Our MR analyses first investigated the causal association between genetically predicted early AMD with the risk of AD. The results from the major estimation method of MR analyses indicated a causal role of early AMD in AD onset, suggesting that early AMD may play a protective role in the onset of AD. The reverse–direction MR analysis also revealed no evidence of genetic liability to AD being related to early AMD.
Our MR analyses demonstrated that early AMD could be a protective factor against AD, which is consistent with the observation that persons with low cognitive function scores were more likely to have early AMD than persons with higher scores.35, 36 However, the influence of confounders may have led to bias in these observational estimates of association. Age is one of the potential confounders, with a positive association with AMD and AD. Jiang et al.37 conducted a two-sample bidirectional MR study recently aiming to evaluate the causal relationship between advanced AMD (16,144 advanced AMD cases, 17,832controls) and AD in discovery data set (late-onset AD, 21,982 cases and 41,944 controls) and validation data set (71,880 clinically diagnosed AD/AD-by-proxy cases and 383,378 control). Interestingly, they found no evidence to support a bidirectional causal association between advanced AMD and AD and then concluded that the MR study decreased the probability of a clinically relevant relationship between the two neurodegenerative diseases and that the associations observed in epidemiological studies should not be considered causal. However, we took it with a grain of salt. Therefore, we performed an MR study to further explore the causal association between AMD and AD. As expected, Our MR study confirmed a moderate causal effect of early AMD on AD development after avoiding these age–related risk factors. This may give us a little prompt message that we should account for the influence of phenotype (IVW, p = 0.032, OR = 0.945, 95% CI = 0.898–0.995 per SD–increase). Contrary to previous studies that adopted a cross-sectional study design, some longitudinal studies investigated the association between AD and AMD. A cohort study with 4,097 patients in the eye disease group and 20,475 in the control group depicted that general AMD was associated with an increased risk of AD (adjusted HR = 36.94, 95% CI = 4.62–295.46).9 Our study focused on genetic risk factors and demonstrated some shared genes between AMD and AD. However, the same genes may play a different role in AMD and AD, for example, the famous gene ApoE, which has three alleles (ε2, ε3, and ε4). Individuals carrying the APOE-ε4 allele seem to have a lower risk for AMD but the strongest thus far identified risk factor for AD. This may be one of the reasons causing the result that we obtained.38
In reverse–direction MR analysis, we could not determine a causal association between AD and AMD (IVW, p = 0.432, OR = 0.974, 95% CI = 0.911–1.041 per SD–increase). The MR analysis result was consistent with the above in validation (IVW, p = 0.760, OR = 1.034, 95% CI = 0.835–1.279 per SD–increase; MR–Egger regression, p = 0.573, OR = 0.861, 95% CI = 0.514-1–442 per SD–increase; WMed, p = 0.486, OR = 0.899, 95% CI = 0.666–1.213 per SD–increase; WMod, p = 0.560, OR = 0.908, 95% CI = 0.659–1.252 per SD–increase). A cross–sectional study that included 592 AD patients aged 50 and above showed AD was not associated with AMD.2 A similar conclusion was reported in a population–based study with 2,088 participants aged 69 to 97 years, which demonstrated that there was no association of AD with early AMD.35
Interestingly, MR–Egger analysis suggests a pleiotropic effect between AD and AMD (p = 0.009, OR = 0.855, 95% CI = 0.763–0.958 per SD–increase), suggested by MR–PRESSO, 2 pleiotropic SNP—rs1800978, encoding ABCA1 and rs61679753, and encoding TOMM40 might influence the result. ATP–binding cassette transporter A1 (ABCA1) is one of several proteins involved in cholesterol homeostasis. ABCA1 gene variation has been demonstrated to contribute to AMD susceptibility.39–42 A GWAS study found that TOMM40 was associated with advanced AMD (p = 3.1 × 10− 6) and combined replication samples (p = 8.4 × 10− 8) with European ancestry.43 This variant has been identified, showing an association with early AMD in a GWAS meta-analysis of the European population (p = 1.1 × 10− 6), and the signal remained in a secondary meta-analysis combining two Singapore-based Asian cohorts (p = 3.2 × 10− 6).44
However, previous observational studies investigated the influence of neurological diagnoses such as AD on AMD but with inconclusive results. The potential causality between AD and AMD remains unclear owing to bias or various confounders inherent to observational studies. To date, no RCT study has been performed to assess the effect of AD on AMD. Our MR study adds to the existing body of knowledge by confirming the causal effect of AD on AMD, which is informative. The causal association between AMD and AD has implications for understanding disease pathogenesis, referral, and treatments and can make the most effective use of scarce resources.
The major strength of our study is its two–sample MR design, which could circumvent the limitations of observational studies with measurement errors and residual confounding. It can minimize residual confounding and reverse causality and genetic validation of the wide range of AD, which can afford high–quality evidence. However, our study has a few limitations that should be addressed. First, the estimates might be biased toward the observational associations if the exposure and outcome data came from the same sample.45 Second, residual pleiotropy might remain, despite the range of sensitivity analyses (i.e., WM, MR–PRESSO, and MR–Egger intercept) conducted to explore and account for pleiotropy. Third, we could not distinguish the effects of all different kinds of AD and AMD and thus only examined general AD, early AMD and AMD progression. A further limitation was that our study was restricted to participants of European ancestry; however, the observational and causal effects of AMD on AD may differ in other ethnic groups.
Therefore, the causal effect in our principal MR analysis (IVW) was robust and unbiased, which was evidenced by neither the MR–Egger intercept test nor the MR–PRESSO analysis. Conclusively, our primary MR and sensitivity analyses, with IVW, WMed, and WMod methods, consistently establish the existence of a causal effect of suffering early AMD on the decreased risk of AD occurrence.