Introduction
Cognitive impairment is a common and often debilitating feature of bipolar disorder (BD) that persists across different mood states and confers substantial functional and psychosocial burden (Bora, Yucel, & Pantelis, Reference Bora, Yucel and Pantelis2009; Bortolato et al., Reference Bortolato, Miskowiak, Köhler, Vieta and Carvalho2015; Burdick et al., Reference Burdick, Russo, Frangou, Mahon, Braga, Shanahan and Malhotra2014; Cullen et al., Reference Cullen, Ward, Graham, Deary, Pell, Smith and Evans2016; Karantonis et al., Reference Karantonis, Rossell, Carruthers, Sumner, Hughes, Green and Van Rheenen2020; Simonsen et al., Reference Simonsen, Sundet, Vaskinn, Ueland, Romm, Hellvin, Melle, Friis and Andreassen2010). Most cognitive research on BD has analyzed case–control differences in measures of central tendency, which, as recently critiqued by Sánchez-Torres et al. (Reference Sánchez-Torres, García de Jalón, Gil-Berrozpe, Peralta and Cuesta2023), can mask individual fluctuations and variations in cognitive performance that are clinically relevant in psychiatry. Indeed, this within-person variation in cognitive performance, otherwise known as intraindividual variability (IIV) (MacDonald, Hultsch, & Dixon, Reference MacDonald, Hultsch and Dixon2003), maybe a feature of cognition in BD.
IIV can be measured across tasks and/or time but is most typically operationalized as trial-to-trial response consistency within a single reaction time task using the metrics of individual standard deviation (iSD) and/or the individual coefficient of variation (CoV) (Christensen et al., Reference Christensen, Dear, Anstey, Parslow, Sachdev and Jorm2005; Hultsch & MacDonald, Reference Hultsch and MacDonald2004). The former reflects the iSD of item-by-item response times. The latter reflects the ratio of the iSD of response times to the mean of those response times and is calculated by dividing the iSD by the individual mean. Increases in these IIV measures indicate irregular cognitive performance, potentially mediated by abnormalities in top-down executive control (MacDonald, Hultsch, & Dixon, Reference MacDonald, Hultsch and Dixon2003).
Originally explored in the context of aging, IIV was first considered to reflect psychometric noise, but is now recognized as a distinct cognitive construct that reliably indexes meaningful cognitive differences between individuals (MacDonald, Hultsch, & Dixon, Reference MacDonald, Hultsch and Dixon2003; Ram, Rabbitt, Stollery, & Nesselroade, Reference Ram, Rabbitt, Stollery and Nesselroade2005). Increased IIV has been associated with poorer socio-occupational functioning (Fuermaier et al., Reference Fuermaier, Tucha, Koerts, Aschenbrenner, Kaunzinger, Hauser and Tucha2015; Rajji, Miranda, & Mulsant, Reference Rajji, Miranda and Mulsant2014) and quality of life (Mitchell, Kemp, Benito-León, & Reuber, Reference Mitchell, Kemp, Benito-León and Reuber2010), as well as increased rates of mortality (Deary & Der, Reference Deary and Der2005). IIV increases have also been observed in healthy elderly people (Christensen et al., Reference Christensen, Dear, Anstey, Parslow, Sachdev and Jorm2005; Hultsch, MacDonald, & Dixon, Reference Hultsch, MacDonald and Dixon2002; MacDonald, Li, & Bäckman, Reference MacDonald, Li and Bäckman2009) and found to predict executive function, memory, processing speed, and sustained attention impairments within them several years later (Bielak et al., Reference Bielak, Hultsch, Strauss, MacDonald and Hunter2010a, Reference Bielak, Hultsch, Strauss, MacDonald and Hunter2010b; Cherbuin, Sachdev, & Anstey, Reference Cherbuin, Sachdev and Anstey2010; MacDonald, Hultsch, & Dixon, Reference MacDonald, Hultsch and Dixon2003). Increased IIV (abnormal to that expected by age) is also a marker of the Alzheimer’s disease prodrome (Roalf et al., Reference Roalf, Rupert, Mechanic-Hamilton, Brennan, Duda, Weintraub and Moberg2018), and has been found to index early cognitive changes in people with other neurodegenerative conditions who are otherwise not yet demonstrating cognitive impairments (Jones, Burroughs, Apodaca, & Bunch, Reference Jones, Burroughs, Apodaca and Bunch2020; Kälin et al., Reference Kälin, Pflüger, Gietl, Riese, Jäncke, Nitsch