Effects of Peroxisome Proliferator-Activated Receptor-Gamma Agonists on Cognitive Function: A Systematic Review and Meta-Analysis

Diabetes mellitus (DM) is known to be a risk factor for dementia, especially in the elderly population, and close associations between diabetes and Alzheimer disease (AD) have been determined. Peroxisome proliferator-activated receptor-gamma (PPAR-γ) agonists are insulin-sensitising drugs. In addition to their anti-diabetic properties, their effectiveness in preventing and decreasing cognitive impairment are the most recent characteristics that have been studied. For this study, we conducted a systematic review and meta-analysis to critically analyse and evaluate the existing data on the effects of PPAR-γ agonist therapy on the cognitive status of patients. For this purpose, we first analysed both early intervention and later treatment with PPAR-γ agonists, according to the disease status. The involved studies indicated that early PPAR-γ agonist intervention is beneficial for patients and that high-dose PPAR-γ therapy may have a better clinical effect, especially in reversing the effects of cognitive impairment. Furthermore, the efficacy of pioglitazone (PIO) seems to be promising, particularly for patients with comorbid diabetes. PIO presented a better clinical curative effect and safety, compared with rosiglitazone (RSG). Thus, PPAR-γ agonists play an important role in the inflammatory response of AD or DM patients, and clinical therapeutics should focus more on relevant metabolic indices.


Introduction
Diabetes mellitus, which has a troublingly high prevalence, affects 463 million individuals throughout the world (International Diabetes Federation, 2019), and over one-quarter of the population aged ≥ 65 years. Diabetes mellitus (DM) is known to be a risk factor for dementia [1][2][3], especially in the elderly population. It has been reported to lead to inferior cognitive performance, when compared to age-matched healthy controls (HCs) [4]. Several meta-analyses have indicated that individuals with diabetes have a 73% increased risk of developing dementia and a 56% risk of developing Alzheimer disease [2], and research has suggested that type 2 diabetes mellitus (T2DM) results in a more rapid rate of cognitive decline than that typically associated with natural aging [5][6][7]. Furthermore, patients with pre-diabetes also have a higher risk for all-cause dementia and Alzheimer disease (AD) [3]. Diabetes is associated with AD through common pathophysiological mechanisms, such as increased levels of inflammatory indices, amyloid β-protein (Aβ) deposition, and increased oxidative stress [8]. Hence, aggressive management is necessary to prevent AD progression. Several studies have reported that anti-diabetic agents could potentially treat cognitive symptoms and modulate disease progression and cognitive decline in mild cognitive impairment (MCI)/AD [9]. Peroxisome proliferator-activated receptor-gamma (PPAR-γ) agonists are insulin-sensitising drugs designed for diabetes patients with insulin resistance; furthermore, they can regulate several cellular processes, such as Aβ degradation and diabetes patients with insulin resistance; furthermore, they can regulate several c processes, such as Aβ degradation and the anti-inflammatory response [10,11]. In years, treatment of animal models of AD or MCI with thiazolidinediones, a PPARnists, demonstrated an improvement in memory performance and a reduction in loid burden and inflammation [12][13][14][15]. Several observational studies and rando controlled trials (RCTs) have also reported the effectiveness of PPAR-γ agonists i venting and decreasing cognitive impairment. However, the reported results still p some considerable variabilities, and the effects of existing PPAR-γ agonists, in ter preventing cognitive impairment, have not yet been summarized and analysed in vational research.
Hence, the primary objective of this study was to conduct a systematic revie meta-analysis to critically analyse and evaluate data on the effects of PPAR-γ agonis apy on cognitive status. To this end, we first analysed both early intervention and treatment with PPAR-γ agonists, according to the disease status. In addition, we co hensively evaluated the efficacy and safety of PPAR-γ agonists in the treatment of mild to moderate AD, and AD based on clinical trials.

Search Strategy
We performed a literature search in July 2022 on the Elsevier, EMBASE, Web ence, and PubMed databases. The following search terms were used: (1) "Cogniti pairment" or "Dementia" or "Alzheimer Disease"; (2) "PPAR-γ agonists/Peroxisom liferator-activated receptor gamma agonists" or "Thiazolidinedione" or "Rosiglitaz "pioglitazone"; (3) "the cohort studies" or "the Case-Control studies" or "the Rando Controlled Trial studies". In addition, the references of the retrieved papers and reviews were reviewed. The flow diagram of the search strategy is presented in Fig   Figure 1. Flowchart of the meta-analysis study selection process.

