Abstract

Background:

Mitochondrial dysfunction is widely recognized as a feature of aging, but the term encompasses heterogeneous processes, including altered energy production, redox imbalance, substrate handling, respiratory capacity, and mitochondrial quality control. Biological resilience depends on the capacity to respond to stressors, restore homeostasis, and preserve maintenance and repair. However, within mechanism-oriented aging literature, how mitochondrial and bioenergetic mechanisms are represented across aging hallmark domains, and whether they can be organized along a continuum of impairment, remains incompletely defined.

Methods:

This study used a secondary evidence map and narrative synthesis based on a previously curated, mechanism-oriented literature dataset; it was not designed as a comprehensive hallmark-by-hallmark systematic review. The original search captured stress adaptation, bioenergetic regulation, systemic dysfunction, and metabolic strain. Included reports were recategorized by hallmark domain, stage of mitochondrial or bioenergetic impairment, mechanistic domain, evidence type, and evidence tier. The final synthesis included 433 reports.

Results:

Within this mechanism-enriched dataset, mapped evidence was concentrated in regulatory and systems-level hallmark domains, particularly altered intercellular communication, chronic inflammation, and deregulated nutrient sensing. The most frequent mechanistic labels involved redox imbalance, substrate reallocation, oxidative throughput limitation, limited adenosine triphosphate availability, and mitochondrial quality-control impairment. Stage-based mapping showed that functional, adaptive, and structural labels often coexisted across hallmark domains rather than forming discrete categories.

Conclusion:

These findings identify recurrent co-representation of mitochondrial and bioenergetic mechanisms with selected aging hallmark domains visible within a mechanism-oriented evidence set. The results support a cautious, hypothesis-generating interpretation in which bioenergetic constraint may provide a conceptual lens for organizing stage-like patterns, rather than reducing mitochondrial aging biology to a binary distinction between “function” and “dysfunction.” However, this analysis does not establish biological centrality, causal direction, temporal sequence, tissue specificity, or generalizability across the broader aging literature.

Systematic Review Registration: https://www.crd.york.ac.uk/PROSPERO/view/, identifier CRD420251033154.

1 Introduction

Mitochondria are important regulators of cellular and systemic homeostasis, extending beyond their canonical role in adenosine triphosphate (ATP) production to include redox regulation, metabolic integration, intracellular signaling, stress adaptation, and intercellular communication (Mootha et al., 2003; Calvo and Mootha, 2010; Cohen, 2014; Picard and Shirihai, 2022). Mitochondrial signaling is highly dynamic and context-dependent, allowing cells to adjust energy production, substrate use, calcium handling, apoptotic sensitivity, and stress-response pathways in response to changing physiological demands (Mochly-Rosen et al., 2012; Skalka et al., 2024). Mitochondrial dysfunction remains one of the most consistently observed features across aging tissues and age-related diseases (López-Otín et al., 2013; Sun et al., 2016). However, the term remains mechanistically imprecise. It is often used as an umbrella descriptor for heterogeneous alterations, including reduced ATP production, impaired respiratory reserve, redox imbalance, increased reactive oxygen species, altered substrate oxidation, and defective mitochondrial quality control (Brand and Nicholls, 2011; Nicholls and Ferguson, 2013; Wallace, 2013; Picard et al., 2016). This broad usage can obscure whether observed mitochondrial changes represent early functional constraint, compensatory adaptation, structural damage, or a mixture of these processes.

Recent work has further emphasized that mitochondria should not be interpreted through a binary framework of “function” versus “dysfunction.” Mitochondria are multifunctional, cell-type-specific, and dynamically recalibrated in response to metabolic and environmental conditions (Monzel et al., 2023). Changes in mitochondrial properties may therefore reflect adaptive shifts in molecular features, enzymatic activities, integrated functions, or organelle-level behaviors rather than irreversible damage alone. Greater conceptual precision is needed to distinguish state-dependent bioenergetic limitations from persistent structural impairment and to clarify how mitochondrial alterations relate to broader aging processes.