and Hock2014; Mazerolle, Wojtowicz, Omisade, & Fisk, Reference Mazerolle, Wojtowicz, Omisade and Fisk2013; Wojtowicz, Omisade, & Fisk, Reference Wojtowicz, Omisade and Fisk2013). This may relate to the demonstrated correlations of IIV with the brain’s white matter (MacDonald, Nyberg, & Bäckman, Reference MacDonald, Nyberg and Bäckman2006; Nilsson, Thomas, O’Brien, & Gallagher, Reference Nilsson, Thomas, O’Brien and Gallagher2014), changes in which have been found to precede the onset of observable cognitive impairments by several years (Silbert et al., Reference Silbert, Dodge, Perkins, Sherbakov, Lahna, Erten-Lyons and Kaye2012). Indeed, several studies have found that increases in IIV are associated with reduced white matter volume (Jackson, Balota, Duchek, & Head, Reference Jackson, Balota, Duchek and Head2012; Lövdén et al., Reference Lövdén, Schmiedek, Kennedy, Rodrigue, Lindenberger and Raz2013) and microstructural integrity in major frontal, parietal, and central white matter networks (Fjell, Westlye, Amlien, & Walhovd, Reference Fjell, Westlye, Amlien and Walhovd2011; Halliday, Gawryluk, Garcia-Barrera, & MacDonald, Reference Halliday, Gawryluk, Garcia-Barrera and MacDonald2019; Mella, de Ribaupierre, Eagleson, & de Ribaupierre, Reference Mella, de Ribaupierre, Eagleson and de Ribaupierre2013; Moy et al., Reference Moy, Millet, Haller, de Bilbao, Weber and Delaloye2011; Tamnes et al., Reference Tamnes, Fjell, Westlye, Østby and Walhovd2012).
It has been argued that IIV may index clinically and biologically meaningful information better than measures of central tendency alone (Davis, Sivaramakrishnan, Rolin, & Subramanian, Reference Davis, Sivaramakrishnan, Rolin and Subramanian2025; Dykiert, Der, Starr, & Deary, Reference Dykiert, Der, Starr and Deary2012; MacDonald, Hultsch, & Dixon, Reference MacDonald, Hultsch and Dixon2003; Sánchez-Torres et al., Reference Sánchez-Torres, García de Jalón, Gil-Berrozpe, Peralta and Cuesta2023; Tamnes et al., Reference Tamnes, Fjell, Westlye, Østby and Walhovd2012; Williams et al., Reference Williams, Hultsch, Strauss, Hunter and Tannock2005). Hence, examining IIV in BD could expand our insights into the cognitive profile of the disorder and help to elucidate the putative mechanisms contributing to the associated cognitive impairments, which, to date, remain unknown. Only a handful of preliminary studies have examined IIV in BD, finding it to be increased, on average, in middle-aged adults (Gallagher et al., Reference Gallagher, Nilsson, Finkelmeyer, Goshawk, Macritchie, Lloyd and Watson2015; Haatveit et al., Reference Haatveit, Westlye, Vaskinn, Flaaten, Mohn, Bjella and Ueland2023; Krukow et al., Reference Krukow, Szaniawska, Harciarek, Plechawska-Wojcik and Jonak2017) and youth samples (Brotman et al., Reference Brotman, Rooney, Skup, Pine and Leibenluft2009) compared to age-matched controls. One study showed that increased IIV in BD persisted longitudinally and was negatively associated with a global index of cognition (Depp et al., Reference Depp, Savla, de Dios, Mausbach and Palmer2012), while another study found that IIV increased even further as a function of the complexity of the cognitive task used (Moss et al., Reference Moss, Finkelmeyer, Robinson, Thompson, Watson, Ferrier and Gallagher2016). Only a few studies have examined whether IIV in BD is associated with clinical variables, presenting mixed findings regarding the role of mood symptoms, medication load, age of onset, or illness duration (Depp et al., Reference Depp, Savla, de Dios, Mausbach and Palmer2012; Gallagher et al., Reference Gallagher, Nilsson, Finkelmeyer, Goshawk, Macritchie, Lloyd and Watson2015; Haatveit et al., Reference Haatveit, Westlye, Vaskinn, Flaaten, Mohn, Bjella and Ueland2023; Krukow et al., Reference Krukow, Szaniawska, Harciarek, Plechawska-Wojcik and Jonak2017; Moss et al., Reference Moss, Finkelmeyer, Robinson, Thompson, Watson, Ferrier and Gallagher2016). Moreover, no studies have examined how IIV in BD relates to other cognitive domains or to indices of white matter, abnormalities of which are observed in BD and linked, to some extent, to its cognitive symptoms (Caruana et al., Reference Caruana, Carruthers, Berk, Rossell and Van Rheenen2024).
