Alzheimer’s disease (AD) remains one of the most prevalent neurological disorders worldwide, accounting for 60%–70% of all dementia cases ¹ . In 2021, an estimated 57 million people were living with dementia, with this figure being projected to nearly double every 20 years ² .

Despite significant investment in basic and translational research, treatments that effectively and safely act on the evolution of the disease are currently lacking. Available drugs such as N-methyl-D-aspartate receptor antagonists (memantine) and cholinesterase inhibitors (donepezil, galantamine, rivastigmine) are largely symptomatic and do not suppress disease progression. Since 2021, the first anti-amyloid antibodies (aducanumab, lecanemab, donanemab) have been approved to target amyloid-β (Aβ) plaques, a pathological hallmark of AD ³ . These are considered disease-modifying therapies as they can slow cognitive decline in patients with mild cognitive impairment or mild dementia due to AD, but are associated with significant risks, including brain swelling and bleeding ⁴ .

The heterogeneity of patients and absence of reliable early diagnostic biomarkers hinder progress, as demonstrated by age-related shifts in Aβ positivity (Young and Mormino, 2022). Adding to the complexity, AD is a highly multifaceted condition, varying in age of onset, genetic risk factors, pathological processes and progression patterns, as well as the presence of comorbidities (Duara and Barker, 2022). This diversity makes it unlikely that treatments aimed at a single target will be effective across the entire patient population.

Over the past 15 years, new hypotheses beyond the amyloid cascade have emerged to explain the complex etiopathology of AD (Zhang et al., 2024). Several are interrelated, with dysfunctional metabolism, chronic inflammation, and environmental factors playing a critical role, along with ageing and genetics. These hypotheses include: (i) systemic inflammation caused by pathogen infections (Seaks and Wilcock, 2020; Bruno et al., 2023; Brown and Heneka, 2024), (ii) the impact of long-term exposure to environmental pollutants (Dhapola et al., 2024), and (iii) the role of microbiota and gut dysbiosis in the induction of neuroinflammation through the gut-brain axis (GBA) (Logan et al., 2023; Seo and Holtzman, 2024; Dhanawat et al., 2025).

Traditional nonclinical models of AD have relied on both in vivo and in vitro models that replicate key disease hallmarks. Transgenic mice that express genetic variants identified in familial AD (Zhong et al., 2024) can develop amyloid plaques, neurofibrillary tangles, gliosis, and mild cognitive deficits. Yet, even with amyloid accumulation, they frequently fail to show significant neuronal loss (Sanchez-Varo et al., 2022). Moreover, animal models do not replicate disease pathogenesis as it occurs in humans (Veening-Griffioen et al., 2019), failing to develop the comorbidities and risk factors commonly associated with sporadic, late-onset AD (Marshall et al., 2023). Even with optimized or humanized animal models, inherent interspecies differences in metabolism, gut microbiome, immune function, and epigenetic regulation considerably limit their external validity (Pound and Ritskes-Hoitinga, 2018). The inadequacy of traditional research models, coupled with their design being grounded in flawed or reductionist theories of disease etiology, has likely contributed to hindering a full understanding of disease complexity and played a relevant role in AD drug development failures.

In recent years, the use of human-centered complex in vitro models (CIVMs) such as cerebral organoids and organ-on-a-chip (OOC) systems, often derived from human induced pluripotent stem cells (hiPSCs) or primary cells, have deepened our understanding of AD pathology. CIVMs offer improved physiological relevance over traditional 2D monolayer cell cultures and animal models by better mimicking the complexities of the human brain and disease processes (Sreenivasamurthy et al., 2023; Dolciotti et al., 2025). Although 2D models have contributed significantly to AD research and remain widely used, they fall outside the scope of this review, which focuses on CIVMs that recapitulate multicellular and microenvironmental features more comprehensively. Furthermore, patient-derived CIVMs can be used to study individual differences in disease progression and response to treatment, possibly informing personalized medicine approaches (Lopes and Guil-Guerrero, 2025). These human-centered platforms can also be leveraged to explore novel etiological hypotheses underlying AD onset and to elucidate their potential interconnections.

In this mini review, we highlight recent applications of human-centered CIVMs (Figure 1), particularly cerebral organoids and single- or multi-OOC systems, and how they have been applied to investigate: (i) the potential impact of pathogen infections on neuroinflammation and AD; (ii) the effects of environmental pollutants on neurodegeneration and AD risk; and (iii) how alteration of gut microbiota and the GBA may drive neuroinflammation and neurodegeneration in AD. Together, these CIVM-based studies (summarized in Table 1) offer valuable human-relevant insights into AD pathogenesis, helping to uncover disease mechanisms and identify potential therapeutic targets.

Despite decades of research investment, the number of drugs available for managing AD remains limited, with most providing only symptomatic relief and benefiting a restricted subset of patients (Oxford et al., 2020). The recent approval of amyloid-targeting antibodies marks an important step toward disease-modifying therapies. However, these treatments are indicated primarily for individuals with early-stage AD or mild cognitive impairment, and they are not curative, carrying their own risks and limitations ⁵ .

New hypotheses have emerged to explain the complex etiopathology of AD (Zhang et al., 2024), which should be taken into account when designing novel therapeutic and preventive strategies. This is particularly relevant given that nearly half of all dementia cases could be prevented by addressing modifiable risk factors (Livingston et al., 2024).