Study Criteria
The inclusion criteria for studies were as follows: (1) cohort studies or case-control studies providing data on PPAR-γ (peroxisome proliferator-activated receptor-gamma) agonist treatments ("Treatment dose," "Treatment duration," "Effect on cognition, such as HR (hazard ratio),"); (2) validated diagnosis of "Cognitive impairment" or "Dementia" or "Alzheimer Disease"; (3) studies that provided information about PPAR-γ agonist treatments in cognitive impairment patients; and (4) articles that reported a clear comparison of PPAR-γ agonist treatments versus no PPAR-γ agonist treatments in population controls with a direct effect on cognitive or metabolic responses.
The exclusion criteria were as follows: (1) duplicate studies; (2) studies such as systemic reviews, meta-analyses, and comments; and (3) studies of PPAR-γ agonist treatments without detailed research data concerning clinical responses.

Data Extraction
The data extracted from each study included the first author's name, the publication year, the country of study origin, the number of patients, median age, the neuropsychological response: Alzheimer Disease Assessment Scale Cognitive Score (ADAS-COG), Clinical Dementia Rating sum of boxes (CDR-SB), Digital symbol test (DST), Mini-Mental State Examination (MMSE), Rey Auditory Verbal Learning Test (RAVLT), metabolic response:fasting plasma glucose (FPG), homeostasis model assessment (HOMA %), Insulin, tumour necrosis factor (TNF-α), interleukin-6 (IL-6), C-reactive protein (CRP), adverse events (AEs), and HR. If a study did not clearly mention any of the above key points, it had not performed the required methods. Two of the authors (Hongfei Zhong and Qiuyun Tu) independently reviewed the selected studies and extracted the data. Discrepancies were resolved by discussion.

Statistical Analysis
The data were abstracted and analysed using Stata (version 12), in order to make the outcomes more comprehensive. The binary variable outcomes were the AEs (anaemia, peripheral oedema, bone fractures, cardiac failure, diarrhoea, dizziness, hepatic disorders, headache, hyperlipidaemia, hypoglycaemia, insomnia, muscle pain, nasopharyngitis, nausea), and the data are expressed as a RR (risk ratio) with 95% CI (confidence interval). The effect was estimated using a random effects model. Other continuous variable outcomes were the rating scale (ADAS-COG, CDR-SB, DST, MMSE, and RAVLT), and the metabolic response (FPG, HOMA%, Insulin, TNF-α, IL-6, CRP). Data are expressed as the SMD (standardized mean difference) with 95% CI. In addition, we performed a random-/fixed-effect meta-analysis using maximally adjusted HRs with 95% CI for the observational studies. When combining studies, the random effects model was used to account for study heterogeneity. We used the Q statistic and I 2 tests to evaluate the heterogeneity. Low, moderate and high heterogeneities were represented by thresholds of <25%, 25-75%, and >75%, respectively. p ≤ 0.05 was considered significant in all statistical tests.

Data Analysis
The binary variable outcomes were the incidence of the AEs (anaemia, peripheral oedema, bone fractures, cardiac failure, diarrhoea, dizziness, hepatic disorders, headache, hyperlipidaemia, hypoglycaemia, insomnia, muscle pain, nasopharyngitis, nausea). In addition, the data of dementia risk are expressed as the hazard ratio (HR) or odds ratio (OR) with 95% confidence interval (CI); the estimation of the effect was performed using a random effects model.

Effect of PPAR-γ Agonists on Cognition Based on the Observational Studies
Nine observational studies were included (two case-control studies and seven cohort studies) in the meta-analysis, where the HR for observational studies was equal to 0.91 (95% CI = [0.88, 0.94], I 2 = 39.5%) in the random-effect analysis (Figure 2A

Effect of PPAR-γ Agonists on Cognition Based on Clinical Trials
In 13 double-blind randomized controlled studies, the effect of PPAR-γ agonists versus placebo on cognitive performance was evaluated in 5102 cognitive impairment patients with and without DM.

Neuropsychological Outcomes/Neuropsychological Scales
Neuropsychological outcomes were measured in all studies, mainly including ADAS-Cog, CDR-SB, DST, MMSE, and RAVLT. Details of the outcomes are listed in Table  3; due to the complexity of the source of the scales, the data are expressed as the SMD (standardized mean difference) with 95% CI (confidence interval). Furthermore, the subgroup analysis considered drug, dose, disease status, treatment duration, and region.

Effect of PPAR-γ Agonists on Cognition Based on Clinical Trials
In 13 double-blind randomized controlled studies, the effect of PPAR-γ agonists versus placebo on cognitive performance was evaluated in 5102 cognitive impairment patients with and without DM.

Neuropsychological Outcomes/Neuropsychological Scales
Neuropsychological outcomes were measured in all studies, mainly including ADAS-Cog, CDR-SB, DST, MMSE, and RAVLT. Details of the outcomes are listed in Table 3; due to the complexity of the source of the scales, the data are expressed as the SMD (standardized mean difference) with 95% CI (confidence interval). Furthermore, the sub-group analysis considered drug, dose, disease status, treatment duration, and region.