Considering the above, the characterization of IIV in BD remains in its infancy. The replication and expansion of existing preliminary studies focused on IIV is thus required to determine the extent to which IIV can inform our broader understanding of cognition within the disorder. In this study, we aimed to do this by further characterizing IIV and its correlates with BD. We specifically sought to replicate prior findings showing increased IIV in patients with BD compared to controls using a larger sample than most previous research. We also aimed to determine whether IIV (i) is particularly related to any specific cognitive domain, (ii) covaries with clinical symptoms, and (iii) is associated with whole-brain white matter integrity. We hypothesized that IIV would be increased in people with BD compared to controls, and that this increased IIV in BD would be associated with poorer cognitive performance across a range of domains as well as decreased whole-brain white matter integrity. The extent of associations between IIV and clinical variables remained an open question.
Methods
The study was approved by the local Human Ethics Review Committee and adhered to the Declaration of Helsinki.
Participant characterization
The data from 217 participants (n = 100 with BD and n = 117 healthy controls [HCs]) were included in this study. All participants had participated in studies led by the authors (e.g. see Karantonis et al., Reference Karantonis, Rossell, Carruthers, Sumner, Hughes, Green and Van Rheenen2020; Neill & Rossell, Reference Neill and Rossell2013; Tan & Rossell, Reference Tan and Rossell2014; Van Rheenen & Rossell, Reference Van Rheenen and Rossell2014b) and had been recruited using general advertisements as well as online websites and social media, with the BD participants also being recruited through community support groups. All participants had given prior informed consent for the analysis of their data.
Participants were aged between 18 and 65 years, were proficient in English, and had no known neurological disorders, acute medical illnesses, or significant hearing or visual impairments, no current alcohol or substance abuse/dependence, and none were pregnant. HCs also had no first-degree relatives with a psychiatric diagnosis. BD diagnosis and HC eligibility were confirmed using the Mini-International Neuropsychiatric Interview for BD (Sheehan et al., Reference Sheehan, Lecrubier, Sheehan, Amorim, Janavs, Weiller and Dunbar1998), with 83 BD participants meeting criteria for BD-I and 17 meeting criteria for BD-II. BD participants were all clinically stable outpatients at the time of assessment, and none were experiencing symptoms of psychosis. Current mood symptom severity was measured using the Montgomery and Åsberg Depression Rating Scale (MADRS; Montgomery & Åsberg, Reference Montgomery and Åsberg1979) and the Young Mania Rating Scale (YMRS; Young, Biggs, Ziegler, & Meyer, Reference Young, Biggs, Ziegler and Meyer1978). Sixty-six percent of participants were considered effectively stable with MADRS scores <12 and YMRS <8. The remaining 34% were symptomatic with mild–moderate symptoms (76% of these had MADRS scores >12, and a further 24% had YMRS scores >8). Self-reported use of mood stabilizers, antipsychotics, and antidepressants in the sample was also recorded, as was the age of illness onset, illness duration, psychiatric hospitalizations, and mood episode history (Table 1).
Table 1. Demographic, clinical, and cognitive characteristics of the full sample
Note: CoV, coefficient of variation; IIV, intraindividual variability; iSD, individual standard deviation; MADRS, Montgomery and Åsberg Depression Rating Scale; YMRS, Young Mania Rating Scale; + total years of education missing from five BD and seven HC participants; ^ premorbid IQ score missing from two BD participants; # number of mood episodes missing from 10 BD participants; p-values reported in the table reflect raw values, but are designated with a * if remaining significant after FDR correction; ~ Mann–Whitney U-test; ! higher trail-making test-B scores indicate worse performance.