The high historical failure rate in AD drug development (Cummings et al., 2014) may have stemmed from an overreliance on reductionist disease hypotheses and inadequate preclinical models, including transgenic animals and simplistic in vitro systems. Emerging human-centered models now offer powerful tools to elucidate the role of risk factors in triggering and exacerbating neurodegeneration and AD (Sreenivasamurthy et al., 2023; Lopes and Guil-Guerrero, 2025; Dolciotti et al., 2025). Particularly in the field of AD research, where animal experimentation continues to feature prominently, CIVMs have been at the forefront of a recent paradigm shift towards the broader adoption of human-centered and non-animal methodologies (Taylor et al., 2024; Vashishat et al., 2024) to drive progress and increase the translatability of preclinical research findings (Mehta et al., 2025).

The enhanced applicability of CIVMs for modeling human biology at both cellular and molecular levels has progressively enabled comprehensive investigations into novel hypotheses underlying AD etiology. These include neuroinflammatory responses to pathogens (Seaks and Wilcock, 2020; Catumbela et al., 2023), the effects of environmental toxicants (Dhapola et al., 2024), and the involvement of interactions along the GBA (Seo and Holtzman, 2024), which are increasingly recognized as potential drivers of neurodegeneration. As reported in this review, human-centered CIVMs, particularly cerebral organoids and OOC systems, can be applied to explore these new etiological hypotheses.

Each model has its strengths and weaknesses. CIVMs offer the potential advantage of being patient-specific, preserving individual (epi)genetic traits that support personalized and precision medicine approaches (Lopes and Guil-Guerrero, 2025). While they hold promise for overcoming the constraints of animal and simplistic in vitro models, technical optimization remains essential to realize their full potential. For example, brain organoids model early brain development rather than the aging brain, so their relevance to age-related neurodegeneration remains uncertain and needs rigorous validation through complementary approaches (Cerneckis et al., 2023). In addition, lack of vascularization can lead to internal hypoxia and cellular stress, resulting in necrosis and the impaired specification of cellular subtypes (Parthasarathy et al., 2026). CIVMs also face reproducibility issues due to biological variability, limited standardization and scalability, and poor reporting, hindering their validation and adoption by industry (Pamies et al., 2024).

Despite these limitations, legislation in the EU (European Medicines Agency, 2023), the US (Food and Drug Administration, 2024; Food and Drug Administration, 2025) and other regions is increasingly supporting the integration of non-animal approaches in pharmaceutical development, signaling a broader shift toward human-centered research.

Recent evidence suggests that dynamic culture conditions can substantially improve the physiological relevance of in vitro models. For example, perfusion-based systems have been shown to better recapitulate in vivo vascularization by enhancing nutrient delivery, waste removal, and biomechanical cues compared with static culture (Quintard et al., 2024). Similarly, oxygenation has been identified as a critical determinant of organoid viability and maturation, with insufficient oxygen supply markedly impairing cellular health and disrupting normal tissue development in brain organoids (Leung et al., 2022; Mohapatra et al., 2025). Moreover, variations in extrinsic forces and spatio-temporal dynamics can influence cellular composition and metabolic parameters in brain organoids (Aiello et al., 2025), thereby affecting both synaptogenesis and the proportion of specialized neuronal cells and glia (Saglam-Metiner et al., 2023), which represent critical endpoints for faithfully modeling neurodegenerative conditions and AD. Together, this highlights that microenvironmental parameters are central to maintaining homeostatic and pathophysiological processes in complex 3D cultures. Therefore, it is imperative not only to optimise CIVM culture parameters systematically, but also to implement rigorous, longitudinal characterisation of these test systems. Such characterisation should include quantitative assessment of culture conditions, cell viability, and disease-relevant biomarkers to ensure reproducibility and predictive validity.

To enhance reproducibility, fit-for-purpose guidelines have been introduced to promote the consistent performance of CIVMs (Pamies et al., 2024). Furthermore, minimum reporting standards for in vitro models, including organoids and microphysiological systems (Pamies et al., 2018; Mohapatra et al., 2025), have been proposed to enhance study quality for safety and regulatory assessments, with relevance also for basic and translational research and drug efficacy testing.

We acknowledge two main limitations of this study. Firstly, due to word count constraints, only studies employing organoids and OOC models were considered, whereas other CIVMs such as co-culture models could also be relevant. Secondly, for the same reason, we focused on selected hypotheses regarding AD pathogenesis, while recognizing that CIVMs have also been used to investigate additional etiological hypotheses not addressed in this review.

CIVMs are playing an increasingly pivotal role in AD research. Their capacity to recapitulate key aspects of human brain physiology and pathology offers unprecedented opportunities to deepen our understanding of AD etiopathogenesis, including the contribution of genetic, molecular, and environmental risk factors.

Modern research increasingly combines CIVMs with advanced computational approaches, AI, and digital twin technologies to integrate and interpret the rapidly expanding body of biological and clinical data, thereby accelerating target discovery and drug development (Li et al., 2018; Cheng et al., 2025). Such integrative frameworks hold great promise for revealing novel disease mechanisms and optimizing therapeutic interventions in a human-relevant context.

Future research should focus on disentangling correlations from causal relationships among risk factors and comorbidities to more precisely identify pathogens and environmental contributors that directly drive neurodegeneration and AD onset. Elucidating these causal pathways at the molecular and cellular levels will be critical for developing effective preventive and therapeutic strategies. These efforts should be underpinned by the broader adoption of human-centered CIVMs and other new approach methodologies which can improve translatability, reduce reliance on animal models, and enable truly patient-centered approaches to AD research, prevention, and treatment development.