A pooled CSA intervention targeting community-dwelling older adults without dementia slowed cognitive decline and improved global cognition and multiple cognitive domains. Notably, improvements were noted in immediate memory, one of the earliest cognitive domains implicated in early cognitive decline. Exploratory subgroup analyses showed that the beneficial effects of the intervention were higher in MCI across all cognitive domains, with a medium effect size of 0.42 in global cognition and the visuospatial construction domain. Furthermore, involvement in a higher number of CSAs was associated with greater improvements in global cognition and several cognitive domains. On the other hand, involvement in the CSAs did not elicit improvements in depressive symptoms, regardless of baseline cognitive status and dose-response effect. In all, these findings suggest the high specificity and malleability of immediate memory in response to CSAs. Furthermore, greater effects of CSAs with MCI subgroup highlighted MCI as a potential interventional target population for CSAs.
A notable contribution of this study to the extant literature is the duration of follow-up and a control group. Till date, there are two existing main lines of evidence on the positive long-term effects of CSAs. First, several epidemiological studies have shown that participation in CSAs was associated with improved cognitive outcomes at approximately 5-year follow-up, but these studies were methodologically limited by the absence of control groups30–32. Second, in contrast to the overwhelmingly positive evidence from epidemiological studies on improving cognitive outcomes, RCTs on CSA interventions still present mostly mixed evidence11,12,14,33. They suffer from short follow-up periods and small sample sizes, raising concerns regarding the sustainability of the interventions. In the contrary, in our study conducted with pooled RCTs on CSAs in the cohort with mean follow-up of approximately five years, comparisons were made with a control group who had no prior involvements in any structured CSAs. We have thus addressed the limitations of short to intermediate-term follow-ups in prior studies, which also lacked control groups. Another noteworthy point is that not only did we show the positive effects of the CSAs on cognition, our effect size of 0.42 for global cognition in the MCI sub-group provide further evidence to the effect sizes summarized in a recent meta-analysis on RCTs on CSAs and total MMSE score13, extending the applicability to a real-world setting. Thus, for the first time, we showed pilot data on the long-term sustainability of CSAs’ effects on improving cognitive functions over five years in a longitudinal and naturalistic setting. To the best of our knowledge, no other studies involving interventions comprising multiple CSAs, of similar large sample size, follow-up duration, study population, and study setting have been reported.
Next, we interpret our findings in the context of previous NPI RCTs on cognitive outcomes. The FINGER trial showed overwhelmingly positive evidence of improvements in global and several domains of cognition following NPIs targeting traditional risk factors3. In contrast, the MAPT, preDIVA, LipiDiDiet, and ACTIVE trials4–6,34, despite similarly incorporating NPIs and conducted over relatively long follow-up periods, did not result in cognitive improvements, and had small effect sizes. Compared to the FINGER trial’s effect size of 0.12 for the combined neurocognitive test battery, the improvement of 0.102 in total MMSE in our total sample is strikingly similar. Notably, we extended the literature by showing pilot data on an effect size of 0.42 for the improvement in global cognition in the MCI sub-group upon undergoing CSA intervention. Furthermore, it is noteworthy that the relatively larger effect sizes detected in this study should be considered in the context of a less intensive nature of our CSA intervention. This is in stark contrast with previous studies, which involved intensive interventions modalities, such as regular exercises and multiple follow-ups with physicians (4). Our CSA intervention also improved immediate memory, a cognitive domain tapping on lower cognitive load, with an effect size of approximately 0.2 in the total sample and both the healthy aging and MCI subgroups. This phenomenon might be attributed to the distinct advantages of CSAs, in that interventions with solely CSAs could be more efficacious, compared to a mix of intervention modalities targeting the traditional risk factors of dementia4–6,34. Concurring with previous studies, compared to controls, cognitive functions tapping on higher cognitive load, such as executive functions, were not improved in our study. We did not follow up with the participants on whether they continued pursuing the CSAs at the end of participation in the RCTs. With this caveat, by showing improved cognition after five years, we showed the sustained beneficial effects conferred by the CSAs, independent of continued pursuance of CSAs upon conclusion of the facilitated sessions.
