Brain activity during Stroop task performance at age 74 after exposure to the Dutch famine during early gestation

Objective: Poorer performance on the Stroop task has been reported after prenatal famine exposure at age 58, potentially indicating cognitive decline. We investigated whether brain activation during Stroop task performance at age 74 differed between individuals exposed to famine prenatally, individuals born before and individuals conceived after the famine. Method: In the Dutch famine birth cohort, we performed a Stroop task fMRI study of individuals exposed (n = 22) or unexposed (born before (n = 18) or conceived after (n = 25)) to famine in early gestation. We studied group differences in task-related mean activation of the dorsolateral prefrontal cortex (DLPFC), anterior cingulate cortex (ACC) and posterior parietal cortex (PPC). Additionally, we explored potential disconnectivity of the DLPFC using psychophysiological interaction analysis. Results: We observed similar activation patterns in the DLPFC, ACC and PPC in individuals born before and individuals exposed to famine, while individuals conceived after famine had generally higher activation patterns. However, activation patterns were not significantly different between groups. Task-related decreases in connectivity were observed between left DLPFC-left PPC and right DLPFC-right PPC, but were not significantly different between groups. Conclusions: Although not statistically significant, the observed patterns of activation may reflect a combined effect of general brain aging and prenatal famine exposure.


Introduction
The Stroop Color-Word Interference test is a well-established cognitive task that measures an individual's ability to inhibit automatic responses and selectively focus attention on relevant stimuli.This task is based on the Stroop interference effect, where the ability to name the ink color of a word is conflicting with the meaning of the word by presenting color words printed in an incongruent ink color (e.g. the word "yellow" printed in blue ink).By introducing conflicting information, the Stroop task challenges the cognitive system to inhibit the prepotent response of automatic word reading and instead focus on naming the incongruent ink color (Kahneman & Chajczyk, 1983;MacLeod & Dunbar, 1988;Scarpina & Tagini, 2017;Stroop, 1935).Thereby, the Stroop task provides a means to measure selective attentional control, interference and response inhibition as domains of executive functioning, giving insight into an individual's cognitive flexibility (Langenecker, Nielson, & Rao, 2004).
The Stroop task activates brain regions which are mainly involved in attentional control and response inhibition.Activity patterns of these brain regions during Stroop task performance have been observed to be different in elderly people and in individuals with MCI, suggesting a change of these activity patterns with increasing age.Brain regions related to Stroop task performance and brain aging include the dorsolateral prefrontal cortex (DLPFC), anterior cingulate cortex (ACC), inferior frontal gyrus (IFG) and posterior parietal cortex (PPC).The DLPFC and ACC are important in task inhibition and showed higher activation during the (numerical) Stroop task in a population with MCI compared to healthy controls (Kaufmann et al., 2008;Mostofsky et al., 2003;Parris, Wadsley, Hasshim, Benattayallah, Augustinova, & Ferrand, 2019;You, Shahar, Mohamad, Yahya, Haron, & Abdul Hamid, 2019).Previous studies suggested a pathway in which the ACC signals the presence of conflicting information to the DLPFC, which subsequently imposes cognitive control for conflict resolution (Miller & Cohen, 2001;Parris et al., 2019;van Veen & Carter, 2002).The IFG is part of the DLPFC and shows higher activation in elderly participants compared to young controls during Stroop task performance (Langenecker, Nielson, & Rao, 2004;Milham et al., 2002;Nielson, Langenecker, & Garavan, 2002).The IFG is specifically thought to be of importance in attentional setting, regulating the focus on either color or word processing (Parris et al., 2019).Lastly, the PPC is associated with response inhibition, visuospatial selection and conflict resolution and has higher activation in individuals with MCI during an executive task (Liu, Banich, Jacobson, & Tanabe, 2004;Nee, Wager, & Jonides, 2007;Parris et al., 2019).Thereby, activity within these brain regions during Stroop task performance can serve as an indicator of ongoing processes related to cognitive aging.
Cognitive aging exhibits substantial inter-individual variability.Some individuals can maintain cognitive health well into old age, whereas others experience varying degrees of cognitive decline.Environmental factors may be a large contributor to shaping these interindividual differences, particularly during the critical period of prenatal development and early postnatal life.During this period of rapid brain growth, environmental factors can shape brain development through numerous pathways.For instance, availability of essential nutrients for brain cell formation significantly impacts early brain growth and brain reserve (Georgieff, 2007;Godfrey & Barker, 2001;Marques, Bjørke-Monsen, Teixeira, & Silverman, 2015).Furthermore, prenatal programming of the immune system and epigenetic, endocrine and metabolic pathways can alter brain health and cognition across the lifespan (Cao-Lei et al., 2020;Marques et al., 2015;Moreno-Fernandez, Ochoa, Lopez-Frias, & Diaz-Castro, 2020).
Through these pathways, prenatal nutrition can impact cognitive and brain aging.Results from the Dutch famine birth cohort (DFBC) have suggested that men and women exposed to undernutrition in early gestation may be prone to accelerated brain aging.In their late fifties, exposed participants of the DFBC performed worse on an adapted Stroop Color-Word Interference test, having a longer response time and lower accuracy (de Rooij, Wouters, Yonker, Painter, & Roseboom, 2010).Moreover, exposed men had lower cerebral blood flow in regions related to neurodegeneration and displayed premature brain aging on the BrainAGE (brain age gap estimation) biomarker (de Rooij et al., 2019;Franke, Gaser, Roseboom, Schwab, & de Rooij, 2018).Altered restingstate functional connectivity was observed in the DFBC in both men and women at age 68 (Boots et al., 2022).Taken together, these results indicate a potential process of accelerated cognitive and brain aging after exposure to undernutrition during early gestation.However, taskrelated brain activation has not been studied in the DFBC.Studying brain activation during Stroop task performance can give further insight into the neural mechanisms underlying cognitive and brain aging in these individuals.
The current study investigated whether late-life brain activation during the Stroop Color-Word Interference test differed between individuals who had or had not been exposed to famine prenatally.First, we performed a region of interest (ROI) analysis using the regions listed above (DLPFC/IFG, ACC, PPC).As previous studies in the DFBC suggested potential accelerated brain aging after prenatal exposure to famine, we hypothesized higher activation levels in the selected ROIs in exposed participants compared to unexposed controls, similar to individuals with MCI.Secondly, we performed an exploratory psychophysiological interaction (PPI) analysis to study the changes in functional connectivity between brain regions during Stroop task performance.Since a functional disconnection of the DLPFC during restingstate has been reported in patients with MCI, we studied the group differences in functional connectivity between the DLPFC and the ACC and PPC (Liang, Wang, Yang, Jia, & Li, 2011).We hypothesized that a functional disconnectivity of the DLPFC similar to patients with MCI may be observed in participants exposed to prenatal undernutrition.Given the previously identified sex-specific effects in the DFBC, we additionally explored sex-specific effects in the current study.

