JOURNAL OF HUMAN SCIENCE & TECHNOLOGY
Volume 1 · Issue 2 · Article 5 · ISSN 3144-6914
SCIENTIFIC METHODOLOGY · CRITICAL REVIEW

Pregnancy-Associated Biomarker Changes Are Not Equivalent to Biological Aging

A critical review of the postpartum recovery hypothesis and an independent restricted reanalysis of DunedinPACE

Kai Weng2 · Bomi Joseph, MD, PhD1,*

1 Peak Health Center, San Francisco, California, United States · ROR: 04b23x642
2 James Cook University, Singapore · ROR: 04gsp2c11

ORCID: Bomi Joseph 0000-0001-7626-2734 · Kai Weng 0009-0009-4186-0991
*Corresponding author: Kai Weng · kweng@alkemyindex.com

Submitted September 30, 2026 · Published October 26, 2026
DOI: 10.5281/zenodo.23630711

Abstract

BACKGROUND:

Pregnancy involves profound endocrine, metabolic, immunological, and cardiovascular adaptation. Recent work has interpreted shifts in epigenetic age estimators during gestation and postpartum as accelerated biological aging and recovery.

OBJECTIVE:

To evaluate the evidentiary basis for that interpretation and examine the principal DunedinPACE finding using traceable complete records.

METHODS:

We critically evaluated the Albright–Poganik hypothesis and five relevant studies, and examined Etzel et al.’s article, supplementary documentation, and available longitudinal workbook.

RESULTS:

The original study reported a significant DunedinPACE later-pregnancy-by-time spline interaction (approximately β=0.196, p=0.006), whereas corresponding interactions for three other biomarkers were nonsignificant. In the supplied workbook, 21 rows contained DunedinPACE values at all three nominal visits. Unadjusted paired prenatal-to-final change was −0.0089 (95% CI −0.1602 to 0.1425; p=0.904). However, absent original subject identifiers and uncertain pregnancy-indicator mappings prevented exact reproduction of the published interaction.

CONCLUSIONS:

Reproductive-state-dependent changes in aging estimators do not alone establish cumulative biological deterioration. The postpartum recovery hypothesis requires direct functional validation and more transparent reproducibility of its central statistical result.

1. INTRODUCTION

Pregnancy is a physiological transition characterized by coordinated changes in endocrine signaling, immune function, plasma volume, renal filtration, coagulation, and nutrient allocation. Many of these measurements also correlate with age in nonpregnant populations. A change in the same direction does not imply the same underlying biological mechanism.

Albright and Poganik [1] hypothesize that the extent of postpartum recovery, particularly before another pregnancy, determines cumulative maternal biological aging. This is a testable hypothesis, but the inference from an age-correlated molecular signal to lasting organismal deterioration requires evidence that the measurement is valid during pregnancy and predicts durable functional decline. Our objective is to distinguish demonstrated biomarker changes from claims about underlying aging processes.

2. METHODS AND SCOPE

This is a focused critical review, not a systematic review or meta-analysis. The evidence base comprises the Albright–Poganik review [1], Etzel et al. [2], Poganik et al. [3], Moran et al. [4], Ying et al. [5], Gurven et al. [6], and Sehgal et al. [7]. The full versions and Etzel supplemental documentation supplied for this project were examined. We did not independently verify every cited primary study in the hypothesis review.

For independent reanalysis, we used Trajectories_Data.xlsx, supplemental mmc1.pdf, and regression summaries in mmc3.xlsx. The workbook is wide-format: measurements at three nominal visits occur in the same row. We used original worksheet rows as provisional longitudinal identifiers, without asserting independent verification of uniqueness against the source cohort. We restricted paired comparisons to rows with non missing DunedinPACE at all three time points, without imputation, cell-composition residualization, or reclassification of pregnancy indicators. Paired t-tests and 95% confidence intervals were calculated on within-row differences. These tests do not reproduce the published adjusted natural-spline interaction.

3. EPIGENETIC AGE PREDICTION VERSUS BIOLOGICAL AGING

Conceptual diagram separating observed pregnancy-associated changes in methylation-derived clock output from inferences of persistent biological deterioration and cumulative maternal aging; independent validation is required.

A statistical clock trained to predict chronological age or an age-associated physiological composite is not automatically specific to irreversible biological deterioration. Pregnancy can alter clock inputs through adaptive endocrine regulation, immune-cell composition, and tissue-specific molecular remodeling. The inferential chain from altered methylation to altered clock output, then to accelerated biological aging, and finally to cumulative maternal damage requires validation at each transition.

Ying et al. [5] distinguish methylation changes inferred to be damaging from changes inferred to be protective, and report that conventional clocks were not significantly enriched for CpG sites inferred to causally affect the examined lifespan/healthspan traits. Their Mendelian randomization approach has assumptions and limitations; nonetheless, it shows why age prediction alone cannot establish damage.

3.1. SEHGAL ET AL. (2026): DUNEDINPACE RESPONSIVENESS AND PHYSIOLOGICAL-STATE DEPENDENCE

Sehgal et al. [7] evaluated 16 epigenetic clocks and 94 additional DNA methylation biomarkers across 51 longitudinal intervention studies. DunedinPACE showed significant decreases in 16 interventions and an increase in one, demonstrating that its output is responsive to changing physiological conditions. The authors noted that changes in composite epigenetic biomarkers may reflect intermediate health or inflammatory states rather than direct alterations in aging biology. This distinction is central to the present critique: cytokine activity, immune-cell composition, endocrine signaling, and metabolic regulation may vary during pregnancy and return toward baseline during recovery. A corresponding change in a methylation-derived pace estimate does not, by itself, demonstrate acceleration and reversal of organismal aging. Sehgal et al. did not conclude that methylation is unrelated to aging; rather, their findings underscore the need to validate what a changing clock output measures in each physiological context. In pregnancy, independent evidence of persistent functional deterioration is required before a clock response can be interpreted as accelerated biological aging.

