Subjective cancer-related cognitive impairments and salience network connectivity in breast cancer survivors

Abstract

Purpose

Little is known about the neural basis of subjective cancer-related cognitive changes. The purpose of this study was to explore salience network connectivity in relation to subjective executive and memory dysfunction in breast cancer survivors compared to controls.

Methods

A retrospective cross-sectional analysis of neuroimaging, subjective cognitive, clinical, and demographic data in chemotherapy-treated primary breast cancer survivors compared to frequency matched controls was used. Functional connectivity within salience network hubs (anterior cingulate, bilateral insula) was determined using resting state functional MRI. Mann–Whitney U tests were used to evaluate group differences and Spearman’s rho correlations were examined among the behavioral measures and salience network connectivity.

Results

We included 65 breast cancer survivors and 71 controls. Survivors demonstrated greater subjective executive dysfunction and memory complaints (p < .001) and lower salience network connectivity (p < .05) than controls. Executive functioning correlated with bilateral insula and left anterior cingulate connectivity (rho > − 0.29, p < .05). Distress did not correlate with salience network connectivity.

Conclusion

These findings suggest that salience network connectivity may represent a biomarker of subjective cancer-related cognitive changes.

Implications for Cancer Survivors

Subjective cancer-related cognitive changes are common following treatment and associated with objective changes in brain connectivity

Introduction

Cancer-related cognitive impairments (CRCI) are among the most distressing and burdensome effects of cancer and its treatment that can negatively impact survivors’ daily lives [1-4]. Both objective and subjective CRCI have been observed in patients with cancer, though these rarely correlate [5]. Objective and subjective measures of cognitive impairment likely represent different aspects of CRCI [5, 6], and thus may have different biologic mechanisms. Research suggests that objective impairments in breast cancer survivors are associated with altered neural connectivity [7-17], but less is known about the neural basis of subjective cancer-related cognitive changes.

In breast cancer survivors, correlations between executive function complaints and reduced gray matter in the left middle frontal gyrus [18] and lower activation in the left caudal lateral middle frontal gyrus during a card sorting task [19] have been found in previous studies. Higher spatial variance in the executive control network during a working memory task has also been found to predict greater attention complaints [20]. Higher connectivity between the precuneus and hippocampus was previously correlated with general cognitive concerns [21] in breast cancer survivors. Our group and others have demonstrated that altered default mode network connectivity is correlated with subjective executive and memory dysfunction in breast cancer survivors [22-24]. Very few studies have examined cancers other than breast but one study involving patients with gynecological cancer found that global network structural characteristics (i.e., small-worldness and path length) correlated with subjective cognitive impairments [25].

However, no studies to date have examined the relationship between subjective cognitive impairment and the salience network in cancer survivors. This large-scale functional network is implicated in switching between cognitive resources in addition to processing emotional experiences [26]. The anterior insula and anterior cingulate cortex are hubs of the salience network [27], meaning that these regions conduct most of the information processing for this network. These regions integrate external experiences with internal feelings and/or emotions and are suggested to guide one’s actions or decisions in response to the external/internal interaction [28]. Subjective cognitive dysfunction reflects the individual’s perceptions of cognitively interacting with the environment. Therefore, we hypothesized that altered salience network connectivity would be associated with subjective executive and memory dysfunction in breast cancer survivors.

Materials and methods

Participants

We examined retrospective neuroimaging, subjective cognitive, clinical, and demographic data from 80 chemotherapy-treated primary breast cancer survivors who had completed all primary treatments (surgery, radiation, chemotherapy), excluding hormone blockade, at least 6 months before study enrollment (no maximum number of months). Data were also examined for 82 female controls, matched for age and education and with no prior history of cancer or cancer treatment, for comparison. Survivors and controls participated in prior studies from our laboratory focused on neuroimaging correlates of objective cognitive dysfunction. Breast cancer survivors all received chemotherapy as part of their treatment and were free from disease and had no history of relapse or recurrence at the time of evaluation. Participants were excluded for neurologic, psychiatric, or medical conditions known to affect cognitive function. Informed consent was obtained from all individual participants included in the study. This study was approved by the Stanford University Institutional Review Board.

Measures

Demographic and treatment-related patient information (e.g., estrogen receptor status, tumor burden, treatment regimen) were collected via self-report.