The Dutch famine birth cohort
The Dutch famine birth cohort consists of 2414 men and women who were born between November 1, 1943, andFebruary 28, 1947, at the Wilhelmina Gasthuis in Amsterdam, the Netherlands.All participants met the following inclusion criteria: born as a singleton, minimum pregnancy duration of 259 days, and availability of a medical birth record.
2.1.1.1.Exposure.The Dutch famine was a direct consequence of events which occurred at the end of World War II.During the Dutch famine, which took place between December 1944 and April 1945, official daily food rations varied between 400 and 800 calories.After May 12, 1945, rations increased to more than 1000 calories.In the DFBC, individuals were considered to be exposed to prenatal undernutrition if the average daily food ration was below 1000 calories during any 13-week period of gestation.16-week periods were defined to represent exposure during early, mid, and late gestation.Only participants exposed to undernutrition during early gestation (born between August 19 and December 8, 1945) were included in the neuroimaging studies since most effects on health outcomes have been observed in this study group (Bleker, de Rooij, Painter, Ravelli, & Roseboom, 2021).Individuals born before January 7, 1945, and conceived after December 8, 1945, were included in the study to serve as control groups which were not exposed to undernutrition during gestation.See Bleker et al. for a detailed description of the DFBC (Bleker et al., 2021).The study was approved by the local medical ethics committee and carried out according to the Declaration of Helsinki.All participants provided written informed consent.