The aging hallmarks framework provides useful architecture for organizing these processes. The original and expanded hallmark models describe aging as a networked biological process involving molecular damage, stress-response pathways, metabolic regulation, cellular senescence, impaired communication, inflammation, dysbiosis, extracellular matrix remodeling, and broader contextual influences, including psychosocial isolation (López-Otín et al., 2013, 2023; Kroemer et al., 2025). In the present review, psychosocial isolation is acknowledged as part of this expanded geroscience context but was not included as a primary mapping domain because the analysis focused on biological hallmark domains in which mitochondrial and bioenergetic mechanisms could be directly coded. These biological hallmarks are increasingly understood as interconnected rather than independent. However, the bioenergetic basis of this interconnection remains incompletely defined.

One possible approach is to interpret mitochondrial impairment as a staged bioenergetic continuum. Early changes may involve functional constraints, such as reduced oxidative throughput, altered redox balance, impaired respiratory reserve, or limited ATP availability. These changes may trigger adaptive responses, including substrate reallocation, altered nutrient sensing, inflammatory signaling, glycolytic shift, or changes in intercellular communication. With persistent or unresolved stress, these adaptive states may become associated with structural impairment, including mitochondrial DNA damage, impaired quality control, organelle remodeling, senescence, proteostatic failure, or tissue-level decline (Picard et al., 2016; Sun et al., 2016; Spinelli and Haigis, 2018). This staged interpretation may help clarify how mitochondrial alterations are described in relation to multiple aging hallmark domains without assuming a single linear or universal pathway.

From a stress-adaptation and exposome perspective, biological systems must continuously allocate finite bioenergetic resources toward immediate survival, repair, maintenance, and long-term function (Sterling and Eyer, 1988; McEwen, 1998; McEwen and Wingfield, 2003; Rappaport, 2016). Persistent environmental, metabolic, inflammatory, or psychosocial stress may increase energetic demand while limiting the capacity for full recovery. Because mitochondrial bioenergetics supports both adaptive stress responses and restoration of homeostasis, constrained mitochondrial processing capacity may be reported alongside unresolved stress signaling, impaired repair, altered substrate handling, and progressive loss of physiological resilience (Picard et al., 2018; Picard and McEwen, 2018; Spinelli and Haigis, 2018). Recent conceptual work has proposed bioenergetic limitation as a potential organizing framework for diverse manifestations of physiological decline (Tippairote et al., 2025). In this context, the present review does not assume this framework to be correct but evaluates whether the existing literature shows recurring patterns consistent with such an interpretation.

Despite extensive investigation across mitochondrial biology and aging, it remains unclear whether mitochondrial and bioenergetic mechanisms recur across multiple aging hallmark domains, whether they can be organized by stage of impairment, and whether the literature supports an integrated or fragmented interpretation. Accordingly, this secondary evidence map reclassifies a previously curated, mechanism-oriented dataset to examine how mitochondrial and bioenergetic mechanisms are represented in relation to aging hallmark domains, with particular attention to core mechanisms, stage distribution, evidence type, and cross-domain co-representation. This framing is directly relevant to biological resilience because recovery from stress requires coordinated energy production, redox control, substrate handling, inflammatory resolution, and repair capacity across tissues. Because the analysis uses a previously curated, mechanism-oriented dataset, the review should be interpreted as a secondary evidence map of visible mitochondrial and bioenergetic co-representation, not as a comprehensive hallmark-by-hallmark systematic review.

2 Study objectives and research questions2.1 Primary objective

To systematically map reports within a previously curated, mechanism-oriented dataset describing mitochondrial and bioenergetic mechanisms in relation to aging hallmark domains, with particular attention to whether reported mechanisms show recurring patterns consistent with a functional–adaptive–structural organizing framework.

2.2 Secondary objectives

To identify, within the curated dataset, reports detailing mitochondrial or bioenergetic mechanisms relevant to aging hallmark domains, including redox imbalance, ATP limitation, oxidative throughput limitation, substrate reallocation, and mitochondrial quality control impairment.

To classify reported mitochondrial alterations according to stage of impairment: functional, adaptive, or structural.

To map the distribution of mitochondrial and bioenergetic mechanisms across aging hallmark domains.

To evaluate the extent to which mechanistic labels co-occur across reports, particularly among redox imbalance, substrate reallocation, oxidative throughput limitation, ATP limitation, inflammation, intercellular communication, and deregulated nutrient sensing.

To characterize the evidence base by report type and biological model, including review-based, human, animal, and cellular evidence.

To identify underrepresented hallmark domains, mechanistic areas, and study designs requiring further investigation.