Metabolic Outcomes
We investigated the glycometabolism, lipid metabolism, and inflammatory response in MCI or AD patients after PPAR-γ agonist treatments, data are expressed as the SMD (standardized mean difference) with 95% CI, and the metabolic outcomes in studies have the same unit (FPG: mg/dL; HOMA%: -; insulin: µU/mL; TNF-α: pg/mL; IL-6: pg/mL; CRP: mg/L), as detailed in Table 3. FPG was decreased after PPAR-γ agonist treatments, but the HOMA % and insulin were not significantly changed. As for the inflammatory response, three related indices were considered in our study, and we found that TNF-α was significantly decreased, while IL-6 and CRP were not changed.

AEs
We investigated all types of AEs. A total of 14 subjects were included in the study (anaemia, peripheral oedema, bone fractures, cardiac failure, diarrhoea, dizziness, hepatic disorders, headache, hyperlipidaemia, hypoglycaemia, insomnia, muscle pain, nasopharyngitis, and nausea), as shown in

Discussion
The published research on PPAR-γ agonist therapy has mainly focused only on RCT clinical studies. Liu and Cheng et al. have taken PPAR-γ agonists as sufficient evidence to treat cognitive impairment patients [38,39]. However, various data were not comprehensively assessed, such as treatment dosing, duration, neuropsychological scales, and metabolism. In recent years, more studies have reported updated research on PPAR-γ agonist treatments. In 2021, Burns et al. [16] conducted a phase 3, randomized, doubleblind, placebo-controlled trial which enrolled 3494 participants. Cohort and case-control studies have also been published in recent years. Khan et al. [40] and Jojo et al. [41] have updated the current progress on PPAR-γ agonists as an emerging therapeutic approach for the treatment of Alzheimer's disease. Hence, updating related meta-analysis research is necessary.
In this systematic review, we found that the available evidence provides some support for PPAR-γ agonists having a protective effect against dementia in individuals who are taking them for the management of diabetes. Furthermore, the considered studies have deeply discussed the clinical efficacy of PPAR-γ agonists versus placebo in clinical trials, including the dose-response effect, treatment duration, neuropsychological scales, metabolic outcomes, and full AE analysis.
Observational studies have shown that PPAR-γ agonists have a protective effect against cognitive decline in diabetic patients (HR = 0.91, 95% CI = [0.88, 0.94] and I 2 = 39.5%), where the heterogeneity of the data was acceptable. These results suggest that early PPAR-γ agonist drug intervention in DM patients has potential benefits. Compared with RSG, PIO presented a better prophylactic effect (Figure 2A). At present, there are two major glitazones: rosiglitazone and pioglitazone. Pioglitazone significantly reduces the neuroinflammation response and cerebral oxidative stress and, so, might play a neuroprotective role associated with improvements in inflammation-related neuropathy and insulin signalling pathways. Additionally, Yang et al. demonstrated that treatment with pioglitazone (20 mg/kg; intragastric administration) ameliorated Aβ42 deposition in the insulin-resistant (IR) rat hippocampus by increasing insulin-degrading enzymes (IDE) and PPAR-γ expression, as well as ameliorating Aβ accumulation via the AKT/GSK3β signalling pathway [42], and Aβ accumulation was measured by hippocampus immunohistochemistry. Pioglitazone not only recovered the memory and cognitive deficits, but also ameliorated Aβ deposition. The involved studies indicated that high-dose PPAR-γ agonists may play protective roles in DM patients. Furthermore, DM patients less than 60 have a lower risk of dementia (HR = 0.72, 95% CI = [0.55, 0.94]), suggesting that early PPAR-γ agonist intervention in DM patients has potential benefits. According to the geographical distribution sub-group, Asian DM patients may have a lower risk of dementia with early PPAR-γ agonist intervention.
The established evidence indicated a considerable overlap in the putative pathophysiological mechanisms for DM and cognitive impairment and dementia [43],such as increased levels of inflammatory indices, amyloid β-protein (Aβ) deposition, and increased oxidative stress. The epidemiological trends for dementia are very similar to those observed in DM [44]. Epidemiological studies have established an increased risk of dementia among individuals with diabetes mellitus [45], and type 2 diabetes mellitus (T2DM) might also be a potential risk for MCI progressing into AD, through the induction of oxidative and inflammatory stress and vascular dysfunction [46]. Studies have also shown a positive association between DM and mild cognitive impairment (MCI), and an accelerated progression from MCI to dementia in DM [47]. Hence, early intervention by using hypoglycaemic agents may be beneficial for reducing the risk of dementia, as shown in Figure 4. reflects early arithmetic ability and, so, in early cognitive impairment patients, early intervention with PPAR-γ agonists may effectively better reverse