-targeting agents as a potential disease-modifying therapy (DMT) for mild to moderate AD.3 So far, although no Aβ-targeting agent has been approved by the US Food and Drug Administration, agents directed at amyloid-related targets are leading the drug development process for new AD therapies.4
Some evidence supports the hypothesis that AD is caused by an imbalance between Aβ production and Aβ clearance. Among all strategies to reduce Aβ production, the inhibition of β-secretase and γ-secretase is one of the first therapeutic strategies formulated after the amyloid cascade hypothesis, and it is still being tested today.5 Additionally, studies in agents targeting the Aβ clearance in AD have been published and increasing in recent years.6 7 To date, preclinical studies and some clinical trials have shown the effect of Aβ-targeting agents on cognitive or functional decline in mild to moderate AD. For instance, multiple small molecules can reduce Aβ synthesis in vitro8–11 and facilitate Aβ removal and improve memory in transgenic mice.12 13 Some phase II clinical trials of an Aβ-targeting agent demonstrated its antibody responses, target engagement and attenuated neurodegeneration in mild to moderate AD.14–16
However, findings from some large trials have been inconsistent. Three recent randomised placebo-controlled trials17–19 evaluated the efficacy and safety of Aβ-targeting agents for patients with mild to moderate AD. All three chose the change in the Cognitive Subscale of the Alzheimer’s Disease Assessment Scale (ADAS-Cog) and the Alzheimer’s Disease Cooperative Study Activities of Daily Living Inventory Scale (ADCS-ADL) from baseline to weeks 72–80 as coprimary end points. These trials suggested no cognitive or functional improvement in patients with mild to moderate AD but more adverse events (AEs). The failure of several recent pathology-based strategies has highlighted the urgent need for effective targeting agents.20 In addition, anti-Aβ immunotherapy, one of the most promising approaches to modify the Aβ load, has been shown to effectively remove Aβ plaques from the brain, but this apparent success has failed to reverse the cognitive deficits in patients with AD.21 22
Currently, there is an overwhelming amount of ongoing and soon-to-start randomised controlled trials (RCTs) with innovative drugs in this field,23 but no comprehensive review on current best evidence is available. Moreover, gaps (Aβ-targeting agents can bring clinical benefits and safety, and a certain type of effective drugs exist) need to be filled to identify the direction of future clinical trials using Aβ-targeting agents that may have a high failure risk. We therefore performed a systematic review and meta-analysis of RCTs to assess the efficacy and safety of Aβ-targeting agents for mild to moderate AD.
We included studies that fulfilled the following criteria: study populations defined as patients with mild to moderate cognitive impairment; RCTs that compared anti-Aβ drugs, including inhibitors of synthetic Aβ or agents facilitating the removal of Aβ plaques with placebo or alternative doses of the drugs; and studies that reported on one of the primary or secondary outcomes specified as follows. The exclusion criteria were meeting proceedings and no availability of the full-text article. The primary outcome was cognitive function measured by ADAS-Cog (lower scores mean better cognitive function). Two commonly used approaches to estimate minimal important difference (MID) are an anchor-based approach and distribution-based approaches.24 Both of these approaches have been used in clinical trials for dementia, one study involving a survey of cognitive abilities ranging from normal to moderate–severe AD dementia, on average, 1.4 points for Mini-Mental State Examination (MMSE) and 3.5 for activities of daily living (ADLs), were indicative of a meaningful decline in the clinician’s assessment.24 The MID for ADAS-Cog has been reported as a reduction of 3.1–3.8 points.25 Secondary outcomes included functional outcomes measured by ADCS-ADL (higher scores mean better ADLs) and plasma/cerebrospinal fluid (CSF) biomarkers. Other outcomes included MMSE and CDR-SOB.
We searched MEDLINE, Embase, the Cochrane Library and PsycINFO from inception to 31 April 2020. We established search strategies that combined database-specific subject headings (such as Medical Subject Headings terms) and free text terms (such as AD, cognition, inhibit Aβ synthetic and randomised clinical trials) for potentially eligible studies (online supplementary eMethods presents the search strategy). We also searched ClinicalTrials.gov and the WHO’s International Clinical Trials Registry Platform search portal in April 2020 to identify any registered yet unpublished or ongoing RCTs.