Study sample
Data for the current MRI study were collected between June 2019 and August 2020 as a longitudinal follow-up of the neuroimaging study A. Boots et al. performed in 2012-2013.In 2012, a random subsample of the initial cohort was invited to participate in a home visit and a brain MRI scan.MRI data was collected for 118 individuals.In 2019, when participants were 73-75 years of age, we aimed to include all available participants from the 2012 study for a hospital visit encompassing extensive cognitive testing, blood draw and a brain MRI scan.Neuroimaging data was collected for 70 of the original 118 individuals.To increase the number of participants in 2019-2020, 11 additional eligible cohort members were included who completed MRI scanning (64 invited, 27 % participation rate including participation without MRI).Of the 81 participants who were scanned, 66 were able to successfully complete the Stroop fMRI task (excluded born before n = 8, early exposed n = 3, conceived after n = 4).Reasons for exclusion were blindness or visual difficulties (n = 3), self-reported color blindness (n = 2) and difficulties with task (n = 7) or buttons (n = 3).

Study parameters
Parental socioeconomic status (SES) was obtained from medical birth records and was defined as the occupation (manual yes/no) of the head of the household.To obtain adult characteristics, we conducted a standardized interview during the hospital visit in 2019.Educational level was measured on a 10-point scale (1 = primary education not completed, 10 = university completed).We defined smoking as ever having smoked.In case of inconsistencies between earlier reports of smoking in previous data collections and our most recent interview, the participant was marked as 'yes' for 'ever smoked' if they had confirmed ever smoking in any interview.SES was based on the participant's or their partner's (former) occupation, whichever was highest, according to the International Index of Occupational Status 92 (ISEI-92).

Stroop task. The Stroop fMRI task was specifically designed for
the present study.To ensure the measurement of brain activity during the correct performance of the Stroop paradigm, task difficulty level was set to aim for high accuracy levels across all participant groups.The task consisted of a block design with alternating blocks of a 'squares' control conditionnaming a colored squareand a 'words' conditionnaming an incongruent word color (Stroop).The full task included eight blocks of four trials for each condition, with a total of sixteen blocks.The order of trials was identical for all participants (Fig. 1).The blocks lasted 22.1 s each.Two response boxes were used for the four answer options, where participants used the index and middle finger of both hands to press the buttons.Due to a technical scanner error, the scan ended during the last block of 'squares' resulting in incomplete fMRI data for this block, and no data for the eighth 'words' block.fMRI data analysis therefore included seven full blocks of 'words' trials, and seven blocks plus two additional trials of 'squares'.
Participants did not undergo the fMRI scan if they were unable to complete the instruction task which served as a checkpoint before starting the actual task.The instruction task consisted of two 'squares' trials and two 'words' trials.Three out of four trials had to be answered correctly in order to successfully complete the instruction task before moving on to the Stroop task.If participants failed to pass the instruction task repeatedly (after 3-4 trials), the protocol was aborted (n = 9).
In addition to BOLD fMRI signal, task accuracy and response time were recorded.Accuracy was defined as % correct trials, response time was calculated as the average response time for all correct trials excluding trials with a response time < 200 ms (n = 1), for 'squares' and 'words' trials separately.

Data analysis 2.3.2.1. Data quality assurance and preprocessing.
Participants were excluded if their mean framewise displacement (FD) was 1 mm or above (n = 1) (Power, Barnes, Snyder, Schlaggar, & Petersen, 2012).We additionally performed motion scrubbing using a threshold of FD > 2 mm by adding a confound regressor to the first level analysis.Participants were excluded from further analysis if scrubbing resulted in the removal of > 20 % of their data (n = 0).Preprocessing of the data was performed using fMRIPrep including automatic removal of motion artefacts using ICA-AROMA (Esteban et al., 2019;Pruim, Mennes, van Rooij, Llera, Buitelaar, & Beckmann, 2015).See supplementary methods for the complete methods text provided by fMRIPrep.We performed visual inspection of raw and preprocessed data for artefacts or irregularities.
Fig. 1.Stroop task design.The task consisted of a block design with alternating blocks of a 'squares' control conditionnaming a colored squareand a 'words' conditionnaming an incongruent word color.The full task included eight blocks of four trials for each condition, with a total of sixteen blocks.