Intraindividual variability
IIV measures were derived for all participants from individual responses on the Continuous Performance Test-Identical Pairs (CPT-IP) version (Cornblatt et al., Reference Cornblatt, Risch, Faris, Friedman and Erlenmeyer-Kimling1988). The CPT-IP was collected during the administration of the Matrics Consensus Cognitive Battery (MCCB), a battery of tests validated for use in BD (Burdick et al., Reference Burdick, Goldberg, Cornblatt, Keefe, Gopin, DeRosse and Malhotra2011; Van Rheenen & Rossell, Reference Van Rheenen and Rossell2014a). The CPT-IP is a computerized neurocognitive measure requiring participants to monitor a series of two-, three-, and then four-digit sequences and respond when identical sequences are presented consecutively. Across the two-, three-, and four-digit blocks, a total of 450 rapidly flashed digit sequences (150 per block) are delivered, including 30 ‘target’ digit pairs within each block, as well as 30 ‘catch’ trials that feature two successive similar but not identical digit sequences, and 90 random digit sequences that are in no way similar. Stimuli are flashed on the screen for 50 ms, followed by a 950-ms blank screen (stimulus onset asynchrony = 1,000 ms). Participants are asked to respond by quickly pressing and releasing the left mouse button whenever they are presented with an identical digit sequence pair. The total test time is ~10 min.
IIV was operationalized from the raw reaction time (milliseconds) values for each successful ‘hit’ across the entirety of the CPT-IP. (The size of the case–control difference in the mean standard deviation in response time for the CPT-IP was stronger when calculated across its entirety versus for each individual block. Hence, we decided to use the IIV measures calculated across the full task. See the Supplementary Methods and Supplementary Table S1 for further details.) Hit trials refer to any trial in which the second stimulus in a target pair received a response during the interstimulus interval (i.e. the response window). Thus, any response to the second stimulus in a target pair occurring after the onset of the next trial was not retained for the IIV calculation. Each participant’s mean reaction time was based on their hits, and the standard deviation of the mean of this reaction time (reflecting the participant’s iSD) was computed. Each individual’s CoV was also derived by dividing the iSD by their mean reaction time. These IIV measures are mathematically (and conceptually) distinct from accuracy, speed, or d-prime scores typically derived from the CPT-IP (Cho et al., Reference Cho, Pilloni, Tahsin, Best, Krupp, Oh and Charvet2023).
Other cognitive measures
Premorbid IQ was estimated in all participants using the Wechsler Test of Adult Reading (Wechsler, Reference Wechsler2001). Scores from other relevant and available cognitive tests were also analyzed. These included overall mean d-prime scores, which are the standard metrics of sustained attention from the CPT-IP in that they reflect a ratio of speed, accuracy, and focus when discriminating between target and distractor digit sequences across blocks. Scores from the Brief Assessment of Cognition in Schizophrenia-Symbol Coding were used as a measure of processing speed, and from the Letter Number Span as a measure of working memory. Time to complete scores from the Trail Making Test: Part B (Reitan, Reference Reitan1958) were used as a measure of executive function. All subtests are described in detail elsewhere (Kern et al., Reference Kern, Nuechterlein, Green, Baade, Fenton, Gold and Marder2008; Nuechterlein et al., Reference Nuechterlein, Green, Kern, Baade, Barch, Cohen and Gold2008; Yatham et al., Reference Yatham, Torres, Malhi, Frangou, Glahn, Bearden and Chengappa2010). Better performance was represented by higher scores on the working memory, processing speed, and sustained attention measures and lower scores on the executive measure.
Neuroimaging acquisition and processing
Diffusion-weighted imaging (DWI) scans were available in a subset of the BD sample (n = 55, of which n = 52 had BD I, n = 3 had BD II; 33 males and 28 females). Scans were acquired on a Siemens Magnetom 3T Tim Trio system (Erlangen, Germany) using a 34-channel head coil and a multi-shell protocol (Repetition Time = 9200 ms, Echo Time = 117 ms, voxel size = 2.5 × 2.5 × 2.5 mm3). Sixty diffusion gradient directions were acquired with a b-value of 3,000 s/mm2, 30 directions with a b-value of 2,000 s/mm2, and 30 directions with a b-value of 900 s/mm2. Ten non-diffusion-weighted images (b-value = 0 s/mm2) were also acquired, and to enable the estimation of susceptibility-induced off-resonance fields, five additional non-diffusion-weighted images were acquired with the same imaging parameters but with a reversed-phase encoding direction.