A primary novelty of this study was the subject recruitment from the community rather than clinical settings and the continuous follow-up in the same setting, providing a real-world context. This communal setting also allowed us to examine the naturalistic effect of CSAs with both cognitively healthy and mildly impaired community-dwelling older adult populations in a single study. Compared to cognitively healthy participants, those with MCI had higher gains in global cognition and all cognitive domains. This finding lends empirical evidence to the hypothesis postulating that MCI is a pre-clinical stage at which the older adults can still learn new skills35, and likely reap higher benefit than cognitive healthy. Similarly, a behavioural trial conducted with African-Americans also showed delayed cognitive decline in individuals with MCI upon undergoing intervention36. Concurring with evidence from prior studies6,37, we postulate that the CSAs were most effective with MCI. For cognitive healthy older adults, there could either be a ceiling effect on MMSE, or they require more extended, more intensive and/or interventions other than CSAs to achieve similar effect sizes of cognitive improvement as MCI. Additionally, with higher involvement in the CSAs, there were also higher effect sizes of improvements in global cognition and several cognitive domains. Hence, these showed pilot findings suggesting that the potential greater benefits of participation in a higher number of CSAs in a naturalistic and real-world setting. Coupled with no adverse event related to either the MAP or HEP intervention was reported, the CSA intervention was both safe and feasible.
Limitations & Future Directions
This study has several limitations. The main limitation was the potentially underpowered sample size, which might have rendered statistical insignificance, even with improvements with medium effect sizes, i.e. 0.42, detected in three cognitive outcomes. However, interpreting results solely based on p-values have been de-emphasized, as it informs us statistical significance without providing any useful information on the clinical effects of treatments, and studies have started to report the clinical effects based on effect sizes to inform clinical significance. In our study, we have presented both statistics for readers’ interpretations, and highlighted the medium effect sized improvements in multiple cognitive domains. There was also some degree of heterogeneity in the pooled RCTs. However, the heterogeneity reflects the characteristics of a real-world and naturalistic study6. We also could not exclude the possibility that the participants could have been involved in other unstructured CSAs outside the structured CSAs performed in the study. Since this information was unavailable, it leaves the plausibility of residual confounding. As this issue is an inherent limitation of a naturalistic, pragmatic and real-world study such as ours, it could be mitigated in future studies, such as employing an RCT design for the total study sample. Similarly, potential confounders, such as physical activity and social networks were not assessed in the main cohort and thus were not controlled for in this study. We did not have data for the attendance and adherence rates as covariates, and it was not in our study protocols to follow-up with the participants on whether they continued pursuing the CSAs after the conclusions of the RCTs. Thus, future studies should collect data on and examine continual pursuance and intervention adherence. It is worth noting that had most of the participants in our study continued the CSAs beyond the facilitated sessions, the effect sizes for cognitive improvements could have been even larger than currently observed, further supportive of our findings. Lastly, our sample population was recruited from a single study site, and thus may not be representative of the whole population, warranting future validations in larger and more representative study population.
Strengths of the study
This study represents a significant advancement in the literature on CSA intervention and cognition and depressive symptoms in several aspects. First, all previous studies on CSAs and cognition either employed an epidemiological design with up to five-year follow-up without control groups, or a short-term RCT design without long-term follow-up. Even with an RCT, most studies on CSA did not have long-term follow-up period beyond one year. The previous RCTs were also limited by having small sample sizes (N<100). In contrary, our study contributed to the literature by having conducted the largest and the longest RCTs examining different CSAs on cognition nested within a longitudinal cohort, further incorporating depressive symptom as the secondary outcome. Second, with our a priori inclusions of two distinct neuropsychiatric outcomes, cognition and depressive symptoms, we holistically examined CSAs’ multi-faceted effects, revealing differential effects of the intervention on the measures. Third, the cognitive status of the participants was diagnosed during our team’s monthly consensus panel meetings, using the well-established clinical diagnostic guidelines, lending robustness to our study and derived findings. Conversely, many previous studies utilized MMSE cut-off scores to derive cognitive status, likely introducing inaccuracies in diagnoses25,38. Recruitment on community-dwelling older adults at the pre-clinical dementia stage allowed us to stratify our sample into clinically defined cognitively healthy and MCI sub-groups, showing the superiority of CSAs with MCI over cognitively healthy older adults. Fourth, with multiple CSA interventions, study participants could be involved in more than one CSA, allowing us to examine dose-response effects. Fifth, being conducted in an RCT setting, the CSAs were formally and structurally conducted, with fixed intervention durations, pre-planned programs for each session, and with frequencies of the interventions objectively quantified. This study design mitigates a core limitation present in previous epidemiological studies with CSAs, in which the study participants self-reported information on CSA participation. These previous subjective reports could have introduced recall bias39 and thus compromised previous findings. Lastly, apart from being conducted as a pilot study, a distinctive feature of our study is the unique model and setting in which we employed to recruit participants. Many previous non-pharmacological intervention RCTs on cognition were either recruited from or based entirely in clinical settings. With our research centre located inside a shopping mall, we recruited solely community-dwelling older adults. Hence, this unique setting allowed us to conduct interventions solely in the community and for the community, enhancing the generalizability and external validity of our findings4. Embedded within this pragmatic and naturalistic setting, the Hawthorne effect4,40 could have been minimized with our real-world data.