3.2. AGING IS DIRECTIONAL: THE CONCEPTUAL INCONSISTENCY IN INTERPRETING REVERSIBLE BIOMARKER CHANGES

A fundamental distinction must be maintained between biological aging and reversible physiological change. Aging, understood as the progressive accumulation of age-associated deterioration and loss of physiological capacity, is a directional process. In contrast, physiological adaptation, homeostatic regulation, inflammatory responses, endocrine transitions, and recovery are dynamic processes that may fluctuate over periods ranging from hours to years.

This distinction is particularly relevant to the conceptual framework advanced by Poganik and colleagues.

In Aging vs. Biological Age: Conceptual Considerations for Age Reversal Claims (2026), Poganik et al. explicitly distinguish the underlying biological aging process from measurements of biological age. Their conceptual discussion recognizes the importance of defining what constitutes aging, rather than assuming that changes in an aging-associated biomarker necessarily represent changes in the aging process itself.

However, the postpartum recovery hypothesis advanced by Albright and Poganik (2026) interprets pregnancy-associated increases in biological-age estimates and their subsequent postpartum normalization within a framework of accelerated maternal aging, rejuvenation, and potentially cumulative aging resulting from incomplete recovery.

These interpretations raise a fundamental question of conceptual consistency.

If a biological-age estimator increases during pregnancy and decreases following childbirth, what precisely has aged and subsequently become younger?

A reversible change in DNA methylation, cytokine signaling, immune-cell composition, endocrine activity, or metabolic regulation does not, by itself, establish the accumulation or reversal of age-associated biological deterioration. Such changes may instead represent normal physiological responses to pregnancy and their subsequent resolution.

Consider inflammation. Cytokine concentrations can increase in response to physiological stress and decline as the inflammatory response resolves. This temporal variation reflects inflammatory regulation and recovery. It does not establish that the individual has biologically aged during the inflammatory response and subsequently reversed aging when cytokine activity subsides.

The same inferential limitation applies to methylation-derived aging estimators. Their responsiveness to changing physiological conditions does not establish that the underlying organismal aging process is moving forward and backward.

Aging and recovery must therefore be distinguished operationally. Under a definition of aging as progressive, persistent deterioration, a reversible biomarker fluctuation cannot itself constitute evidence of aging. A transient biological response may coexist with aging, but it should not be classified as aging without independent evidence connecting it to the underlying deterioration process.

This distinction does not exclude the possibility that pregnancy complications or repeated physiological stressors produce lasting biological damage. Rather, it requires such damage to be demonstrated independently instead of inferred from temporary changes in aging-associated measurements.

The conceptual distinction between aging and biological-age estimation articulated by Poganik et al. should be applied consistently when interpreting pregnancy-associated biomarker trajectories. Without that consistency, the postpartum recovery hypothesis risks conflating physiological adaptation and recovery with accelerated aging and rejuvenation.

A measurement that rises and falls with physiological state cannot, solely on that basis, establish a corresponding acceleration and reversal of biological aging.

4. APPRAISAL OF ETZEL ET AL.: THE EVIDENTIARY FOUNDATION OF THE POSTPARTUM HYPOTHESIS

Albright and Poganik [1] cite Etzel et al. [2] in support of a postpartum recovery-window hypothesis linking subsequent pregnancy to altered maternal biological-age trajectories. The evidentiary chain requires careful separation of (i) measured longitudinal biomarker changes, (ii) a modeled pregnancy-by-time interaction, and (iii) an inference of persistent biological aging. Etzel et al. provide evidence for the first two under their analysis choices, but their data do not directly demonstrate the third.

Reported later-pregnancy interaction p-values: DunedinPACE 0.006, telomere length 0.191, GrimAge2 approximately 0.14, and PhenoAge approximately 0.14. Only DunedinPACE meets the unadjusted 0.05 threshold.

4.1. FOUR BIOMARKERS DO NOT PROVIDE FOUR INDEPENDENT CONFIRMATIONS

Among the reported later-pregnancy interactions, DunedinPACE alone met the conventional unadjusted 0.05 significance threshold. Telomere length, GrimAge2 and PhenoAge did not. Non-significance is not proof of zero association, and differences among outcomes may reflect

OutcomeReported interaction pInterpretation
DunedinPACE0.006Significant selected spline-basis interaction
Telomere length0.191No statistically significant corresponding interaction
GrimAge2~0.14No statistically significant corresponding interaction
PhenoAge~0.14No statistically significant corresponding interaction

Table 1. Values reported in the current manuscript from Etzel supplemental regression results. These p-values are not effect sizes, and do not establish absence of an association.

distinct measurement properties. Nonetheless, the three nonsignificant results do not corroborate a shared cumulative-aging trajectory. The hypothesis should therefore not be presented as supported consistently across four aging measures.