Images were processed and analyzed using FMRIB Software Library 6.0.1 (Smith et al., Reference Smith, Jenkinson, Woolrich, Beckmann, Behrens, Johansen-Berg and Matthews2004), adhering to the ENIGMA-DWI protocol. Susceptibility-induced distortions were estimated and corrected for using the TOPUP method (Andersson, Skare, & Ashburner, Reference Andersson, Skare and Ashburner2003). Subject motion and eddy current-induced distortion correction, as well as automated outlier replacement, were performed in line with the methods (EDDY) described in Andersson and Sotiropoulos (Reference Andersson and Sotiropoulos2016) and Andersson, Graham, Zsoldos, and Sotiropoulos (Reference Andersson, Graham, Zsoldos and Sotiropoulos2016). Fractional anisotropy (FA) maps were estimated for each participant using the DTIFIT option with the FMRIB Diffusion Toolbox by fitting a tensor model to the preprocessed diffusion data. Axial diffusivity (AD) (λ1) and radial diffusivity (RD) ([λ2 + λ3]/2) maps were also estimated using the eigenvalues associated with the fitted tensor model. Using tract-based spatial statistics, participant FA maps were then aligned to the custom ENIGMA-DWI FA template derived from 400 adult participants (Jahanshad et al., Reference Jahanshad, Kochunov, Sprooten, Mandl, Nichols, Almasy and Glahn2013) and subsequently projected onto the ENGIMA-DWI template skeleton. The same method was used to project images of FA’s constituent measures: mean diffusivity (MD), AD, and RD onto the skeleton. Voxels along the individual skeletons were averaged across 25 bilateral regions of interest (ROIs) based on the JHU WM atlas (Mori et al., Reference Mori, Oishi, Jiang, Jiang, Li, Akhter and Woods2008). Each of the diffusion measures was then imported into the Statistical Package for the Social Sciences (SPSS) and averaged over all ROIs to generate whole-brain FA, MD, AD, and RD values for each participant.
Statistical analyses
All data were statistically analyzed using SPSS version 27 (IBM). First, variables were visually checked for extreme outliers, and relevant statistical test assumptions were assessed and met using standard methods. (Outliers were considered at the sample level after iSD and CoV for each participant had been calculated, based on the SPSS categorizations of extreme outliers; that is, iSD and CoV scores that were less than/greater than three SDs of the mean were excluded.) In preliminary analyses, two-tailed comparisons (using χ 2-tests and independent-sample t-tests as appropriate) were conducted to compare the BD and HC groups on relevant demographic and clinical variables to characterize abnormalities in the BD sample and identify any potential covariates. The association of these potential covariates and the IIV indices (iSD and CoV) was also examined in the full sample using Pearson’s correlations, with only the variables that were significantly correlated with IIV being covaried in the analyses.
In the primary analyses, group differences in mean iSD, CoV, and other cognitive test scores were ascertained using one-way analyses of variance (ANOVAs). Two-tailed bivariate Pearson’s correlations (or nonparametric equivalent tests) were then conducted to explore associations between the IIV indices and the other cognitive scores in the BD and HC groups separately, with Fisher’s Z-tests used to compare correlations. Correlations were also conducted in the BD imaging subsample to examine the associations between whole-brain white matter (FA, RD, MD, and AD) and global IIV indices, as well as other cognitive test scores. Furthermore, associations between IIV and mood symptom severity scores were examined in the BD sample and in relation to other clinical factors (age of onset, illness duration, number of mood episodes, hospitalizations due to mood disturbance, and medication load). Group comparisons in mean IIV based on BD diagnostic subtype, psychosis history, and usage of key medication types (mood stabilizers, antipsychotics, and antidepressants) were also conducted using ANOVA (or nonparametric equivalent). To examine the influence of mood state and establish state versus trait effects, group comparisons of mean IIV were conducted, excluding the symptomatic BD participants (using the same procedures as used in the full sample) in secondary sensitivity analyses.
A false discovery rate (FDR) of p < 0.05 was applied to the results to account for multiple comparisons using the Benjamini–Hochberg method (see the Supplementary Material for details). The effect sizes in the text are given in Cohen’s d, while the reported p-values reflect raw, uncorrected values.