Subjective cognitive measures

The Behavioral Rating Inventory of Executive Function (BRIEF), a questionnaire of everyday executive function behaviors [29], and the Multifactorial Memory Questionnaire Ability Scale (MQ), a 21-item questionnaire concerning memory abilities [30], were administered to measure subjective CRCI. The BRIEF includes the General Executive Composite score (GEC), Behavioral Regulation Index (BRI), and Metacognition Index (MI) [29]. Higher scores indicate worse executive function and worse memory abilities. Participants also completed the Clinical Assessment of Depression (CAD), a 50-item questionnaire measuring depression, anxiety, and cognitive fatigue [31]. Scores were converted to T scores, which have a mean of 50 and standard deviation of 10, based on the test’s published normative data. To focus the sample with respect to our hypothesis, we excluded breast cancer survivors whose GEC scores were less than 1 standard deviation below the control mean (i.e., low impairment) and excluded controls whose GEC scores were greater than one standard deviation from the control mean (i.e., high impairment), resulting in a final sample of 65 breast cancer survivors and 71 controls (Table 1).

Table 1.

Demographic and clinical sample description

	Breast cancer survivors (n = 65)	Controls (n = 71)	Comparison test statistic
Age in years	50.7 (7.4)	47.9 (13.1)	2503.5
Years of education	16.6 (2.5)	17.2 (2.5)	1982
Minority	n = 12 (18.5%)	13 (18.3%)	0.0005
Post-menopausal	n = 42 (64.6%)	n = 50 (70.4%)	0.52
Time since chemotherapy treatment (in months)	39.2 (32.3)	-	-
History radiation treatment	n = 50 (76.9%)		-
History endocrine therapy	n = 45 (69.2%)		-

Mann–Whitney U tests used for continuous variables and χ² used for categorical

Neuroimaging acquisition and preprocessing

Functional magnetic resonance imaging (fMRI) data were obtained, while participants rested with eyes closed using a T2*-weighted [32] gradient echo spiral pulse sequence: TR = 2000 ms, TE = 30 ms, flip angle = 80° and 1 interleave, FOV = 22 cm, matrix = 64 × 64, in-plane resolution = 3.4375, number of volumes = 216 with a 3 T GE Signa HDx whole body scanner (GE Medical Systems, Milwaukee, WI). A high-order shimming method was employed to reduce field heterogeneity. A high-resolution, 3D IR prepared FSPGR scan was also acquired and used for spatial normalization of fMRI: TR: 8.5, TE: minimum, flip: 15 degrees, TI: 400 ms, BW: ± 31.25 kHz, FOV: 22 cm, phase FOV: 0.75, slice thickness: 1.5 mm, 124 slices, 256 × 256 @ 1 NEX, and scan time: 4:33.

Functional connectivity preprocessing was performed with Statistical Parametric Mapping v12 and CONN v21 Toolboxes [33, 34] implemented in MATLAB v2021b (MathWorks, Inc., Natick, MA). Briefly, this involved realignment, coregistration with the segmented anatomic volume, spatial normalization, artifact detection, and smoothing (FWHM = 8 mm) followed by band-pass filtering (0.008–0.09 Hz) and noise correction [35]. Motion parameters from realignment were included as regressors and images identified as motion or signal outliers were excluded. No volumes had greater than 10% outliers. Pairwise temporal correlations were computed for 90 regions from the Automated Anatomical Atlas [36] and normalized to z-scores using Fisher’s r-to-z transformation. Bilateral insula and anterior cingulate were used to define the salience network [27]. The mean z-score for each of these four regions was calculated, representing each region’s connectivity with the rest of the brain.

Statistical analyses

Descriptive statistics were explored to characterize and describe the sample. Parametric assumptions were not met for the main variables; therefore, Mann–Whitney U tests were used to evaluate group differences in GEC, MQ, CAD, and salience network connectivity. Spearman’s rho correlations were examined among the behavioral measures and salience network connectivity separately in the survivors and controls.

Results

Subjective cognitive function

GEC scores (W = 3309.5, p < 0.001), BRI scores (W = 3513, p < 0.001), MI scores (W = 3751, p < 0.001), and MQ scores (W = 2874.5, p < 0.001) were higher in survivors compared to controls indicating greater subjective cognitive dysfunction. CAD was higher in survivors than controls but not significantly (W = 2703.5, p = 0.056, Table 2).

Table 2.