Study selection and data extraction
Two reviewers (XZ and YG) independently screened titles and abstracts for eligible full-text studies and extracted data from the eligible studies using a predefined data extraction form. Disagreements regarding the eligibility of full-text articles were resolved by discussion or arbitration with a third reviewer (LL).
Information gathered from the included studies included study characteristics (eg, author name, year of publication, study design and sample size); population characteristics (eg, mean age, sex, race, education level, baseline MMSE, baseline Clinical Dementia Rating Scale Sum of Boxes (CDR-SOB) and degree of AD; intervention characteristics (eg, mode of treatment, type of drug, dose, and duration of study treatment), comparison characteristics (eg, type of control, dose and duration of treatment); and outcomes (eg, ADAS-Cog, ADCS-ADL, MMSE, CDR-SOB and length of follow-up). For trials with multiple follow-up points, data or reports, we collected the outcome data at the longest follow-up.
Risk of bias assessment
Two reviewers (XZ and YG) independently assessed the risk of bias in the RCTs based on the Cochrane risk of bias tool.26 This instrument consists of seven domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessments, incomplete outcome data, selective outcome reporting and other sources of bias. The tool ranks the evidence from RCTs as having ‘definitely yes (low risk of bias)’, ‘probably yes’, ‘probably no’ and ‘definitely no (high risk of bias)’ levels of bias. The overall risk of bias for individual studies was classified as low risk of bias if all domains were ranked as definitely yes or probably yes, as high risk of bias if at least one domain ranked as definitely no or probably no.26 Attrition bias was judged as high risk if more than 5% of patients were lost to follow-up.
Trial sequential analysis
We performed trial sequential analysis (TSA) to avoid random errors due to sparse data and repetitive testing of the accumulating data.27 We used the diversity-adjusted required information size estimated from a control event proportion of the included trials and an overall 5% risk of a type I error and 10% risk of a type II error (a power of 90%) and the diversity, which was estimated in the included trials. if the cumulative Z-curve crosses a trial sequential monitoring boundary (TSMB), a sufficient level of evidence is reached and no further trials may be needed. However, if the cumulative Z-curve does not cross the TSMB or does not surpass the futility boundaries, it was probably necessary to continue doing trials in order to detect or reject a certain intervention effect. We conducted trial sequential analysis using software from the Copenhagen Trial Unit.28
To evaluate treatment effectiveness, we conducted our meta-analysis using the metaphor package and R V.3.5.3 statistical software. For continuous outcomes, we present the calculated weighted mean differences (MDs) with 95% CIs using an inverse variance random effects model. If a study with multiple intervention groups (different doses) was included, we combined all relevant experimental intervention groups of the study into a single group (the form of comparison: all active vs placebo), and the means and SD were combined using formulas for continuous outcomes in the Cochrane Handbook.29
Based on the assumption that clinical and methodological heterogeneity was likely to exist and to have an impact on the outcomes, we used the random eﬀects meta-analysis model to estimate the MDs. The DerSimonian and Laird method of moments estimator was used to estimate the between-study variance, and the Wald-type method was used to calculate 95% CIs.9 29 30 Statistical heterogeneity was assessed using the I 2 statistic.
Regardless of the observed statistical heterogeneity, and when the evidence was available (at least two included RCTs in one subgroup), we conducted the following prespecified subgroup analyses: clinical phase (phase II and phase III); length of treatment (≤48 and >48 weeks); target of Aβ (reducing Aβ production and increasing Aβ clearance); and type of drug (r-secretase inhibitor, β-secretase inhibitor and intravenous immunoglobulin). We hypothesised that phase III, >48 weeks, increasing Aβ clearance and β-secretase inhibitor had a larger effect than phase II, ≤48 weeks, reducing Aβ production and other types of drugs. We performed an interaction test between different subgroups. To evaluate the credibility of subgroup effects, we followed recent guidance of criteria to evaluate the credibility of subgroup analyses.31
Small study effects were assessed by Egger’s test32 and contour-enhanced funnel plots33 (the trend of estimated intervention effects in smaller studies is different from that in larger studies and can result from reporting biases, biases in study conduct or other factors).
We rated the certainty of evidence for each outcome using the Grading of Recommendations Assessment Development and Evaluation (GRADE).34
Data are available to qualified investigators on request to the corresponding author.