Results
Demographic, clinical, and cognitive characteristics of the sample
Demographic, clinical, and cognitive characteristics of the full sample are provided in Table 1 and of the DWI subsample in Supplementary Table S2. There were no differences in age, sex, years of education, or estimated premorbid IQ between the BD and HC groups, and none of these characteristics were significantly associated with iSD (age r = .098, p = .150; years of education r = −.040, p = .566; estimated premorbid IQ r = −.127, p = 0.063) or CoV (age r = .017, p = .799; years of education r = −.074, p = .290; estimated premorbid IQ r = −.057, p = .401). The mean MADRS and YMRS scores in the BD sample were low (M = 9.422, SD = 8.861 and M = 4.174, SD = 4.26, respectively). Fifty-eight percent of the BD group had a history of psychosis. Cognitively, the BD group performed significantly more poorly than HCs on measures of sustained attention and processing speed, with moderate effects (d = 0.448 and d = 0.692, respectively). The BD group also had worse mean executive function and working memory than HCs, with small effects (d = 0.401 and 0.316), although the former comparison was not significant initially and the latter did not survive the FDR correction. Moreover, BD participants had significantly higher mean IIV (indexed by both iSD and CoV) than HCs (Figure 1; iSD [F(1,215) = 15.724, p ≤ 0.001] and CoV [F(1,215) = 10.830, p = 0.001]).
Figure 1. Raincloud plots depicting mean comparisons of (a) global iSD and (b) global CoV between bipolar disorder (BD) and healthy control (HC) groups. p-Values reflect raw values, but are significant after FDR correction. CoV, ‘coefficient of variation’; iSD, ‘individual standard deviation’.
Associations between global IIV indices and other cognitive tests in BD and HC groups
In the BD group, a higher mean IIV, as measured by both iSD and CoV, was associated with worse performance in all four cognitive domains analyzed: sustained attention, working memory, executive function, and processing speed (Figure 2a). In the HC group, a higher mean IIV was only associated with worse sustained attention (Figure 2b). A secondary check of these associations using median splitting within each group based on high/low IIV indicated that these results persisted (Supplementary Table S3); however, no Fisher’s Z comparisons of the correlations between groups were significant.
Figure 2. Spearman’s rho correlations between IIV indices and the different cognitive domains for the (a) bipolar disorder (BD) and (b) healthy control (HC) groups.
Note: CoV, ‘coefficient of variation’; iSD, ‘individual standard deviation’; * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001 (FDR-corrected).
Associations between global IIV indices and clinical variables in the full BD sample
Mean iSD was higher in BD participants using antidepressants (on antidepressants: M = 129.949, SD = 32.573; off antidepressants: M = 115.958, SD = 28.854; F(1,86) = 4.277, p = 0.042, d = 0.455). Mean CoV was higher in BD-I than BD-II participants (BD-I: M = 0.225, SD = 0.047; BD-II: M = 0.202, SD = 0.028; Kruskal–Wallis test p = 0.025, d = 0.595) and in BD participants using versus not using antipsychotics (on antipsychotics: M = 0.233, SD = 0.047; off antipsychotics: M = 0.213, SD = 0.045; F(1,84) = 3.956, p = 0.050, d = 0.435). Mean CoV was also negatively correlated with age of BD onset (r = −0.258, p = 0.016). However, none of these results survived FDR correction. No other significant associations were observed. See Supplementary Tables S4 and S5 for details.
Sensitivity analyses including only affectively stable BD participants (n = 66)
There were no substantial differences between the outcomes of the sensitivity analyses that excluded symptomatic BD participants and the analyses conducted using the full BD sample (Supplementary Tables S6 and S7).
Associations between IIV indices and global white matter microstructure in the BD imaging subset (n = 55)
Figure 3 shows a significant negative association between whole-brain FA and mean iSD (r = −.275, p = .042). Significant positive associations were also evident between whole-brain RD and mean iSD (r = .271, p = .045) and mean CoV (r = .296, p = .028). No associations between the IIV indices and whole-brain MD or AD were evident, nor were there associations between any of the DWI measures and sustained attention, executive function, working memory, or processing speed.
Figure 3. Pearson’s r correlations of global IIV indices with diffusion-weighted imaging measures and the different cognitive domain scores in the BD neuroimaging subsample.
Note: AD, ‘axial diffusivity’; CoV, ‘coefficient of variation’; FA, ‘fractional anisotropy’; iSD, ‘individual standard deviation’; MD, ‘mean diffusivity’; RD, ‘radial diffusivity’; *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 (FDR-corrected).