4.2. COHORT ATTRITION AND UNVERIFIABLE ANALYSIS LINKAGE

The source study describes 130 participants, but its supplemental table identifies 45 with observations at all three visits, including six of 19 participants classified as having a later subsequent pregnancy. The separately disclosed wide-format trajectory workbook contains only 21 rows with DunedinPACE present at all three nominal visits. These counts refer to different levels of completeness and should not be conflated. In that workbook, visit-specific measurements occupy the same row, supporting provisional within-row linkage. However, no original subject identifier (FamID) is supplied, and the fields AddPreg_T18, AddPreg_T30, and AddPreg_T36 cannot be mapped unambiguously to the published PregAgain_Late classification. Accordingly, independent reviewers cannot establish unique original identities or exactly reproduce the key exposure grouping from that workbook alone.

DunedinPACE trajectories for 21 complete within-row records at prenatal, first follow-up, and final follow-up visits. Gray lines show records; the blue line shows the mean with standard errors.

4.3. THE MODEL COEFFICIENT IS NOT AN AGING-RATE EFFECT

The reported DunedinPACE result is a second natural-spline-basis-by-later-pregnancy interaction (approximately β=0.196, SE=0.072, p=0.006). A spline-basis coefficient describes a component of a fitted time-dependent group contrast; it is not directly interpretable as a 19.6% increase in aging, a change in biological years, or permanent biological deterioration. The result requires fitted marginal trajectories, uncertainty intervals, and explicit outcome preprocessing for meaningful interpretation. Cell-composition residualization, time coding, inclusion criteria, and missing-data assumptions all affect exact reproducibility.

The study lacks a non-pregnant comparator and postpartum BMI; later-pregnancy gestational age was not standardized. Saliva has variable epithelial and immune-cell content. The authors used reference-free deconvolution and imputed CpG values missing from the EPICv2 array. These decisions require careful validation when applying predominantly blood-trained clocks to pregnancy-associated tissue changes.

Poganik et al. [3] show that several biological-age estimates can change during stress and return toward baseline after recovery. This demonstrates responsiveness but does not distinguish restoration of biological function from normalization of stress-responsive clock inputs. If an algorithm reports that a woman has aged during pregnancy and become younger after childbirth, the first scientific question is not how she reversed aging. It is whether the algorithm was measuring aging in the first place.

Moran et al. [4] analyzed laboratory measurements across approximately 313,000 pregnancies and a large non-pregnant reference population. They show that apparent laboratory age can fall early in gestation, rise toward delivery, and return toward baseline. The mechanisms differ: reduced albumin during gestation can reflect hemodilution, while a similar decline in later life may arise from different mechanisms. Pregnancy-associated hypercoagulation may be adaptive, unlike harmful late-life hypercoagulation. Their aggregate cross-sectional design does not establish individual long-term outcomes.

Gurven et al. [6] report no consistent cumulative deterioration in the measured nutritional and immune indices with greater parity or reproductive pace among Tsimane women. They also document meaningful reproductive health burdens and do not assess epigenetic clocks. Their findings challenge universal depletion models but do not prove reproduction has no health costs.

5. INDEPENDENT RESTRICTED DUNEDINPACE RESULTS

5.1. RECORD SELECTION AND ESTIMAND

We selected the 21 rows of Trajectories_Data.xlsx containing nonmissing DunedinPACE_T0, DunedinPACE_T6, and DunedinPACE_T36 values. Original worksheet row numbers were retained as audit keys, not asserted to be verified participant IDs. No imputation, pregnancy-group reconstruction, or cell-composition residualization was performed. We estimated mean within-row changes between visits with paired t-tests and 95% confidence intervals. This estimand is distinct from the original adjusted nonlinear group-by-time interaction.

Comparison (n=21)Mean change95% CIPaired p
Prenatal → first+0.0586−0.0637 to +0.18090.329
First → final−0.0675−0.2096 to +0.07460.333
Prenatal → final−0.0089−0.1602 to +0.14250.904

Table 2. Previously reported exploratory complete-record DunedinPACE results (n = 21). Unadjusted paired comparisons; these do not test the published nonlinear pregnancy-by-time interaction. Numerical values require final independent code verification.

Three paired DunedinPACE changes in 21 complete records with 95% confidence intervals. All intervals include zero; prenatal-to-final mean change is minus 0.0089, with p equal to 0.904.

The supplied trajectory workbook contains 21 within-row-linked records with DunedinPACE values at all three nominal time points. The mean prenatal-to-first follow-up change was +0.0586 (95% CI −0.0637 to +0.1809; paired p=0.329). The mean first-to-final change was −0.0675 (95% CI −0.2096 to +0.0746; p=0.333). The mean prenatal-to-final change was −0.0089 (95% CI −0.1602 to +0.1425; p=0.904). These results do not show a statistically detectable net mean change within the complete-record subset, but do not establish equivalence or absence of an effect.

Among the 21 records, the workbook field AddPreg_T36 was positive for two and negative for 19. Their unadjusted prenatal-to-final mean changes were approximately −0.2828 and +0.0199, respectively (exploratory Welch p≈0.449). AddPreg_T18 was positive for three records, with mean change approximately −0.3565 versus +0.0490 in 18 other records (p≈0.176). These subgroup contrasts are severely underpowered and the fields have not been established as equivalent to the published PregAgain_Late exposure. They cannot be used to infer a protective effect of pregnancy.

The supplemental code specifies mixed models using natural splines, FamID, PregAgain_Late, PregAgain_Early, and covariates. The trajectory workbook does not explicitly supply the same analysis-ready participant IDs or pregnancy classification, and our unadjusted complete-record contrasts do not test the published interaction. The observed difference between our descriptive result and the original interaction therefore cannot be characterized as a demonstrated statistical error. Before publication, a second analyst should reproduce our numeric results and assess whether a separate publicly deposited analysis-ready dataset resolves the linkage gap.