Regions of interest.
The ROIs were created using the Neurosynth uniformity test voxel-wise z-score map of an automated metaanalysis of 215 Stroop-task studies (search term 'Stroop';Yarkoni, Poldrack, Nichols, Van Essen, & Wager, 2011).This activation map contains an overview of the general pattern of activation observed in Stroop task fMRI studies.Similar to McDermott et al. (2020), the uniformity test zmap was thresholded at a z-score of 5 to create separate clusters (McDermott et al., 2020).For these clusters, the peak activation coordinate was extracted for the clusters located in the areas corresponding to the DLPFC (Brodmann areas 9 and 46; You et al., 2019), the ACC (Brodmann areas 24 and 32; (Lopez, 2015)) and the PPC (Brodmann areas 5,7,39 and 40; (Whitlock, 2017)) using the FMRIB's Software Library (FSL) cluster tool.Subsequently, we constructed spherical ROIs with an 8 mm radius around these peak coordinates using fslmaths (Supplementary Fig. 1).The peak activation coordinate of the ACC was located in the right hemisphere which did not fully represent the bilateral activation of the ACC.Therefore, we mirrored the peak coordinate along the y-axis and created two unilateral spherical ROIs which were then combined to create a bilateral ROI for the ACC.The resulting spherical ROIs were visually inspected to be within the task activation zscore map of our dataset and overlapping with the anatomical regions of interest.ROIs were defined unilaterally for the DLPFC and PPC and tested for statistically significant differences between hemispheres.Given the significant differences between hemispheres, unilateral ROIs were used in the analyses.
For the exploratory PPI analysis, the first-level COPE maps of the main task contrast (words > squares) were used as a psychological regressor of interest, and subject-specific DLPFC time-courses were used as a physiological regressor of interest in the model.Subject-specific time courses were extracted for the left and right DLPFC using fslmeants.An interaction term between the psychological and physiological regressor was created as described here (PPIHowToRun − FslWiki (ox.ac.uk);Jenkinson et al., 2012;Woolrich et al., 2001).Similar to the ROI analysis, between-ROI connectivity was extracted using FSL FEATQuery.

Statistical analysis.
Statistical testing was performed in SPSS Statistics v28 (IBM, Chicago, USA).All Stroop task fMRI analyses were performed on raw scores using the BOLD signal contrast between control (squares) and incongruent (words) blocks.Normally distributed study group characteristics and task outcomes (response time) were analyzed using ANOVA or chi-squared analyses.Accuracy was rank-transformed to account for non-normality.We used one-sample Wilcoxon signed rank tests to test if task-related connectivity in the PPI analysis across groups was significantly different from zero.ROI and PPI group differences were analyzed using a Kruskal-Wallis test because of nonnormality and no equality of variances across groups.We additionally ran a model corrected for covariates (parental SES, smoking and education) for the ROI analysis by rank-transforming the ROI activation values and performing ANCOVA analyses using these transformed outcomes as dependent variables.The born before and conceived after control groups could not be combined as group means and medians were not similar.Therefore, analyses were performed comparing the three study groups separately.We employed a significance threshold of α = 0.05.

Transparency and openness
The analysis protocol of this study was pre-registered on the Open Science Framework (OSF; https://osf.io/7z3mv/).Code and ROI files are available on OSF (https://osf.io/htsk7/).Data is available upon request due to privacy restrictions.We follow JARS-quant reporting guidelines.

Study group characteristics
We collected Stroop task fMRI data in 22 exposed and 44 control subjects of the DFBC at age 74-76.One subject was removed because of excessive motion (mean FD > 1.0 mm), resulting in a total of 65 subjects.No differences in mean FD after scrubbing were observed between groups.Comparing characteristics between individuals born before the famine, exposed to famine in early gestation and conceived after the famine in this cohort subsample, we observed a significant difference in age across groups (by definition), with individuals born before the famine having the highest average age, and individuals conceived after the famine having the lowest average age.No other group differences in baseline characteristics were observed (Table 1).
Data are shown as mean (SD) or as frequencies (%).ANOVA or chisquared analyses were used to create p-values for differences between study groups born before the famine, exposed to famine in early gestation and conceived after the famine.Key: SES = socioeconomic status.

Stroop task analysis
As intended by design, participants performed well on the Stroop task and both accuracy and response time scores were similar across the three study groups (Table 2).Based on visual comparison, the Stroop task whole-brain activation pattern across all subjects was similar to the pattern reported by a meta-analysis of 215 Stroop task fMRI studies, confirming that we successfully measured Stroop task-related brain activity ( (Yarkoni et al., 2011), Supplementary Fig. 3).The three study groups had comparable activation patterns (Supplementary Fig. 4).