Figure 1 shows a flowchart of the literature search and study selection. Of 4656 records identified from the electronic databases, 3517 were excluded on the basis of the title and abstract. Of the 80 records reviewed in full, 55 were excluded for failure to meet the eligibility criteria. The remaining 1917–19 35–40e1-10RCTs, including 12 903 participants, met eligibility criteria.
Table 1 provides a summary of the baseline characteristics of the 19 eligible RCTs. Ten (53%) were phase II studies, and seven (37%) were phase III studies. Typical patients were between 70 and 75 years of age, and the majority were women. The treatment duration ranged from 4 to 104 weeks, and follow-up ranged from 4 to 48 weeks. The treatment groups in all trials were anti-Aβ drug therapy, and the control was placebo. Among all trials, 8 (42%) studies were 2-arm, 8 (42%) studies were 3-arm, 1 (5%) studies were 4-arm and 2 (10%) studies were 5 -arm in treatment groups.
Risk of bias assessment
Seventeen (89.5%) studies were at low overall risk of bias and two (10%) studiese15 were at high overall risk because of insufficient reporting of randomisation, allocation concealment and blinding of outcome assessment. In addition, two studies (10%) e15 at high overall risk due to loss to follow-up (missing AE data). Agreement between reviewers was substantial for six domains and perfect for one domain (figure 2 and online supplementary eTable1).
Primary outcome measure
Cognitive Subscale of the Alzheimer’s Disease Assessment Scale
Figure 3 shows the effect of anti-Aβ drugs in the 18 studies that reported on the mean change in the ADAS-Cog score from baseline. Compared with placebo, anti-Aβ treatment showed no difference in ADAS-Cog score (MD: 0.20, 95% CI: −0.40 to 0.81; I 2=100%; MID 3.1 to 3.8 point, moderate-certainty evidence; table 2).
There was no visual asymmetry in the funnel plot between the anti-Aβ treatment and the placebo groups (online supplementary eFigure 1, Egger’s test p=0.82).
For the subgroup analysis (online supplementary eFigure 2), only one subgroup analysis showed a potential subgroup effect: drugs of increasing Aβ clearance suggested a positive benefit effect (MD: −0.96, 95% CI0.99 to −0.92; I 2=0%) not seen with drugs reduce Aβ production (MD: 0.78, 95% CI: 0.25 to 1.32; I 2=100%) (interaction p<0.00). The apparent effect in the Aβ clearance increasing drugs was about a third of the MID.
Trial sequential analysis on ADAS-Cog in 15 trials was performed (online supplementary eFigure 3). The blue line (Z-curve) shows the cumulative meta-analysis adding the results of individual trials based on the year of publication. The horizontal green line represents the 5% level of significance. The monitoring boundaries (inward sloping red lines) show the significance level after adjusting for the cumulative analysis. The vertical red line shows the required information size (the number of participants needed to determine if firm evidence was established). We conducted the trial sequential analysis with the alpha set to 5%, power to 90%, MD to −3.1, variance to low bias, and heterogeneity correction based on model variance. The diversity-adjusted required information size was 2175 participants. In total, the cumulative meta-analysis included 5974 participants in the anti-Aβ drug group and 4067 in the placebo group. The cumulative Z-curve crossed the monitoring boundary and required information size, implying that a sufficient level of evidence was reached and no further trials may be needed.
Secondary outcome measures
Alzheimer’s Disease Cooperative Study Activities of Daily Living Inventory Scale
Nine studies reported the mean change in the ADCS-ADL score from baseline (online supplementary eFigure 4). The overall pooled effect provided no support for an effect of anti-Aβ drugs on the ADCS-ADL score compared with the placebo treatment (MD: −0.73, 95% CI −1.74 to 0.28; I 2=100%, MID: 3.5, moderate-certainty evidence; online supplementary eFigure 4 and table 2). The subgroup of increasing Aβ clearance versus placebo included only one RCT, so the interaction test was not performed (online supplementary eFigure 5A).
Trial sequential analysis on ADCS-ADL in nine trials was performed (online supplementary eFigure 6). We conducted the trial sequential analysis with the alpha set to 5%, power to 90%, MD to 3.5, variance to low bias and heterogeneity correction based on model variance. The diversity-adjusted required information size was 4457 participants. In total, the cumulative meta-analysis included 5444 participants in the anti-Aβ drugs group and 3664 in the placebo group. The cumulative Z-curve crossed the monitoring boundary and required information size, implying that a sufficient level of evidence was reached and no further trials may be needed.