6. DISCUSSION

The empirical evidence supports a reproductive-state-associated change in aging-clock outputs, not a proven acceleration of persistent organismal deterioration. Even exact reproduction of the published DunedinPACE interaction would not establish that its change during pregnancy reflects the same processes as its change in nonpregnant longitudinal aging cohorts.

Reproducibility dimensionObserved limitationConsequence
Participant linkageWithin-row longitudinal data; no explicit original FamIDOriginal cohort uniqueness cannot be independently checked
Exposure definitionPregnancy indicators not unambiguously mapped to PregAgain_LatePublished interaction cannot be exactly reconstructed
Completeness21 complete three-visit DunedinPACE rowsRestricted analysis has low exposed-subgroup precision
Biological constructMethylation-derived pace estimator applied in pregnancyClock response is not direct evidence of lasting deterioration

Table 3. Traceability and construct-validity limitations of the disclosed reanalysis materials. These concerns do not themselves establish statistical misconduct or falsification.

The statistical distinction is equally important. A significant time-by-pregnancy spline coefficient is not equivalent to a significant mean difference across all three visits. Conversely, a nonsignificant paired mean change in 21 complete records does not refute the interaction estimated using a larger incomplete-case sample and different covariate adjustments. The restricted analysis identifies uncertainty and limited independent corroboration, not falsification of the authors’ model.

A decisive prospective study would begin before conception, include age-matched nonpregnant controls, standardize gestational and lactation state, measure cell composition, use validated tissue-specific assays, and connect biomarker trajectories to durable functional and clinical outcomes. Interpregnancy interval would need to be modeled alongside health selection, socioeconomic resources, nutritional status, pregnancy complications, and baseline physiology.

This distinction matters for clinical translation. Recommending reproductive spacing on the basis of an unvalidated pregnancy-specific interpretation of a clock could confuse adaptive physiology with damage. Conventional obstetric evidence for pregnancy spacing remains important; the present critique addresses only claims that aging-clock shifts provide an independently validated mechanistic measure of cumulative aging.

The evidentiary basis of the Albright–Poganik postpartum recovery hypothesis warrants particular scrutiny. Etzel et al. reported a statistically significant later-pregnancy-by-time interaction for DunedinPACE, but not for telomere length, GrimAge2, or PhenoAge. The subsequent interpretation of this isolated significant interaction as evidence of cumulative maternal biological aging therefore lacks statistically significant corroboration from the other three examined biomarkers. Our independent examination of the disclosed trajectory workbook further identified incomplete participant traceability, uncertainty in reconstructing the published pregnancy classification, and only 21 complete three-visit DunedinPACE records. Within this restricted subset, the prenatal-to-final mean difference was not statistically significant. Although this analysis does not replicate or refute the original adjusted nonlinear interaction, it provides no independent confirmation of sustained DunedinPACE elevation. More fundamentally, neither the reported interaction nor our restricted analysis independently establishes persistent biological deterioration. Consequently, the evidence presented does not substantiate the broader claim that pregnancy itself constitutes accelerated or cumulative biological aging.

Four-panel comparison of Etzel 2026 published findings, the restricted 21-record DunedinPACE reanalysis, data traceability limitations, and implications for the postpartum biological-aging hypothesis. CAPTION: Figure 6. Critical Examination of Etzel et al. (2026): Published DunedinPACE Findings, Independent Reanalysis, and Implications for the Postpartum Biological Aging Hypothesis.

7. LIMITATIONS

This review is focused, not systematic. The reanalysis is restricted to a supplied wide-format workbook rather than the full original study database. Row-level linkage is internally coherent but not independently authenticated. The complete-case sample is small, exposed subgroups are extremely small, and selection bias is possible. The published outcome transformations and pregnancy categories could not be reconstructed exactly from the reviewed workbook. No claim of fabrication, misconduct, or mathematical error is justified by these findings.

8. CONCLUSION

Pregnancy-associated biomarker changes are demonstrable, but equating them with accelerated biological aging is not warranted without independent evidence of lasting functional deterioration. The restricted DunedinPACE analysis does not corroborate a net increase across three visits, while the original study’s sole significant biomarker interaction remains a different, incompletely replicated statistical contrast. The postpartum recovery hypothesis remains plausible as a research question, but its aging interpretation is not established by the available evidence.

DECLARATIONS

Author contributions

Bomi Joseph: Conceptualization, article conception and design and development of the manuscript outline.

Kai Weng: Research and investigation, source identification, collection of supporting documents, compilation of the reference literature and writing of the manuscript.

Both authors reviewed and approved the final manuscript.

Funding

This work was supported by Peak Health Center, San Francisco, California, USA (ROR: https://ror.org/04b23x642).

Competing interests

Bomi Joseph is the inventor of the Deep Health® Device, a health-measurement technology. Kai Weng declares no competing interests.

Data availability

This study presents an independent secondary examination of the supplementary data published by Etzel et al. (2026), Maternal biological aging from pregnancy through four years postpartum with disrupted recovery by subsequent pregnancy, iScience, 29, 116546. DOI: 10.1016/j.isci.2026.116546.

The original supplementary files examined included Trajectories_Data.xlsx and mmc3.xlsx. The accompanying Zenodo dataset, Etzel 2026 — Independent DunedinPACE Reanalysis, provides the extracted longitudinal measurements, a complete-record selection audit, within-record differences, recalculated descriptive statistics, paired statistical comparisons, 95% confidence intervals, and documentation of the analytical methods and limitations.