2.3 Research questions

What mitochondrial and bioenergetic mechanisms are most frequently described in relation to aging hallmark domains within the curated mechanism-oriented dataset?

How are reported mitochondrial mechanisms distributed across functional, adaptive, and structural stages of impairment?

Which aging hallmark domains are most frequently co-represented with mitochondrial and bioenergetic mechanisms in the included reports?

Which mechanistic and hallmark labels most commonly co-occur across the included reports?

What types of evidence — review-based, human, animal, or cellular — contribute to the curated evidence base on mitochondrial mechanisms and aging hallmarks?

What gaps remain in the literature, particularly regarding underrepresented hallmark domains, mitochondrial mechanisms, longitudinal evidence, and interventional studies?

3 Materials and methods3.1 Protocol and registration

The present secondary systematic evidence map was conducted and reported in accordance with PRISMA 2020 and was registered with PROSPERO under registration number CRD420251033154. The present manuscript represents a systematic evidence map and secondary conceptual synthesis of a curated systematic dataset, which was originally developed to examine stress adaptation, bioenergetic regulation, systemic dysfunction, and biomarker patterns of metabolic strain. For the present analysis, this dataset was re-mapped to explore reports detailing mitochondrial and bioenergetic mechanisms in relation to aging hallmark domains. No additional database searches were performed beyond the original search and citation searching described below. Therefore, the present study necessarily inherits the scope and limitations of the original mechanism-oriented search. As a result, the dataset is likely enriched for reports that explicitly use mitochondrial, metabolic, bioenergetic, stress-adaptation, inflammatory, or systemic-dysfunction terminology, whereas hallmark-specific literature framed through alternative vocabularies may be underrepresented.

Any amendments to the original protocol are reported transparently as methodological adaptations. The main protocol adaptations were reclassification of the previously curated dataset according to aging hallmark domains, addition of the functional–adaptive–structural staging framework, and use of descriptive co-occurrence mapping rather than quantitative synthesis.

3.2 Study design

This study was designed as a structured evidence map and secondary conceptual synthesis. Its aim was to identify, classify, and synthesize reports describing mitochondrial and bioenergetic mechanisms in relation to aging hallmark domains.

The purpose was not to estimate pooled effect sizes or test a single intervention-outcome relationship. Instead, the review examined how mitochondrial mechanisms are represented within the curated mechanism-oriented dataset and whether reported mechanisms could be organized according to stage of impairment, mechanistic domain, hallmark domain, evidence type, and biological model.

This evidence-map approach was selected because the included literature was expected to be heterogeneous, including human, animal, cellular, review-based, and conceptual reports with diverse outcome measures and levels of mechanistic detail.

3.3 Conceptual framework

The analytical framework integrated two complementary perspectives.

First, reported mitochondrial impairment was classified using a dynamic process that distinguished functional limitation, adaptive compensation, and structural disruption. Within this framework, early functional changes may include reduced oxidative capacity, impaired respiratory reserve, ATP limitation, redox imbalance, or altered substrate oxidation. Adaptive responses may include substrate reallocation, metabolic inflexibility, altered nutrient sensing, inflammatory signaling, glycolytic shift, stress-response activation, or changes in intercellular communication. Structural changes may include mitochondrial DNA damage, impaired mitochondrial quality control, organelle remodeling, loss of proteostasis, extracellular matrix remodeling, cellular senescence, stem cell exhaustion, or tissue-level degeneration.

Second, included reports were mapped to aging hallmark domains. Domains considered during evidence mapping included genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis, disabled macroautophagy, deregulated nutrient sensing, mitochondrial dysfunction as the entry domain, cellular senescence, stem cell exhaustion, altered intercellular communication, chronic inflammation, dysbiosis, and extracellular matrix remodeling.

Rather than treating mitochondrial dysfunction only as a discrete aging hallmark, this review examined whether specific mitochondrial and bioenergetic mechanisms—such as redox imbalance, ATP limitation, substrate reallocation, oxidative throughput limitation, and mitochondrial quality control impairment—were represented across multiple aging hallmark domains.