Discussion
This study expanded the characterization of IIV in BD, identifying mean increases in IIV in BD as compared to HCs of an equivalent mean age and sex. This increased IIV in BD aligns with our hypothesis and replicates and reinforces the results of prior, albeit smaller, IIV BD studies (Brotman et al., Reference Brotman, Rooney, Skup, Pine and Leibenluft2009; Depp et al., Reference Depp, Savla, de Dios, Mausbach and Palmer2012; Gallagher et al., Reference Gallagher, Nilsson, Finkelmeyer, Goshawk, Macritchie, Lloyd and Watson2015; Haatveit et al., Reference Haatveit, Westlye, Vaskinn, Flaaten, Mohn, Bjella and Ueland2023; Krukow et al., Reference Krukow, Szaniawska, Harciarek, Plechawska-Wojcik and Jonak2017; Moss et al., Reference Moss, Finkelmeyer, Robinson, Thompson, Watson, Ferrier and Gallagher2016). That the average IIV of the BD sample remained increased in sensitivity analyses that removed all symptomatic patients, supports suggestions that IIV abnormalities in BD are trait-like in nature (Depp et al., Reference Depp, Savla, de Dios, Mausbach and Palmer2012). Moreover, these increases were not linked to clinical variables, such as a longer illness duration, a history of more mood episodes, or psychiatric hospitalizations.
IIV increases in BD were found in individuals with worse processing speed and working memory performance, as well as in those with poorer sustained attention and executive function. The largest effects were for the latter two cognitive domains, consistent with evidence that IIV closely covaries with, and may even proxy, top-down attentional and executive control (Cañigueral et al., Reference Cañigueral, Ganesan, Smid, Thompson, Dosenbach and Steinbeis2023; MacDonald, Li, & Bäckman, Reference MacDonald, Li and Bäckman2009). In the HC group, IIV was significantly associated with only sustained attention, which is sensical, given that sustained attention scores were derived from the same cognitive task as the IIV measures. Thus, a correlation between them is expected even though these measures are conceptually and mathematically distinct. The significant correlations between IIV and the other cognitive domain scores were observed solely in the BD group, which may reflect that increased IIV in this group is one index of a more generalized cognitive impairment. This is consistent with theories that domain-level cognitive abilities are more related to each other (less differentiated) at lower levels of general cognitive ability than they are at higher levels (Tucker-Drob, Reference Tucker-Drob2009). This inference should be interpreted with the caveat that the differences in correlations between the BD and HC groups were nonsignificant.
In our study, increased IIV in BD was associated with two measures of white matter integrity: reduced whole-brain FA and increased whole-brain RD. Since concurrent decreases in FA and increases in RD may reflect damaged white matter resulting from reduced myelin integrity/demyelination (Johnson, Diaz, & Madden, Reference Johnson, Diaz and Madden2015; Madden, Bennett, & Song, Reference Madden, Bennett and Song2009), this pattern of findings aligns with previous work positing that one neurobiological mechanism underpinning increases in IIV is reduced action potential conduction efficiency caused by axonal or myelin abnormalities (Fjell, Westlye, Amlien, & Walhovd, Reference Fjell, Westlye, Amlien and Walhovd2011; Moy et al., Reference Moy, Millet, Haller, de Bilbao, Weber and Delaloye2011). It is notable that in our study, FA and RD were only correlated with the IIV metrics of interest and not with any other cognitive domain scores. Thus, IIV appears to provide unique information about the brain-behavior relationship in BD, beyond that of the more commonly used cognitive scores, which have not been robustly linked to white matter microstructure in the disorder to date (Caruana et al., Reference Caruana, Carruthers, Berk, Rossell and Van Rheenen2024).
Taken together, our data demonstrate that increases of IIV in BD that are associated with reductions in performance in other cognitive domains, as well as in white matter microstructural integrity – patterns typically observed in aging samples (Nilsson, Thomas, O’Brien, & Gallagher, Reference Nilsson, Thomas, O’Brien and Gallagher2014). In our data, increases in IIV in BD were evident, despite the BD and HC groups being equated in terms of mean age, and age not being significantly associated with IIV in the BD group or the overall sample. Given that, increased IIV may be considered a marker of advancing age, and considering that this sample largely comprised middle-aged adults within a period of the lifespan generally characterized by cognitive consistency and stability (Ferreira et al., Reference Ferreira, Machado, Molina, Nieto, Correia, Westman and Barroso2017), we speculate that the increased IIV in BD observed here may reflect the outcome of premature or accelerated cognitive aging. An alternative explanation is that elevated IIV in BD is related to a lag in normative cognitive development, given that IIV is known to follow a U-shape curve across the lifespan in which it is initially high during childhood, plateaus during adulthood, and trends upward in the elderly (MacDonald, Li, & Bäckman, Reference MacDonald, Li and Bäckman2009). However, since cognition is not typically impaired in BD during the premorbid period (which typically coincides with childhood and adolescence (Van Rheenen et al., Reference Van Rheenen, Lewandowski, Bauer, Kapczinski, Miskowiak, Burdick and Balanzá-Martínez2020), this explanation seems less likely.