Of the 264 rows in the disclosed trajectory workbook, 21 contained complete DunedinPACE measurements at all three nominal observation points (T0, T6, and T36). These 21 within-row-linked records formed the basis of the restricted paired analysis.

The disclosed trajectory workbook did not contain explicit original participant identifiers, preventing independent verification of participant uniqueness against the original study cohort. Additionally, the available pregnancy indicators could not be unambiguously mapped to the published later-pregnancy classification. Consequently, the independent analysis does not reproduce the original adjusted nonlinear pregnancy-by-time interaction model.

The secondary analysis detected no statistically significant net prenatal-to-final mean DunedinPACE change in the 21 complete records (mean difference −0.0089; 95% CI −0.1602 to +0.1425; p = 0.904). This finding is limited to the restricted complete-record sample and should not be interpreted as a direct statistical refutation of the published interaction.

The original supplementary datasets remain attributable to Etzel et al. The derived analysis workbook, documentation, and reproducibility information are made available through the accompanying Zenodo record.

Zenodo DOI: 10.5281/zenodo.23630711

Use of artificial-intelligence tools

Artificial-intelligence tools were used in the creation of some of the figures.

REFERENCES

[1] Albright SM, Poganik JR. The postpartum recovery window as a determinant of maternal aging. Trends Endocrinol Metab. 2026. doi:10.1016/j.tem.2026.09.002.

[2] Etzel L, et al. Maternal biological aging from pregnancy through four years postpartum with disrupted recovery by subsequent pregnancy. iScience. 2026;29:116546. doi:10.1016/j.isci.2026.116546.

[3] Poganik JR, et al. Biological age is increased by stress and restored upon recovery. Cell Metab. 2023;35:807–820.e5. doi:10.1016/j.cmet.2023.03.015.

[4] Moran R, Pridham G, Toledano Y, Alon U. Pregnancy lab test dynamics resemble rejuvenation of some organs and aging of others. Nat Commun. 2026;17:2838. doi:10.1038/s41467-026-69340-0.

[5] Ying K, et al. Causal Epigenetic Age Uncouples Damage and Adaptation. bioRxiv preprint. 2022. doi:10.1101/2022.10.07.511382. Published version: Nat Aging. 2024. doi:10.1038/s43587-023-00557-0 (final text not cross-checked against supplied preprint).

[6] Gurven M, et al. Health costs of reproduction are minimal despite high fertility, mortality and subsistence lifestyle. Sci Rep. 2016;6:30056. doi:10.1038/srep30056.

[7] Sehgal R, Borrus D, Armstrong JF, et al. Responsiveness of epigenetic aging biomarkers to longevity interventions in humans. Nat Med. 2026;32:3477–3490. doi:10.1038/s41591-026-04562-9.

[8] Poganik JR, Moldakozhayev A, Justice JN, et al. Aging vs. biological age: Conceptual considerations for age reversal claims. Cell Metabolism. 2026;38(9):1746–1750. DOI: 10.1016/j.cmet.2026.06.015

SUPPORTING DATA · VOLUME 1 · ISSUE 2 · ARTICLE 5

Etzel 2026 — Independent DunedinPACE Reanalysis

Supporting data for Pregnancy-Associated Biomarker Changes Are Not Equivalent to Biological Aging. All seven sheets of the accompanying workbook are reproduced below, including the 21 complete records and the complete 264-row source audit.

These unadjusted complete-record comparisons do not reproduce the published adjusted nonlinear pregnancy-by-time interaction. Worksheet rows are provisional record keys, not independently verified original participant identities. Empty cells remain empty; visit dates remain Excel serial values.

Article and accompanying dataset · DOI 10.5281/zenodo.23630711

START HERE

Etzel 2026 — Independent Data Examination

Read this sheet first. This workbook distinguishes published regression outputs from independently recalculated complete-record comparisons.

SectionContentsHow to interpret
01 — Key findingsOur paired analysis of 21 complete within-row recordsNo statistically detectable net prenatal-to-final mean change; not the published spline interaction
02 — Paired testsThree comparisons, confidence intervals and p-valuesCalculated directly from disclosed DunedinPACE_T0/T6/T36 fields
03 — 21 complete recordsExact original row numbers, measurements, differences and pregnancy flagsWorksheet row is a provisional record locator, NOT a verified original participant ID
04 — All recordsAll 264 rows: DunedinPACE values, completeness and pregnancy indicatorsProvides transparent selection audit and preserves missing values
05 — Published modelsTranscription of published mmc3.xlsx D. Results interaction coefficientsPublished model output; not independently reconstructed
06 — Data dictionaryField definitions and interpretive cautionsDo not infer pregnancy grouping from AddPreg flags without source code
07 — Methods and limitsProvenance, calculation equations and missing-data restrictionsImportant limitations for scientific publication
Paired tests

Independent complete-record paired DunedinPACE comparisons

Two-sided paired t-tests, df=20; 95% t confidence intervals; these do NOT test the published later-pregnancy × spline interaction.

ComparisonnMean changeSE of changet statisticp-value95% CI low95% CI highInterpretation
Prenatal (T0) → first follow-up (T6)210.0586436551475140.05863121302372151.000212209898970.329156405354782-0.06365891208882730.180946222383855Not significant at 0.05
First follow-up (T6) → final follow-up (T36)21-0.06753161940256840.068127522908121-0.9912531165068740.33340396891523-0.2096431419416770.0745799031365398Not significant at 0.05
Prenatal (T0) → final follow-up (T36)21-0.008887964255054470.0725557453299424-0.1224984212433770.903726865537269-0.1602365969024450.142460668392336Not significant at 0.05
21 complete records

21 complete within-row DunedinPACE records

All values transcribed from the disclosed source workbook; visit dates are original Excel serial values; original subject identifiers are unavailable.