In this review, “bioenergetic constraint” is used as an operational interpretive term, not as a directly measured outcome. It refers to a reported or inferred condition in which mitochondrial or cellular energy-processing capacity appears insufficient relative to physiological demand, reflected by reported features such as impaired oxidative throughput, altered redox balance, limited ATP availability, reduced respiratory reserve, substrate rerouting, or impaired energy-dependent maintenance and repair. These features were coded, when reported, as specific mechanistic labels. The broader term “bioenergetic constraint” was used only in the Discussion and Conclusion as a hypothesis-generating synthesis of recurrent co-labeling patterns, not as an independent analytical category or causal claim.

3.4 Eligibility criteria

Reports were eligible for inclusion if they met the following criteria.

3.4.1 Study type and scope

Eligible reports included original research studies, systematic reviews, narrative reviews, and conceptual or theoretical papers relevant to mitochondrial biology, bioenergetics, stress adaptation, aging biology, or systemic physiological decline. Original research could include observational, interventional, translational, animal, cellular, or mechanistic reports.

3.4.2 Biological model

Eligible reports included human reports, animal models, and cellular or in vitro reports. Human studies could include clinical, population-based, interventional, or observational designs. Animal and cellular studies were included when they addressed mechanisms relevant to aging, metabolic adaptation, mitochondrial regulation, or systemic physiological stress.

3.4.3 Mitochondrial or bioenergetic relevance

Reports were eligible if they described mechanisms relevant to mitochondrial function or bioenergetic regulation, including oxidative phosphorylation, respiratory capacity, redox regulation, ATP production, NAD+/NADH balance, substrate utilization, metabolic flux, mitochondrial quality control, mitochondrial dynamics, mitonuclear signaling, mitochondrial stress responses, oxidative stress, or energy-dependent maintenance and repair.

Reports were excluded if they mentioned mitochondrial dysfunction only generically without describing a relevant mechanism, or if they lacked a clear mitochondrial, metabolic, or bioenergetic component.

3.4.4 Aging hallmark or system-level relevance

Reports were required to address systemic physiological adaptation or at least one process relevant to aging hallmarks. Eligible domains included genomic stability, epigenetic regulation, proteostasis, autophagy, nutrient sensing, cellular senescence, stem cell function, inflammation, intercellular communication, extracellular matrix remodeling, dysbiosis, metabolic adaptation, or functional decline.

Disease-specific studies were included when they provided mechanistic insight relevant to broader aging biology, systemic adaptation, or hallmark-related processes. Reports focused exclusively on disease-specific outcomes without relevance to systemic or aging-related mechanisms were excluded.

3.5 Information sources and search strategy

The literature dataset used in this review was derived from a systematic search designed to capture four interrelated conceptual domains:

stress adaptation and systemic strain;

bioenergetic and metabolic trade-offs;

system-level dysfunction across physiological domains; and

biomarker patterns associated with metabolic stress and functional impairment.

The original search strategy was intentionally broad and mechanism-oriented rather than restricted to aging hallmark terminology. It was designed to capture literature at the intersection of stress adaptation, bioenergetic regulation, systemic dysfunction, and biomarker patterns of metabolic strain. For the present evidence map, included reports were subsequently reclassified according to mitochondrial and bioenergetic mechanisms, stage of impairment, and aging hallmark domains.

Because the present review was not designed as a comprehensive hallmark-by-hallmark systematic review or catalogue, the search strategy directly shaped the evidence map. Reports using mitochondrial, metabolic, bioenergetic, inflammatory, senescence-related, or stress-adaptation terminology were more likely to be captured, whereas hallmark-specific studies framed primarily through telomere biology, chromatin regulation, extracellular matrix remodeling, microbiome biology, tissue-specific degeneration, or other non-bioenergetic vocabularies may have been missed.

Database searches were conducted in PubMed, Scopus, Web of Science Core Collection, and the Cochrane Library. The final database search was completed on May 15, 2025. Search strategies combined controlled vocabulary, where available, and free-text terms related to chronic stress, resilience, allostatic load, energy metabolism, mitochondrial function, substrate allocation, inflammation, senescence, sarcopenia, oxidative stress, and metabolic biomarkers. Database searches were limited to reports published between 2005 and 2025. Reports published after the final database search were not added through a new systematic database search, but a limited number were retained when identified through citation searching, final update checks, or evidence-map verification and when they provided high-yield mechanistic context directly relevant to interpretation. Although PubMed and Cochrane searches applied human-subject and language filters consistent with the original clinical/systemic review scope, animal and cellular reports were eligible when identified through Scopus, Web of Science, citation searching, final update checks, or evidence-map verification and when they provided direct mechanistic relevance to the review questions.