The results of the MMSE meta-analyses suggested no effect of anti-Aβ drugs versus the placebo group (MD: −0.29, 95% CI −0.76 to 0.17, I 2=100%, MID: 1.4, moderate-certainty evidence; online supplementary eFigure 7 and table 2). The interaction test showed that increasing Aβ clearance had better efficacy than reducing Aβ production when compared with the placebo in MMSE (interaction p<0.000001, online supplementary eFigure 5B).
Trial sequential analysis on MMSE in nine trials was performed (online supplementary eFigure 8). We conducted the trial sequential analysis with the alpha set to 5%, power to 90%, MD to 1.4, variance to low bias and heterogeneity correction based on model variance. The diversity-adjusted required information size was 4422 participants. In total, the cumulative meta-analysis included 3438 participants in the anti-Aβ drugs group and 2772 in the placebo group. The cumulative Z-curve crossed the monitoring boundary and required information size, implying that a sufficient level of evidence was reached and no further trials may be needed.
Clinical Dementia Rating Scale Sum of Boxes
In the meta-analysis of the high-dose group versus the placebo group, no difference in CDR-SOB was observed in favour of the anti-Aβ drugs (MD: 0.25, 95% CI −0.03 to 0.52, I 2=100%, MID: 1.5, moderate -certainty evidence, online supplementary eFigure 9 and table 2). The interaction test showed that increasing Aβ clearance had better efficacy than reducing Aβ production when compared with the placebo in CDR-SOB (interaction p<0.000001, online supplementary eFigure 5C).
CSF and plasma biomarkers
The effect of anti-Aβ drugs on the plasma and CSF biomarkers of AD was presented in online supplementary eFigure 10. Of the 19 trials, 2 (10.5%) reported that there was no significant difference between groups regarding the mean change in CSF Aβ1–42 from baseline (MD: 0.02, 95% CI −0.49 to 0.52, I 2=0%, very low -certainty evidence; online supplementary eFigure 10A and table 2); similar results were found for the phosphorylated-tau (P-tau) in CSF (MD: −3.30, 95% CI −28.34 to 21.74, very low-certainty evidence; online supplementary eFigure 10B and table 2). A meta-analysis of two trials showed that the amount of total tau in CSF was decreased in drugs that inhibited Aβ aggregation vs placebo (MD: −0.07, 95% CI: −2.62 to 2.47, I 2=0%, very low-certainty evidence; online supplementary eFigure 10C and table 2).
We summaried the common anti-Aβ drug-related AEs (online supplementary eFigure 11 and eTables 2 and 3). Compared with placebo, the risk ratios of anti-Aβ drug-related AEs were the following: 2.86 for anxiety (95% CI 2.03 to 4.01, high-certainty evidence), 0.65 for cerebral artery infarct (95% CI 0.08 to 5.21, low-certainty evidence), 1.64 for depression (95% CI 1.06 to 2.53, high-certainty evidence), 1.37 for diarrhoea (95% CI 1.07 to 1.75 high-certainty evidence), 1.09 for dizziness (95% CI 0.61 to 1.94, moderate-certainty evidence), 1.32 for fall (95% CI 1.00 to 1.73, moderate-certainty evidence), 2.42 for fatigue (95% CI 1.15 to 3.96, high-certainty evidence), 0.92 for headache (95% CI 0.65 to 1.31, moderate-certainty evidence), 2.18 for nausea (95% CI 1.01 to 4.71, low-certainty evidence), 3.32 for rash (95% CI 1.97 to 5.59, high-certainty evidence), 0.65 for stroke (95% CI 0.08 to 5.21, moderate-certainty evidence), 1.97 for syncope (95% CI 1.15 to 3.37, high-certainty evidence) and 2.17 for vomit (95% CI 1.29 to 3.62, high-certainty evidence).
We also pooled the deaths serious AE rates. Compared with placebo, the risk ratios of anti-Aβ drug-related deaths was 1.24 (95% CI 0.77 to 1.99, moderate- certainty evidence; online supplementary eFigure 12 and eTable 3).