Source rowProvisional record labelT0 prenatalT6 firstT36 finalT6 − T0T36 − T6T36 − T0AddPreg T18AddPreg T30AddPreg T36Visit T0 (serial)Visit T6 (serial)Visit T36 (serial)
16Record at original worksheet row 161.291367454357151.114086534809351.34663702664913-0.17728091954780.232550491839780.05526957229198000434814372844733
18Record at original worksheet row 181.403201650644121.298129957901291.56642373052096-0.105071692742830.268293772619670.16322207987684000434764376144700
30Record at original worksheet row 301.765424664213211.885127040153721.75117863618530.11970237594051-0.13394840396842-0.0142460280279102000434794373444781
40Record at original worksheet row 401.738714540359711.865052968054081.721370360985610.12633842769437-0.14368260706847-0.0173441793741001440224427045375
58Record at original worksheet row 581.405626804047061.806564652757551.845501527964960.400937848710490.038936875207410.4398747239179000440344427645240
64Record at original worksheet row 641.364303598354341.442841291099241.296263367865670.0785376927449002-0.14657792323357-0.0680402304886698000435224384645038
65Record at original worksheet row 651.733545621489061.626916331545711.10863470996392-0.10662928994335-0.51828162158179-0.62491091152514100439864420445346
83Record at original worksheet row 831.47995823518821.645479097569181.560897443117390.16552086238098-0.08458165445178990.0809392079291902000435454384545030
94Record at original worksheet row 941.42233667456561.36989864836461.27116370944613-0.052438026201-0.09873493891847-0.15117296511947000436014386844869
97Record at original worksheet row 971.429974062533351.607738323497091.239962319613580.17776426096374-0.36777600388351-0.19001174291977000435434383244972
115Record at original worksheet row 1151.766406391615021.67033244530761.94919038807937-0.09607394630742010.278857942771770.18278399646435000435784383344837
117Record at original worksheet row 1171.820844474691.112691758698811.27264023500571-0.708152715991190.1599484763069-0.54820423968429001435814386145040
121Record at original worksheet row 1211.609597911152291.762620038506571.521471691520420.15302212735428-0.24114834698615-0.0881262196318700436024384845152
127Record at original worksheet row 1271.438594196179281.722607252995561.532456971937380.28401305681628-0.190150281058180.0938627757580999010435774387945109
137Record at original worksheet row 1371.04651859156310.8461329669566731.24592078200801-0.2003856246064270.3997878150513370.19940219044491000436414403445068
148Record at original worksheet row 1481.466189037637941.292782380026270.976295143130586-0.17340665761167-0.316487236895684-0.489893894507354100438184409245138
173Record at original worksheet row 1731.541783192442111.728555028840772.334352789436690.186771836398660.605797760595920.79256959699458000438854420545220
201Record at original worksheet row 2011.775528509486291.887357783677681.525114899510020.11182927419139-0.36224288416766-0.25041360997627000438374414045083
227Record at original worksheet row 2271.599944118786041.774573838019171.397269514241160.17462971923313-0.37730432377801-0.20267460454488000437854407745185
235Record at original worksheet row 2351.356582337934671.61898617336471.761615702510020.262403835430030.142629529145320.40503336457535000440014428245321
250Record at original worksheet row 2501.278262270177441.887746583368161.323696138367820.60948431319072-0.564050445000340.0454338681903801100440704432845358
All source records

Disclosed DunedinPACE completeness audit — all 264 rows

One row per original worksheet row. No claim that rows are verified unique individuals. Blank source values remain blank.