Forward and backward citation searching was conducted on high-yield reports to identify additional mechanistic or conceptual literature relevant to mitochondrial mechanisms, bioenergetic impairment, and aging hallmark mapping. Citation searching identified additional records that were screened and assessed using the same eligibility framework. A limited number of foundational mechanistic reports published before 2005, as well as newly available reports published after the final database search, were retained when identified through citation searching, final update checks, or evidence-map verification, provided that they offered high-yield mechanistic context directly relevant to mitochondrial bioenergetic mechanisms, aging hallmark domains, or interpretation of the functional–adaptive–structural staging framework. These reports were not used to expand the database search window systematically but were retained to preserve conceptual and mechanistic completeness within the evidence map.

The full search strategy for each database, including filters and limits, is provided in Supplementary Table 1.

3.6 Study selection and screening process

Search results were imported into Rayyan for de-duplication and screening management. Duplicate records were removed before screening. Records marked as ineligible before screening were recorded in the PRISMA flow diagram. Rayyan was used to facilitate organization, de-duplication, and screening; eligibility decisions were made by reviewers and were not replaced by automated classification.

A two-stage screening process was conducted by two independent reviewers. First, titles and abstracts were screened against predefined eligibility criteria. Second, full-text reports were retrieved and assessed for eligibility based on the inclusion and exclusion criteria and conceptual relevance to the study objectives.

Disagreements between reviewers were resolved through discussion. When consensus could not be reached, a third reviewer adjudicated. Reports that could not be retrieved after reasonable effort were classified as not retrieved and excluded from full-text assessment.

Exclusion reasons were not systematically recorded for all records excluded during title/abstract screening; therefore, the PRISMA diagram reports aggregate exclusion counts for this phase. Supplementary Table 2 provides selected examples of excluded reports with reasons, focusing on reports that appeared potentially eligible or required closer adjudication, and should not be interpreted as a complete exclusion-reason log for all excluded records.

3.7 Data extraction

Data were extracted using a standardized and piloted extraction form. Extraction was performed independently by two reviewers, and discrepancies were resolved by consensus. Coding decisions were reviewed through consensus checking, and disagreements or uncertain cases were logged during extraction. Unresolved cases were adjudicated by a third reviewer. Formal inter-rater agreement statistics were not calculated because the labeling framework used non-mutually exclusive, context-dependent categories rather than mutually exclusive diagnostic classes. Most disagreements involved boundary decisions between adaptive and structural labels, especially for inflammation, nutrient sensing, mitochondrial quality control, and senescence-related processes. Representative ambiguous cases and final coding decisions are provided in Supplementary Table 3.

Extracted variables included:

bibliographic information: first author, publication year, journal, and report type;

study characteristics: design, population or model system, sample size where applicable, and setting;

biological model: human, animal, cellular, review-based, or conceptual;

mitochondrial and bioenergetic variables: oxidative phosphorylation, respiratory capacity, ATP production, redox status, NAD+/NADH balance, substrate utilization, metabolic flux, mitochondrial quality control, mitochondrial dynamics, or oxidative stress;

aging hallmark-related processes: hallmark domain or system-level process assessed;

mechanistic findings: pathways linking mitochondrial or bioenergetic alteration to biological outcome;

clinical, molecular, cellular, or functional outcomes, where applicable;

interventions, exposures, confounders, and study limitations, where reported;

reviewer notes relevant to classification.

Missing or unclear information was recorded as “not available.” When a report included multiple relevant biological systems, hallmark domains, or mechanistic domains, all applicable labels were retained.

The data extraction template is provided in Supplementary Table 3A.

3.8 Evidence labeling framework

Following data extraction, each included report was classified using a structured, multi-dimensional evidence-labeling framework. Labels were not mutually exclusive; therefore, a single report could receive multiple stage, mechanistic, hallmark, and evidence-type labels. The labeling framework included four primary analytical dimensions: stage of mitochondrial or bioenergetic impairment, mechanistic domain, aging hallmark domain, and evidence type.

When a process could function both as a stage-related feature and as a hallmark domain, stage labels were assigned according to the role of the process in the report’s mechanistic narrative, whereas hallmark labels were assigned according to the broader aging domain represented. This distinction was used to reduce circular interpretation between stage classification and hallmark mapping.