ROI analysis outcomes
No statistically significant differences in median ROI activation during Stroop task performance were observed between individuals born before the Dutch famine, exposed to famine in early gestation and conceived after the famine across all participants (Table 3, Fig. 2) or in men and women separately (Table 4, Supplementary Fig. 5 and 6).Adjusting for covariates did not alter the results.Median ROI activation scores of the group born before the famine were generally similar to median scores of the group exposed to famine in early gestation, whereas median ROI activation scores in the group conceived after the famine were higher, with large individual variability.

PPI analysis outcomes
Task-related decreases in connectivity were observed between the left DLPFC-left PPC (significant; p = 0.005) and the right DLPFC-right PPC (near significance; p = 0.065).For these regions, no statistically significant differences in median PPI scores during Stroop task performance were observed between individuals born before the Dutch famine, exposed to famine in early gestation and conceived after the famine across all participants (Table 5, Fig. 3) or in men and women separately (Table 6, Supplementary Fig. 7 and 8).

Discussion
In this study, we aimed to investigate the association between prenatal undernutrition and late-life brain activation during the Stroop Color-Word Interference test.Although we had hypothesized patterns of higher ROI activation during Stroop task performance in individuals exposed to the Dutch famine compared to individuals born before or conceived after the famine, we observed similar patterns of activation of the DLPFC, ACC and PPC in individuals born before the famine and exposed to famine in early gestation.None of the differences between our study groups reached statistical significance.Task-related connectivity was observed between the left DLPFC-left PPC and right DLPFCright PPC, but was not significantly different between individuals who had or had not been exposed to famine prenatally.

Interpretation of ROI analysis
While the activation of the investigated brain areas was not significantly different between groups and individual variability was large, noteworthy discrepancies in median ROI activation − particularly in the right PPC − were evident.We are of the opinion that these patterns may be of interest and are worth discussing, especially given the limited sample size and statistical power of our study.The trends in the median ROI activation mostly showed higher ROI activation in the individuals conceived after the famine compared to individuals born before the famine or exposed to famine during early gestation.The median activation score of the right PPC was nearly twofold higher in the conceived after group (15.2) compared to the early exposed group (8.6), although the IQR was broad in both groups.The individuals born before the famine and exposed to famine in early gestation had relatively similar activation scores.Thus, the group of individuals exposed to famine in early gestation appeared more similar to the older group born before the famine compared to the younger group conceived after the famine.Since all groups had comparable task activation maps, it is unlikely that these variations in median ROI activation are explained by different brain regions being recruited in each group or a lack of selectivity in recruitment as a result of brain aging (Supplementary Fig. 4).The nonsignificant tendencies of lower ROI activation in individuals exposed to famine in early gestation compared to those conceived after the famine are contrary to our hypothesis of higher ROI activation in the individuals exposed to famine in early gestation, although both increases and decreases in task-related brain activity are observed in aging and dementia, also depending on the stage of cognitive decline (Li, Zheng, Wang, Gui, & Li, 2009;Milham et al., 2002).Speculatively, the observed nonsignificant trends of lower task-related ROI activation in exposed individuals may point to a more pronounced ongoing process of brain aging in individuals exposed to famine in early gestation, similar to unexposed individuals who were on average 1.2 years older.Alternatively, these observations may potentially reflect developmental effects of prenatal undernutrition on brain functioning which have persisted into late life, as has been discussed previously in a resting-state functional connectivity analysis in the DFBC at age 68 (Boots et al., 2022).
Although it did not reach statistical significance, the observed pattern of aberrant outcomes in the individuals conceived after the famine deviates from previous findings in the DFBC.In previous studies investigating long-term health outcomes after prenatal famine exposure, individuals born before and conceived after the famine exhibited remarkable similarity and could be pooled into a single control group, while those exposed to famine during early gestation often displayed more unfavorable health outcomes (Bleker et al., 2021).In previous studies, combining both unexposed groups resulted in a similar average age as for the group of individuals exposed to famine in early gestation, which enabled the interpretation of study results in the context of exposure rather than age.It could be hypothesized that the patterns Data are shown as mean (SD) or as frequencies (%).ANOVA or chi-squared analyses were used to create p-values for differences between study groups born before the famine, exposed to famine in early gestation and conceived after the famine.Key: SES = socioeconomic status.observed in the present study reflect a combination of an effect of aging (with the group of participants born before the famine being 2 to 3 years older than the group of participants conceived after the famine) and prenatal famine exposure (with the group of participants exposed to famine in early gestation showing patterns that are similar to those in the older group of controls), although we cannot currently disentangle these factors.In addition to the unanticipated differences between unexposed group scores, we observed a large variability in individual ROI activation scores, with each group having a broad interquartile range (IQR) and the group of individuals conceived after the famine displaying the largest spread in data points.In the sex-specific ROI analysis, the abovementioned trends were visible in women, but not in men, and   findings seemed to be mainly driven by the high ROI activation levels in women conceived after the Dutch famine.However, as stated above, none of these group differences were statistically significant, which warrants the need for caution in trying to interpret these findings.