Some limitations of this study should be considered. First, IIV metrics were calculated from individual responses to target stimuli that occurred within a 949-ms response window. This would have disadvantaged particularly slow participants whose responses outside this period would not have been captured. Second, the use of cross-sectional data precluded our ability to test the directionality of relationships between IIV and other variables of interest. Third, the imaging subsample was modest in size and comprised only those with BD, limiting our ability to conduct white matter tract-specific analyses or group comparisons with HCs. Fourth, the effects in this study were small to moderate in size, which should be taken into consideration when interpreting the results. That said, the absence of large effects suggests that other factors of relevance to IIV may be relevant to future research on this topic, such as peripheral inflammation, stress, and trauma, which are implicated in BD and known correlates of cognitive performance and white matter pathology (Li, Xu, & Wang, Reference Li, Xu and Wang2023).
Finally, it should be mentioned that our use of the CPT-IP to examine IIV was based on the availability of this test within a widely used cognitive battery, the MCCB. iSD and CoV are easily calculated from the CPT-IP and were thus used to operationalize IIV here. However, these IIV indices assume that the response times across the CPT-IP are Gaussian (i.e. normally distributed) (Moss et al., Reference Moss, Finkelmeyer, Robinson, Thompson, Watson, Ferrier and Gallagher2016). We initially reasoned that the use of such metrics was most suitable because ex-Gaussian analyses require ~100 trials as a rule of thumb, and the absolute number of trials from which valid responses can be recorded from the CPT-IP for each digit-sequence block is limited to 30. However, subsequent preliminary analyses in our data suggested that IIV is best measured across the totality of the CPT-IP, since group differences in IIV were found not to be affected by the increasing cognitive load of each digit block, and the largest effects were evident when all valid responses across the task were used to calculate the IIV metrics (see Supplementary Materials for details). This suggests that the CPT-IP may be suitable for ex-Gaussian analyses, as there are 90 possible hit trials across all blocks. Thus, future extensions of our work could benefit from examining ex-Gaussian parameters, such as mu, sigma, and tau, in addition to iSD and CoV.
In summary, this study complements an existing, albeit small, evidence base showing that IIV is increased in BD. It extends it by demonstrating that IIV elevations can be elicited from a widely used cognitive test from the MCCB using easily calculated metrics that are detrimentally associated with cognitive performance across other domains, as well as with a proxy of underlying myelin damage in the neural white matter. Given the unique links between IIV and white matter, but not between white matter and more traditionally used cognitive scores, IIV may be considered a neurobiologically informative cognitive measure for BD that is worthy of future research.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291725100731.
Acknowledgments
The authors would like to acknowledge the facilities and scientific assistance of the National Imaging Facility, a National Collaborative Research Infrastructure Strategy (NCRIS) capability, at the Swinburne Neuroimaging Facility, Swinburne University of Technology. The authors would also like to sincerely thank the participants in this study for their time and contributions. The authors note that for the purposes of open access, a CC BY public copyright licence has been applied to any Author Accepted Manuscript version arising from this submission.
Funding statement
In completing this work, TVR was supported by the University of Melbourne and an Al and Val Rosenstrauss Fellowship from the Rebecca L Cooper Medical Research Foundation. SLR was supported by National Health and Medical Research Council (NHMRC) Senior Fellowship (GNT1154651) and project grant (GNT1060664). GFC was supported by the University of Melbourne/Fay Marles Bursary. JK was supported by a Swinburne University/postgraduate scholarship and LF was supported by an Australian Rotary Health/Ian Parker Bipolar Research Fund/Brunslea Park Estate postgraduate scholarship. Data collection was supported by the Jack Brockhoff Foundation, University of Melbourne, Barbara Dicker Brain Sciences Foundation, Rebecca L Cooper Foundation, the Society of Mental Health Research, Helen McPherson Smith Trust and Australian Rotary Health.
Competing interests
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest.
Open access