To reduce circular interpretation, labels were assigned from the reported role of the process rather than from the biological term alone. Processes such as inflammation, nutrient sensing, intercellular communication, mitophagy, and mitochondrial quality control were therefore not automatically assigned to a fixed stage. For example, inflammatory signaling was coded as adaptive when described as a stress-responsive or compensatory response to mitochondrial or bioenergetic strain, but chronic inflammation was coded as a hallmark domain when it represented the broader aging-related domain under study. Similarly, mitophagy activation or mitochondrial quality-control remodeling was coded as adaptive when presented as compensatory organelle renewal, whereas impaired mitophagy, failed mitochondrial turnover, or accumulation of damaged mitochondria was coded as structural when presented as persistent organelle-level disruption. When a report described more than one role, multiple non-mutually exclusive labels were retained. Ambiguous coding examples and decision rules are provided in Supplementary Table 3.

3.8.1 Stage of mitochondrial or bioenergetic impairment

Reports were categorized according to whether their findings aligned with one or more of the following stages.

Functional stage: reports describing altered mitochondrial or bioenergetic performance without clear structural damage. Examples included reduced oxidative capacity, impaired respiratory reserve, ATP limitation, redox imbalance, altered NAD+/NADH state, impaired substrate oxidation, or reduced metabolic flexibility.

Adaptive stage: reports describing compensatory, transitional, or stress-responsive changes in response to mitochondrial or bioenergetic strain. Examples included substrate reallocation, metabolic inflexibility, glycolytic shift, altered nutrient sensing, inflammatory signaling, stress-response activation, mitonuclear signaling, or altered intercellular communication.

Structural stage: reports describing persistent cellular, organelle-level, or tissue-level disruption. Examples included mitochondrial DNA damage, impaired mitochondrial quality control, altered mitochondrial dynamics, organelle remodeling, loss of proteostasis, extracellular matrix remodeling, cellular senescence, stem cell exhaustion, or tissue degeneration.

3.8.2 Mechanistic domain

Reports were assigned one or more mechanistic labels according to the mitochondrial or bioenergetic processes described. Core mechanistic domains included:

redox imbalance;

substrate reallocation or metabolic inflexibility;

oxidative throughput limitation or impaired oxidative capacity;

ATP limitation;

mitochondrial quality control impairment.

Additional mechanistic labels, including calcium dysregulation, mitochondrial dynamics, environmental inputs, mitonuclear signaling, and oxidative stress, were recorded where relevant. However, only the most recurrent core mechanistic domains were displayed in the main mechanistic distribution figure.

3.8.3 Aging hallmark domain

Reports were mapped to one or more aging hallmark domains, including altered intercellular communication, chronic inflammation, deregulated nutrient sensing, loss of proteostasis, cellular senescence, genomic instability, stem cell exhaustion, epigenetic alterations, extracellular matrix remodeling, disabled macroautophagy, dysbiosis, and telomere attrition.

Because mitochondrial dysfunction constituted the primary mitochondrial/bioenergetic entry domain for inclusion and mechanistic classification, it was not treated as a separate final hallmark outcome in the main evidence-map figures.

3.8.4 Evidence type

Reports were categorized by evidence type as review, human, animal, or cellular evidence. These categories were not mutually exclusive. For example, a review could synthesize human and animal evidence, and an original report could include more than one experimental model.

The full evidence-labeling framework and operational definitions are provided in Supplementary Tables 3B, E.

3.9 Evidence tiering

To contextualize the strength of mechanistic inference, included reports were also categorized using a four-level evidence-tier framework. This framework assessed the degree to which each report supported a relationship between mitochondrial or bioenergetic mechanisms and aging-related processes.

Tier 1 – Direct evidence: reports directly linking mitochondrial or bioenergetic mechanisms to aging-related processes through experimental, longitudinal, interventional, or clearly mechanistic evidence.

Tier 2 – Mechanistic support: reports describing mitochondrial, metabolic, redox, substrate-handling, ATP-related, or quality-control mechanisms relevant to aging or systemic adaptation, but without directly demonstrating causal or temporal progression across stages of impairment.

Tier 3 – Downstream associations: reports describing aging-related outcomes, systemic dysfunction, inflammation, frailty, senescence, functional decline, or disease-related changes with indirect or limited mechanistic linkage to mitochondrial or bioenergetic impairment.