Source rowT0T6T36Observed of 3Complete all 3?AddPreg T18AddPreg T30AddPreg T36Visit T0 (serial)Visit T6 (serial)Visit T36 (serial)
20NO000433644362144656
30NO
41.907188083603161.293038565153812NO000433634362644676
51.473629903151691NO000433824365744693
61.550689392023531NO010433824364844726
70NO04338943666
80NO000440134435045236
90NO
100NO
111.330604676161861NO010433894367344681
120NO
131.631636687710691.611515977495712NO00434124366844662
141.160457122165151NO010434034368644693
151.546308473842871NO000439874422545246
161.291367454357151.114086534809351.346637026649133YES000434814372844733
170NO104348843743
181.403201650644121.298129957901291.566423730520963YES000434764376144700
190NO04402244257
201.353347750581371.9875676763042NO0004401144391
210NO04339843668
221.769204834184041.441915126469782NO000433714362944666
231.76408408525261.395214186719432NO000434204369344768
240NO0044203
250NO04343943676
260NO
271.882540246059531.392121808556212NO0004417645126
280NO
290NO0004343443720
301.765424664213211.885127040153721.75117863618533YES000434794373444781
311.808289586328361.125122073066482NO000434034367244663
320NO04343843729
330NO000434064366144738
340NO004339743684
350NO
361.68221715512311.208627225849972NO000434404372244801
370NO00439984425745645
380NO004350743806
391.728009436548791.76541843634252NO000434384371544793
401.738714540359711.865052968054081.721370360985613YES001440224427045375
410NO
420NO
430NO100434134368344673
440NO14349543778
450NO
460NO14341643728
470NO
480NO4341943699
491.73390292548921.578576729487972NO000434124370444786
501.570778015348981.269685075483822NO000435234381944881
511.34359327009561.310868763565612NO100434944377044816
520NO
531.783415680111211.978771580957722NO010440224425245318
541.730213740092591.572276441522072NO000434484372444694
550NO004351743791
560NO000434454371444701
570NO04344043707
581.405626804047061.806564652757551.845501527964963YES000440344427645240
590NO4351143818
601.351255029997711.654622188386512NO000435774385744862
610NO04346843747
621.65174700497591.270297687678642NO011435004380544879
630NO000434814375944772
641.364303598354341.442841291099241.296263367865673YES000435224384645038
651.733545621489061.626916331545711.108634709963923YES100439864420445346
661.84122553509431NO000434794376344845
672.043912397202321.371459936135712NO000439854422345204
681.468377398268181.800107142493372NO000434744374344815
690NO0004349343763
700NO
711.707528797339871NO010440074426645244
721.699173969323341NO000435654384144850
730NO14348743775
741.490095855821041.70553716073032NO000434974377744743
750NO0004417445245
761.456805352829451NO001435394383944841
770NO4350343791
780NO04353043813
790NO
800NO004354543818
810NO4350243810
821.559612703921111NO000435084379244886
831.47995823518821.645479097569181.560897443117393YES000435454384545030
841.549356545139411NO000435034376844819
850NO001435584384344847
860NO04351143785
871.530974373887131.089456084707832NO010435284380844964
881.513645071750581.924797370023842NO100438924421045243
891.570864040199881.596432375931632NO004351443806
900NO0
910NO
921.776001440718011.568821484878192NO000439924420445107
930NO
941.42233667456561.36989864836461.271163709446133YES000436014386844869
951.390104975228961NO001435254378944822
960NO004351743805
971.429974062533351.607738323497091.239962319613583YES000435434383244972
981.755512279883971.769374676888042NO004401244314
990NO004400744293
1000NO
1010NO
1020NO000439934423945394
1030NO004388944106
1041.481851561573961NO0004353243811
1050NO
1061.669006961859241NO01435854388345114
1070NO
1080NO
1090NO
1101.856990468757441NO0004361345153
1111.740191302585631.608068047988322NO010436064389244861
1120NO0004359445121
1130NO043601
1140NO
1151.766406391615021.67033244530761.949190388079373YES000435784383344837
1160NO
1171.820844474691.112691758698811.272640235005713YES001435814386145040
1180NO4356043854
1191.716926557527521.816677032495942NO000436854402145178
1201.456726795247211.610279518619762NO000436214389645086
1211.609597911152291.762620038506571.521471691520423YES00436024384845152
1221.8141032032581NO0143620
1230NO
1240NO
1251.761785515876791.776941272332792NO004356143855
1260NO
1271.438594196179281.722607252995561.532456971937383YES010435774387945109
1280NO000435954386945171
1291.305567231773931.29031124649962NO100436824402845106
1300NO04361443895
1311.853869294516791NO001440394428645328
1321.61823555855851.112054388525682NO00436274404245096
1330NO
1340NO004359843876
1351.810701076696181NO0044249
1361.324457010670041.20548594378782NO000436784404044876
1371.04651859156310.8461329669566731.245920782008013YES000436414403445068
1380NO0043572
1390NO
1400NO
1410NO043630
1420NO043665
1431.506803783594611.270818631805572NO000436274409045084
1440NO
1450NO04370544061
1461.665246737774371NO10440294428745233
1470NO
1481.466189037637941.292782380026270.9762951431305863YES100438184409245138
1490NO
1501.391011423048811NO043665
1510NO
1520NO
1531.37160268719511.317823238262872NO0004364445016
1541.740664537746281NO000438944425245276
1551.355367988985511NO000439854424445124
1560NO4400044358
1570NO000436564406044907
1580NO043626
1590NO043623
1600NO000438764416845262
1610NO
1621.378290698032111.313189788870842NO00437424407545145
1630NO4364144091
1641.91801990917261NO000440624427845346
1650NO00436434405645197
1660NO
1670NO
1680NO
1691.526746346156411NO00436754390345018
1700NO004371344096
1710NO
1721.3355415356881.863342330209772NO000437284409745157
1731.541783192442111.728555028840772.334352789436693YES000438854420545220
1740NO043673
1750NO43742
1760NO
1770NO43677
1780NO43676
1790NO
1801.284913744535991.2518518189672NO000437884408645076
1810NO
1820NO
1831.640820240569271NO0437414409145078
1841.582687910699441NO0043754
1851.30549815203271NO104390144243
1860NO
1871.756857831932471.909326427154072NO00
1880NO004371744056
1891.750708793132321NO04398744276
1900NO04378344085
1910NO0043836
1921.365870919406131.386386520066012NO100438954421245126
1931.726313084739581NO000439864421845115
1940NO04378044202
1950NO
1961.379340500075391NO000438894410645166
1970NO
1981.601386582769661.284159548309412NO000438094410545079
1991.684753871453341.704001862386062NO000438104411745092
2000NO000437894421645174
2011.775528509486291.887357783677681.525114899510023YES000438374414045083
2021.785169060827991.653298262499482NO04401244273
2031.247049605758311NO004381744114
2041.890308735990281NO043815
2050NO
2060NO
2071.600805528652111.921563339761142NO000438674414745160
2080NO0
2090NO
2101.486796717510681.755470818763932NO104380844090
2110NO0
2121.781403082598751NO010439874423245175
2131.468160412189311.291761170812172NO000438154413245211
2140NO0
2150NO04399044224
2161.915424318844241NO004388844174
2170NO044305
2180NO
2191.431009719074911NO004416745112
2200NO4376144114
2210NO
2220NO
2230NO
2240NO043776
2250NO0
2260NO00
2271.599944118786041.774573838019171.397269514241163YES000437854407745185
2280NO
2291.405944276028731.692377177446562NO14385244130
2300NO1044007
2310NO
2320NO
2330NO
2340NO0004381045201
2351.356582337934671.61898617336471.761615702510023YES000440014428245321
2360NO
2370NO100439984425345324
2381.883976591566621NO100440014423145404
2391.299554067713911.745998713840162NO000440224425845238
2400NO44230
2411.318003019841561.842229640769682NO04402644274
2421.614581550512161.452277924107132NO100440434428645258
2431.817823751839611.709941261741092NO0004410445386
2441.373120450725241.608627002893412NO0004406944310
2451.379734480058571.708421852149322NO004406944305
2461.493586307083831.932830461072NO004411244371
2470NO
2481.27965537481131.774583811317212NO000441264442345397
2490NO04410644404
2501.278262270177441.887746583368161.323696138367823YES100440704432845358
2511.43524437527341.905996100830982NO000440764431745260
2520NO
2531.505985106590271NO000440914439445353
2540NO4412444419
2550NO
2560NO
2570NO004413444322
2580NO00
2590NO001
2600NO00
2610NO0
2620NO00
2630NO000
2640NO0
2650NO0
Published models