For the AEs- subgroup analysis, compared with placebo, the risk ratios of drugs of increasing Aβ clearance was 0.89 (95% CI 0.51 to 1.54, moderate-certainty evidence; online supplementary eFigure 13 and eTable 3) and drugs of reduce Aβ production was 1.90 (95% CI 1.53 to 2.37, moderate-certainty evidence; online supplementary eFigure 13 and eTable 3).
In this systematic review and meta-analysis, we focused on examining the effects of drugs that inhibit Aβ synthesis or increase Aβ clearance in patients with mild to moderate AD. We found moderate quality evidence that these drugs have no important influence on cognitive function as measured by the ADAS-cog nor on ADLs as measured by the ADCS-ADL (table 2). We found an apparent subgroup effect suggesting cognitive function improves with drugs that increase Aβ clearance but not those that reduce Aβ production. However, the credibility of the subgroup hypothesis is limited (online supplementary eFigures 2 and 5) and even if it is real, the magnitude of the effect on the ADAS-cog is about a third of the minimal important difference. Adverse effects of the drugs include anxiety, depression, diarrhoea, fatigue, rash, syncope and vomit.
Strengths and limitations of this study
The strengths of our study include the following: the comprehensive approach to the search for RCTs that found 19 RCTs with 143 participants investigating the beneficial effect of anti-Aβ drugs on the ADAS-Cog; a pre-planned subgroup analyses to explore the relationship between ADAS-Cog and four baseline factors; and use of the GRADE approach for assessing the quality of evidence.
Our meta-analysis also has several limitations. First, this study is based on study-level data but not patient-level data. Therefore, problems of missing detailed information for some patients are inevitable and could undermine the degree of internal and external credibility. Second, we could not assess the long-term effects of these anti-Aβ drugs; among the 19 eligible studies, the longest treatment duration in the large trials was 104 weeks, and most studies followed up patients for a year or less. It seems extremely implausible, however, that effects would be trivial or absent at 1 year and important benefits would emerge thereafter. Third, several recent anti-Aβ agents (eg, aducanumab, atabecestat, elenbecestat, crenezumab and bapineuzumab) are not included in our analysis due to inconsistent outcome indicators, observed data (not RCT) and results not yet published (online supplementary eTable 4). Because reported results of these new drugs are mostly negative, combined with our TSA analysis, we believe that adding data of several anti-Aβ agents will not affect the stability of results of our current meta-analysis.
Relation to previous work
At present, Aβ hypothesis is widely known as the most important pathogenesis of AD. The clearance of Aβ depends on the immune mechanism of the organism. Two meta-analyses21 22 were established to investigate the efficacy of anti-Ab immunotherapy for AD. However, these anti-Aβ drug treatments may have no significant effect and cannot reverse the cognitive deficits. Consistent with our subgroup analysis result, negative symptomatic effects in ADAS-Cog scores was found in the type of drug of intravenous immunoglobulin (online supplementary eFigure 2). Besides, one previous systematic review examined the effectiveness and safety of passive immunotherapy targeting Aβ42e11. This review, including 10 anti-Aβ42 therapies tested in clinical trials, reported that anti-Aβ42 could not prevent neuronal loss but provided significant clinical benefits to patients with AD. This article only focused on antibodies targeting Aβ42, omitting other treatments, such as small molecular compounds that reduce Aβ production and active immunotherapy anti-Aβ42. In addition, the authors did not undertake a meta-analysis.
Treatments that increase Aβ clearance, most of which are Aβ immunotherapies including active and passive immunisation strategies, reduce the burden of deposited Aβ through binding to the mid-domain of the Aβ peptide, binding to the N-terminal portion of Aβ or affecting the toxic oligomerisation of Aβe12 e14. Small molecular compounds, including avagacestat, semagacestat and verubecestat, reduce Aβ overproduction by inhibiting γ-secretase or BACE-1; lithium hampers the formation of amyloid plaques and neurofibrillary tangles via the negative regulation of GSK3βe12. Our subgroup analysis indicated that drugs that promote Aβ clearance may have an effect on cognitive performance (ADAS-Cog), but drugs that reduce Aβ production do not.