Interpretation of PPI analysis
We performed an exploratory PPI analysis to explore patterns of connectivity between ROIs.Disconnectivity of the DLPFC during restingstate has previously been associated with age-related changes in cognitive task performance (Liang et al., 2011).Therefore, we had hypothesized to observe disconnectivity of the DLPFC in individuals exposed to famine in early gestation compared to unexposed individuals.Our analysis showed significant task-related connectivity between the left DLPFC-left PPC and right DLPFC-right PPC, which did not differ between individuals exposed or unexposed to famine.Analyses across all subjects and sex-specific analyses did not reveal any clear patterns across groups, with PPI scores varying in direction and amplitude across the three exposure groups.Comparable to the ROI analyses, IQRs were broad with large individual variability.Taken together, the current study did not reveal any patterns of decreased connectivity of the DLPFC during Stroop task performance after prenatal undernutrition.It should be noted that PPI analyses are known to have limited statistical power and are most suitable for factorial design fMRI tasks, which is why this analysis was performed as an exploratory analysis.Based on our PPI covariance matrices, a signal change of approximately 1.4-1.9% was needed to observe a significant effect on the interaction term.

Stroop task performance context
An adapted Stroop Color-Word Interference test has repeatedly been administered in the DFBC outside of the MRI scanner.Participants exposed to famine during early gestation performed worse on this task at age 58, having a longer response time and lower accuracy compared to unexposed participants (de Rooij et al., 2010).At ages 68 and 74, we no longer observed group differences in task performance, although exposed individuals did report experiencing more cognitive problems at age 72 (Wiegersma, Boots, Roseboom, & de Rooij, 2022, 2023).Potentially, the execution of the Stroop task within a protocol specifically designed to evoke a stress response at age 58 may have uncovered variations in task performance which are only visible under a high cognitive demand caused by having to perform the cognitive task under stressful circumstances (de Rooij et al., 2010).Alternatively, the absence of group differences may have been related to selective participation and mortality, which is discussed in the strengths and limitations section below (Wiegersma et al., 2023).
Although the current Stroop fMRI task was performed in the same cohort, it was adapted to be suitable for fMRI measurement.The Stroop fMRI task was specifically designed to achieve high accuracy scores to enable measuring a Stroop-task related fMRI response and to prevent measuring a stress-related response.Thereby, we specifically aimed to detect ROI activation and connectivity during the correct execution of the Stroop paradigm, potentially revealing exposure-related group differences related to a higher cognitive processing load and brain aging.As a result, we cannot evaluate any group differences in task accuracy and response time in the current study and we should be careful in comparing the results from the current study one-on-one with the previous Stroop task results obtained in our cohort outside of the MRI scanner.The current fMRI task design was more similar to the original Stroop task, which was used by De Groot et al. in a different cohort of individuals prenatally exposed to the Dutch famine.In their study, De Groot and colleagues did not observe differences in Stroop task performance at age 59 between people who were and were not exposed to famine, although results pointed towards a trend of worse cognitive performance after famine exposure in early gestation ( de Groot et al., 2011).Despite the challenges in comparing this fMRI study to the existing literature, the trends observed in the ROI analysis do indicate that the recruitment of brain regions essential in Stroop task-related cognitive processing in individuals exposed to famine in early gestation closely resembles that of a control group which has a higher average age.