Tier 4 – Contextual evidence: reviews, conceptual frameworks, or background literature providing theoretical or interpretive support.

Tier classification was performed independently by two reviewers, with discrepancies resolved through discussion and consensus. The evidence-tier framework was used to contextualize mechanistic contribution and inferential strength and was not designed as a formal certainty-of-evidence rating.

The evidence-tiering framework is provided in Supplementary Table 3C.

3.10 Data synthesis and evidence mapping

Because of substantial heterogeneity in report type, biological model, outcome measures, and mechanistic focus, quantitative meta-analysis was not performed. No pooled effect estimates were calculated. This heterogeneity supported the use of structured evidence mapping and narrative synthesis rather than quantitative meta-analysis.

Instead, a structured narrative synthesis and descriptive evidence map were generated. Synthesis was organized across the following dimensions:

The synthesis focused on identifying recurring mitochondrial mechanisms, distributional patterns across hallmark domains, stage-related patterns of impairment, label co-occurrence, and evidence gaps.

Descriptive counts and proportions were calculated for non-mutually exclusive labels. Because reports could receive multiple labels, counts represent labeled report associations rather than mutually exclusive study totals.

3.11 Graphical and network visualization methods

Evidence-map findings were displayed using horizontal bar charts, proportional stacked bar charts, absolute stacked bar charts, and a co-occurrence network.

Horizontal bar charts were used to summarize evidence type and core mechanistic-domain frequencies. Absolute stacked bar charts were used to show the number of report-level stage-by-hallmark label associations across aging hallmark domains. Proportional stacked bar charts were used to display the relative distribution of functional, adaptive, and structural labels within each aging hallmark domain.

A co-occurrence network was generated to visualize relationships between mechanistic domains and aging hallmark domains. In the network, node size reflected label frequency, node color distinguished mechanistic and hallmark categories, and edge thickness reflected the frequency of co-occurrence between labels within the same report.

All graphical analyses were descriptive. They were intended to support evidence mapping and hypothesis generation rather than causal inference.

3.12 Quality appraisal, risk-of-bias considerations, and certainty assessment

Given the heterogeneity of included report types, a single formal risk-of-bias tool was not applied across the entire dataset. However, heterogeneity was not treated as a reason to omit quality consideration. Instead, we conducted a structured quality-appraisal summary stratified by major evidence category. Primary human studies, animal studies, cellular or mechanistic studies, systematic reviews, narrative reviews, and conceptual or theoretical papers were considered separately. For primary empirical studies, appraisal focused on study design, population or model relevance, exposure and outcome definition, mechanistic specificity, confounding or experimental control, and directness to the mapped bioenergetic–hallmark relationship. For systematic reviews, appraisal focused on search transparency, inclusion criteria, synthesis method, and whether risk-of-bias assessment was reported. For narrative reviews and conceptual papers, appraisal focused on source transparency, balance of evidence, risk of selective citation, dependence on recycled field narratives, and whether claims were grounded in primary evidence.

This quality-appraisal summary was used to contextualize interpretation rather than to exclude reports or generate pooled certainty estimates. It is provided in Supplementary Table 3I. The evidence-tiering framework was retained as a separate classification of inferential role and was not treated as a substitute for methodological quality appraisal, formal risk-of-bias assessment, reporting-bias assessment, or GRADE certainty grading. This category-level appraisal does not replace design-specific tools such as ROBINS-I, RoB 2, SYRCLE, QUADAS, AMSTAR 2, or GRADE. Because the present study was designed as an evidence map rather than an intervention-effect, diagnostic-accuracy, or prognosis review, the appraisal was used to contextualize evidentiary contribution rather than to generate pooled certainty ratings. Accordingly, the findings should not be interpreted as graded certainty-of-evidence conclusions.

3.13 Sensitivity and heterogeneity assessment

No formal statistical sensitivity analyses were conducted because the synthesis was descriptive and evidence-map based. However, we conducted a descriptive original-study-only sensitivity analysis to examine whether the main co-representation patterns remained visible after excluding reviews and conceptual papers. Robustness of observed patterns was considered qualitatively by comparing distributions across evidence type, biological model, hallmark domain, mechanistic domain, stage of impairment, and evidence tier.

Heterogeneity was explored descriptively rather than statistically. Sources of heterogeneity included study design, model system, population, disease context, mechanistic focus, measurement approach, and level of inference.