the loss in arithmetic cognitive ability. Unfortunately, due to the small number of studies reporting MMSE and RAVLT scores, we found that there was no apparent effect of PPAR-γ agonists on MMSE (SMD = 0.37, 95% CI = [−0.09, 0.83], I 2 = 0%) or RAVLT (SMD = −0.15, 95% CI = [−0.44, 0.13], I 2 = 0%) scores. .8%) were significantly changed after PPAR-γ agonist treatment. The most significant metabolic change was the inflammatory response. In DM and AD mouse models, chronic treatment of pioglitazone (15 and 30 mg/kg) significantly attenuated TNF-α and IL-6, compared to an Aβ-treated group (p < 0.05), the increase in IL-6 and TNF-α expression which indicates the involvement of inflammatory activity, pioglitazone significantly reduced the neuroinflammation response (e.g., reducing the inflammatory cytokines TNF-a, IL-6, IL-8, and so on) and cerebral oxidative stress and, thus, might play a neuroprotective role associated with improvements in inflammation-related neuropathy and insulin signalling pathways [48,49]. Studies have also reported that rosiglitazone treatment may decrease the inflammatory markers TNF-α and IL-6a [15,50,51]. Furthermore, as investigated in the metaanalysis, TNF-α decreased after PPAR-γ agonist treatment (SMD= −0.60, 95%CI = [−1.04, −0.15], I 2 = 19.8%). The inflammatory response may be crucial in the AD progression of DM patients. The involved studies indicated that the anti-inflammatory properties of PPAR-γ agonists may have potential roles in the treatment of AD. Furthermore, evidence from various studies has suggested that insulin sensitizers (TZDs) are key therapeutic targets in MCI and AD patients, with an associated Aβ degradation and anti-inflammatory response [10,11]. Hence, we collected the RCTs in the above disease which were treated with PPAR-γ agonists (Figure 4), and analysed the clinical effects from the perspective of drug treatment. We enrolled all outcomes, including neuropsychological outcomes, metabolic outcomes, and AEs.
We also discussed the AEs during PPAR-γ agonist treatment, and found that the risk of suffering from AEs increased with the dose and duration of medication (Table 4). Other AEs, such as anaemia and peripheral oedema, presented the same tendency. Previous articles have reported that PIO treatment may lead to a higher risk of AEs; in their subgroup analyses, a borderline result was detected in pioglitazone-versus placebo-treated subjects (RR =7. 43 14.50], I 2 = 6.8%). Clinicians must pay attention to these safety events.
We must note that there were some limitations of our study. First, several clinical studies on PPAR-γ agonist therapy are still ongoing [52][53][54], and we could not enrol these yet-unpublished studies. Second, some of the studies that were included failed to directly present concrete data in the text, such as changes in standard deviation and mean values from baseline; in order to ensure the reliability of the data, we bypassed the second transformation entirely, which may have led to a reduced amount of data. Third, as for the AE data, a large number of data were included; however, for the neuropsychological outcomes, the data are still limited. We expect that more RCTs will help us to expand the sample size and make a more accurate conclusion in the future. Fourth, in the observational studies, the time span of the drug intervention was huge and the data were limited, from 45-90 days to 11 years and, thus, a sub-group analysis of the intervention time could not be performed. Due to this, further studies using larger samples and detailed intervening measures are required to provide clearer results.

Conclusions
This systematic review and meta-analysis study indicated the beneficial effects of PPAR-γ agonists on the improvement of cognitive function in cross-sectional and cohort studies, as well as RCTs. We divided these results into two parts: PPAR-γ agonist intervention and therapy. Cognitive impairment is a progressive disease, with patients suffering progressive deterioration of cognitive and functional skills and having difficulties with memory, language, problem-solving, and other thinking skills. Such diseases include MCI, mild to moderate AD, and AD. Continuous monitoring is essential during PPAR-γ agonist treatment. The enrolled studies indicated that early PPAR-γ agonist intervention is beneficial for patients, and that high-dose PPAR-γ therapy may have a better clinical effect, especially in the reversing the effects of cognitive impairment. Furthermore, the efficacy of PIO seems to be promising, particularly for patients with comorbid diabetes. PIO presented a better clinical curative effect and safety, compared with RSG, and it deserves more well-designed trials with large sample sizes in the future. PPAR-γ agonists play an important role in the inflammatory response of AD or DM patients, and clinical therapeutics should focus more on relevant metabolic indices.  Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.
Data Availability Statement: The data sets used and analysed during the current study are available from the corresponding author.