The different effects on ADAS-Cog may be due to several reasons. First, increasing Aβ clearance may be more important than reducing Aβ production to slow cognitive decline in the early stage of AD, but there is no direct evidence to support this hypothesis. Second, drugs that decrease Aβ production were more commonly associated with SAEs, resulting in patients withdrawing from the clinical trials of anti-Aβ drugs. Finally, agents that increase Aβ clearance directly bind to bound or unbound harmful Aβ peptides (Aβ40 and Aβ42), which induce neurotoxicity in neurons and then reduce amyloid aggregation and subsequent deposition. The pharmacological mechanisms of the two kinds of anti-Aβ drugs are different, and none of the drugs could significantly reduce cognitive decline in patients with mild to moderate AD from our results.
However, these findings must be interpreted cautiously since the subgroup analysis has limited credibility. When we checked against 11 criteria for assessing the credibility of a significant subgroup effect, the following findings support the hypothesis31 (online supplementary eTable 5). The characteristics were measured at baseline; the interaction test results were highly significant; the subgroup findings seemed to be consistent across related outcomes (such as ADAS-Cog, MMSE and CDR-SOB); and in terms of biological rationale, current studies pay particular attention to microglia and some proteases of cerebral Aβ clearance, which was considered as a potential therapeutic target.6 7 In contrast, the subgroup analysis was not pre-specified; no individual study mentioned that the direction of the subgroup effect specified a priori; made inferences about different effect sizes in different groups (one that compared drugs increasing Aβ clearance with placebo and another that compared drugs reducing Aβ production with placebo) on the basis of between-study differences; multiple metaregression could not be performed due to the limited number of included studies; the apparent benefit in ADAS-Cog with increasing Aβ clearance was below the MID; and no additional similar exploratory subgroup analysis of large trials was performed. It cannot be determined for the criteria ‘was the subgroup effect one of a small number of hypothesised effects tested’ and ‘is the interaction consistent across studies’ due to insufficient information within studies. We thus conclude that the subgroup hypothesis has limited credibility.
Implications for clinicians and pharmaceutical companies
The findings from this study emphasise the role of meta-analyses in examining focused clinical questions. After spending millions of dollars and more than 10 years on drug development and clinical trials, none of the anti-Aβ drugs have shown therapeutic benefits, such as halting or reversing cognitive decline or the ability of patients to perform daily activities. Additional large-scale and well-designed randomised and placebo -controlled trials will be necessary to explore the benefit of a certain type of drugs that increase Aβ clearance as treatment in populations with mild to moderate AD. In general, further research on Aβ clearance pathways, including non-enzymatic and enzymatic pathways, should be carefully explored for AD research.
AEs were more common with drugs that reducing Aβ production than increasing Aβ clearance (online supplementary eFigure 13-14).We speculate that small molecular drugs reduce Aβ production by inhibiting γ-secretase, β-secretase or other enzymes involved in Aβ production and may hamper the production of some proteins that are necessary for normal physiological processes. Further in vitro and in vivo experiments are needed to prove this hypothesis.
In conclusion, in patients with mild to moderate AD, current evidence showed that anti-Aβ interventions is unlikely to have an important impact on slowing cognitive or functional decline. Although the subgroup analysis suggested that increasing the Aβ clearance subtype could be associated with better ADAS-Cog, MMSE and CDR-SOB, the subgroup analysis has limited credibility and a benefit from drugs that increase clearance, if real, is very small.
LL, XZ, SW and CT contributed equally.
Contributors Study concept and design: NX, CT and LL. Acquisition of data: XZ and GY. Analysis and interpretation of data: LL, JZ, SG, HW and MX. Drafting of the manuscript: LL, XZ, SW and YZ. Critical revision of the manuscript for important intellectual content: GG, NX and CT. All authors critically revised successive drafts of the paper and approved the final version.
Funding This study was supported by the Youth Scientific Research Training Project of GZUCM (2019QNPY02), the National Natural Science Foundation of China (81873375), the Young Top Talent Project of Scientific and Technological Innovation in Special Support Plan for Training High-level Talents in Guangdong (2017TQ04R627), the Key Program of the First-Class Discipline of Guangzhou University of Chinese Medicine (XK2018001), the Key Laboratory Program of Universities in Guangdong Province (2018KSYS006) and the Natural Science Foundation of China(81904275).
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information. Data are available to qualified investigators on request to the corresponding author.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.