To address the heterogeneous composition of the evidence base and the inclusion of reviews and conceptual papers, we performed an original-study-only sensitivity analysis restricted to primary empirical reports, including human, animal, cellular, and mixed-model studies. Systematic reviews, narrative reviews, and conceptual papers were excluded from this subset. The purpose of this analysis was to evaluate whether the main mechanism–hallmark co-representation patterns remained visible when secondary and conceptual literature were removed.

For this subset, we recalculated mechanistic-domain frequencies, hallmark-domain frequencies, stage-label frequencies, biological-model distribution, evidence-tier distribution, and mechanism–hallmark co-occurrence counts, using the same report-level labeling approach as the main analysis. We also generated an original-study-only mechanism–hallmark co-occurrence network.

3.14 Data and materials availability

The full database search strategies are provided in Supplementary Table 1, and selected excluded reports with reasons are provided in Supplementary Table 2. The data extraction template, evidence-labeling framework, evidence-tiering framework, included-report characteristics, label matrix, co-occurrence matrix, figure-generation notes, data availability notes, and quality-appraisal summary by evidence category are provided in Supplementary Table 3.

4 Results4.1 Study selection

The study selection process is summarized in Figure 1. After screening, retrieval, and eligibility assessment, 408 reports from database searches and 25 reports from citation searching were included, yielding a final synthesis set of 433 reports. Consistent with PRISMA 2020 terminology, results are presented at the report level because formal study-level deduplication across all publication types was not performed. The final dataset was therefore treated as a structured evidence map of included reports rather than as a set of mutually exclusive primary studies.

PRISMA 2020 flow diagram of study selection. Flow diagram showing identification, screening, retrieval, eligibility assessment, and inclusion of reports from database searches and citation searching. The final synthesis included 433 reports, comprising 408 reports from databases and registers and 25 reports identified through citation searching.

4.2 Characteristics of the included evidence base

The included reports represented a heterogeneous evidence base spanning review-based, human, animal, and cellular literature. Review-based evidence was the largest category, followed by human, animal, and cellular evidence. Table 1 summarizes the composition of the evidence base and the dominant evidence-map labels across evidence categories. Figure 2 provides a visual overview of the evidence-base composition and evidence-tier distribution. Unless otherwise stated, all counts in the Results represent non-mutually exclusive report-level labels rather than mutually exclusive study totals; therefore, a single report could contribute to more than one evidence type, stage, mechanism, or hallmark domain.

Overview of evidence-base composition and evidence-tier distribution. (A) Distribution of included reports by report-level evidence category. Categories were non-mutually exclusive; therefore, totals exceed the number of included reports. Percentages indicate the proportion of the full evidence map represented by each category. (B) Distribution of included reports by evidence tier. Tiers summarize the interpretive strength of evidence within the evidence map, ranging from direct evidence to contextual or conceptual evidence.

Values are report-level counts. Evidence categories were non-mutually exclusive; therefore, human, animal, cellular, and review/conceptual categories may overlap and should not be summed across rows or columns. Original empirical reports were defined as reports not labeled as review, conceptual, or systematic review. Dominant labels indicate the most frequently assigned evidence-map labels within each category and do not imply causal direction, biological centrality, or mutually exclusive classification. MQC, mitochondrial quality control.

As shown in Table 1, 194 reports were classified as original empirical reports and 239 as review/conceptual reports. Human evidence was represented in 163 reports, animal evidence in 58 reports, and cellular evidence in 50 reports. Across the full evidence map and across major evidence categories, the dominant mechanistic labels were redox imbalance, substrate reallocation, oxidative throughput limitation, and ATP limitation, while the most frequent hallmark labels were altered intercellular communication, chronic inflammation, and deregulated nutrient sensing.

Evidence categoryOriginal empiricalReview/conceptualDominant mechanistic labelsDominant hallmark labelsDominant stage labelsFull evidence map
(433 reports)194239Redox imbalance; substrate reallocation; oxidative throughput limitationAltered intercellular communication; chronic inflammation; deregulated nutrient sensingAdaptive; functional; structuralHuman evidence
(163 reports)12637Redox imbalance; ATP limitation; substrate reallocationAltered intercellular communication; chronic inflammation; deregulated nutrient sensingFunctional; adaptive; structural