Strengths and limitations
Strengths of this study include a solid task design resulting in an overall good task performance.Almost every participant was able to sufficiently perform the task with high accuracy scores, which was challenging considering the performance of a cognitively demanding task in an MRI setting in an elderly population.Additionally, a clear Stroop task brain activation pattern was measured matching with the Neurosynth Stroop task activation pattern confirming the successful measurement of Stroop task-related brain activity.Furthermore, we defined the ROIs based on a meta-analysis of activation patterns of Stroop task fMRI studies, thereby ensuring the functional relevance of our ROIs during task performance.
Several study limitations should be discussed.First, the number of study participants was limited by the availability of individuals exposed to undernutrition in early gestation in our cohort who were still alive and fit enough to participate.As a result, this study may have been underpowered to detect a true effect, although determining a specific threshold for clinical relevance for this specific exposure and outcome is challenging.Nevertheless, previous studies with a smaller sample size have observed differences in ROI activity between healthy older adults and individuals with MCI in the PPC, ACC and DLPFC, and our ROIbased approach limited multiple testing issues (Li et al., 2009;Poldrack, 2007;Rosano et al., 2005).Secondly, selection bias may have occurred.Selective participation should be considered as we previously observed an increased risk of dying at a younger age in women exposed to famine in early gestation (van Abeelen et al., 2012).Similarly, selective participation may be the result of other medical complaints previously related to prenatal famine exposure (Wiegersma et al., 2023).We have previously found that birth weight did not differ between cohort members in the original cohort and the 2019 subsample, and head circumference was smaller for exposed individuals who participated in 2019 compared to those who did not participate.Furthermore, exposed individuals in the current subsample had higher scores on the Stroop-like task at age 58 compared to those who did not participate in the current study (Wiegersma et al., 2023).This would have likely resulted in an underestimation of the differences between exposed and unexposed individuals (Bleker et al., 2021).Additional selective participation may have occurred due to the nature of the fMRI task.Individuals who were not able to correctly complete the trials at the beginning of the task because they had difficulties with understanding the task were not included, as we were not able to measure Stroop taskrelated brain activity.Although this number of individuals was small (n = 7, all unexposed), it did exclude the individuals with the poorest cognitive performance from this study.Lastly, as discussed above, comparability with previous Stroop-like task measurements in this cohort is challenging given the necessary adaptations made to the task design for compatibility with fMRI analysis.

Conclusions
Activation of the DLPFC, ACC and PPC during Stroop task performance in individuals exposed to the Dutch famine in early gestation was more similar to the older control group born before the Dutch famine compared to younger individuals conceived after the famine.These patterns potentially reflect a combined effect of aging and prenatal famine exposure on brain activity during Stroop task performance.However, it is important to note that none of these trends reached statistical significance.The current study did not reveal any patterns of decreased connectivity of the DLPFC during Stroop task performance after prenatal undernutrition.

Declaration of generative AI in scientific writing
During the preparation of this work the authors used ChatGPT in order to improve the readability and language of the introduction section.After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.A. Boots et al.

Fig. 2 .
Fig. 2. Violin plot of Stroop task ROI activation scores across all subjects according to prenatal exposure to famine.Individual observations are displayed as circles, with a box plot and median displayed per group.Key: ACC, anterior cingulate cortex; BOLD, blood oxygen level dependent; BB, born before; CA, conceived after; DLPFC, dorsolateral prefrontal cortex; EE, exposed in early gestation; PPC, posterior parietal cortex.

Fig. 3 .
Fig. 3. Violin plot of Stroop task PPI scores for regions with task-related connectivity according to prenatal exposure to famine.Individual observations are displayed as circles, with a box plot and median displayed per group.Key: BB, born before; CA, conceived after; DLPFC, dorsolateral prefrontal cortex; EE, exposed in early gestation; PPC, posterior parietal cortex.

Table 1
General, birth and adult characteristics according to prenatal exposure to famine.

Table 2
Stroop task accuracy and response time according to prenatal exposure to famine.
ANOVA analyses were used to create p-values for differences between study groups born before the famine, exposed to famine in early gestation and conceived after the famine.ANOVA was performed on rank-transformed accuracy scores.aControl condition.b Correct responses, excluding trials with RT < 200 ms.

Table 3
Median ROI activation scores in response to the Stroop task according to prenatal exposure to famine.

Table 4
Sex-specific median ROI activation scores according to prenatal exposure to famine.

Table 5
Median PPI scores in response to the Stroop task according to prenatal exposure to famine.
Differences between exposed and unexposed groups were tested for regions with task-related connectivity.P-values are based on Kruskal-Wallis tests.Key: ACC, anterior cingulate cortex; DLPFC, dorsolateral prefrontal cortex; IQR, interquartile range; PPC, posterior parietal cortex.