Etzel supplemental model coefficients — verbatim source table

Source: mmc3.xlsx, D. Results. Values are published estimate (SE) strings; stars denote source significance markings, not independently calculated p-values.

OutcomeNo interaction: spline 1No interaction: spline 2No interaction: later pregnancyInteraction: spline 1 × pregnancyInteraction: spline 2 × pregnancyWith interaction: spline 1With interaction: spline 2With interaction: later pregnancy
aTL0.49 (.54) -0.81 (.37)* -0.55 (.42)0.27 (1.68) -1.27 (.97)0.452 (.57) -0.58 (.40) -0.55 (.42)
GrimAge -2.11 (1.06)*0.77 (.80) -0.81 (.76) -1.06 (3.80)2.97 (2.02) -1.94 (1.11)0.21 (.87) -0.832 (.77)
PhenoAge -2.22 (2.20) -0.16 (1.64) -1.17 (1.51)5.29 (7.80)6.21 (4.21) -2.47 (2.29) -1.36 (1.81) -1.22 (1.51)
DunedinPACE -0.24 (.04)***0.05 (.02) -0.03 (.03) -0.04 (.13)0.20 (.07)** -0.23 (.04)***0.01 (.03) -0.03 (.03)
Data dictionary

Variables and definitions

The source field names are preserved. Descriptions reflect the available workbook, not an independently verified codebook.

FieldMeaningImportant qualification
Source rowOriginal row number in Trajectories_Data.xlsxLocator only; not Subject ID/FamID
DunedinPACE_T0Prenatal DunedinPACE valueUnadjusted disclosed score
DunedinPACE_T6First follow-up DunedinPACE valueT6 is nominal study label; actual elapsed time may vary
DunedinPACE_T36Final follow-up DunedinPACE valueT36 is nominal study label; actual elapsed time may vary
T6 − T0Within-row difference: first minus prenatalCalculated from source values
T36 − T6Within-row difference: final minus firstCalculated from source values
T36 − T0Within-row difference: final minus prenatalCalculated from source values
AddPreg_T18/T30/T36Additional pregnancy indicator fields in disclosed sourceDo not assume they reproduce published PregAgain_Late variable
VisitDate_T0/T6/T36Original numeric Excel date serialsRetained as source values; dates not inferred
mmc3.xlsx D. ResultsPublished spline regression coefficient summariesNot independently replicated from raw individual-level model inputs
Methods and limits

Reproducibility and interpretive limits

For manuscript / Zenodo readers: provenance and restrictions on inference.

ItemDescription
Data sourceEtzel et al. (2026), supplementary Trajectories_Data.xlsx and mmc3.xlsx provided with the publication.
Source rows264 data rows in Trajectories_Data.xlsx (Excel rows 2–265).
SelectionRows where DunedinPACE_T0, DunedinPACE_T6 and DunedinPACE_T36 are all numeric; n = 21.
Record linkageWithin-row measurements treated as a provisional longitudinal record. Original participant IDs absent; uniqueness cannot be verified independently.
StatisticsFor each record, calculate paired change. Mean, sample SD, SE=SD/sqrt(n), t=mean/SE, two-sided Student t p-value (df=20), 95% t CI.
Published model comparisonThe paired analysis is not a reproduction or refutation of the adjusted nonlinear pregnancy-by-time spline interaction in mmc3.xlsx.
Exposure classificationPregnancy indicators are retained but not recoded to PregAgain_Late because the mapping was not sufficiently documented.
MissingnessComplete-record analysis can be affected by selection bias and is not representative of all 264 source rows by assumption.
Meaning of nonsignificanceA p-value >0.05 does not prove absence of a biological relationship or equality of groups.
Biological inferenceA changing methylation-derived score alone does not demonstrate lasting biological deterioration or rejuvenation.
No fabricated dataNo imputation, guessed participant identities, invented pregnancy labels, or synthetic observations have been added.
Reproduction noteThis workbook is a documented secondary analysis. The original author-supplied supplementary files should be separately cited and linked, not represented as newly collected data.

Source workbook: Etzel_2026_Independent_Reanalysis_Zenodo.xlsx. Published model coefficients are reproduced as source strings; they are distinct from the independently recalculated paired comparisons. Original supplementary data remain attributable to Etzel et al.