Comparison of subjective cognitive function and salience network connectivity for breast cancer survivors and controls

	Breast cancer survivors (n = 65) Mean (SD)	Controls (n = 71) Mean (SD)	Mann–Whitney U test statistic
GEC	53.68 (13.27)	43.96 (6.18)	3309.5***
BRI	52.66 (9.99)	44.30 (8.0)	3513.0***
MI	56.83 (12.60)	44.68 (6.67)	3751.0***
MQ^	44.38 (10.70)	32.21 (7.14)	2874.5***
CAD	45.28 (11.33)	42.64 (8.6)	2703.5
RINS	0.28 (0.11)	0.33 (0.08)	1727.0*
LINS	0.31 (0.13)	0.37 (0.10)	1683.0**
LACC	0.15 (0.08)	0.19 (0.07)	1657.0**
RACC	0.33 (0.14)	0.39 (0.11)	1644.0**

BRI, Behavioral Regulation Index; CAD, Clinical Assessment Of Depression; GEC, general executive function; LACC, left anterior cingulate; LINS, left insula; MI, Metacognition Index; MQ, multifactorial memory questionnaire ability scale; RACC, right anterior cingulate; RINS, right insula.

^{^}n = 60 for breast cancer survivors, n = 57 for controls.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Salience network connectivity

Bilateral insula and anterior cingulate connectivity were all significantly lower in survivors compared to controls (p < 0.05, Table 2).

Correlations between salience network connectivity and subjective cognitive function

In the survivors, GEC correlated with bilateral insula and left anterior cingulate connectivity (rho > −0.29, p < 0.05). MQ was not correlated with the GEC, CAD, or salience network connectivity. CAD was correlated with GEC, MI, and BRI (rho > − 0.49, p < 0.001), but did not correlate with salience network connectivity (Table 3).

Table 3.

Spearman’s rho correlations among subjective cognitive function and salience network connectivity in breast cancer survivors (n = 65)

	CAD	GEC	BRI	MI	MQ	RINS	LINS	LACC	RACC
CAD	1
GEC	− 0.49***	1
BRI	− 0.50***	0.58***	1
MI	− 0.49***	0.62***	0.62***	1
MQ	0.17	0.02	− 0.17	0.10	1
RINS	− 0.00	− 0.31*	− 0.09	− 0.17	0.05	1
LINS	− 0.03	− 0.29*	− 0.06	− 0.11	0.01	0.94***	1
LACC	0.03	− 0.33**	− 0.07	− 0.17	0.13	0.72***	0.62***	1
RACC	0.06	− 0.21	− 0.15	− 0.12	0.10	0.81***	0.81***	0.54***	1

BRI, Behavioral Regulation Index; CAD, Clinical Assessment Of Depression; GEC, general executive function; LACC, left anterior cingulate; LINS, left insula; MI, Metacognition Index; MQ, Multifactorial Memory Questionnaire Ability Scale; RACC, right anterior cingulate; RINS, right insula.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

In controls, CAD was correlated with GEC, MI, and BRI (rho > 0.32, p < 0.01) and GEC, MI, and BRI correlated with MQ (rho > 0.39, p < 0.01) but there were no correlations with salience network connectivity (Table 4).

Table 4.

Spearman’s rho correlations among subjective cognitive function and salience network connectivity in controls (n = 71)

	CAD	GEC	BRI	MI	MQ	RINS	LINS	LACC	RACC
CAD	1
GEC	0.49***	1
BRI	0.32**	0.69***	1
MI	0.33**	0.71***	0.75***	1
MQ	0.22	0.39**	0.42**	0.42**	1
RINS	− 0.05	− 0.01	0.11	− 0.10	− 0.01	1
LINS	0.07	0.07	0.19	0.00	− 0.05	0.84***	1
LACC	− 0.03	− 0.05	0.15	0.08	0.10	0.59***	0.49***	1
RACC	0.04	0.09	0.17	− 0.00	− 0.01	0.78***	0.76***	0.41***	1

BRI, Behavioral Regulation Index; CAD, Clinical Assessment Of Depression; GEC, general executive function; LACC, left anterior cingulate; LINS, left insula; MI, Metacognition Index; MQ, Multifactorial Memory Questionnaire Ability Scale; RACC, right anterior cingulate; RINS, right insula.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Discussion

We examined the neural connectivity associated with subjective cognitive functioning in breast cancer survivors. We hypothesized that salience network connectivity would be different in survivors compared to controls and associated with executive and memory complaints. We found greater subjective executive dysfunction, worse memory abilities, and lower salience network connectivity in the survivors compared to controls. We also found correlations between self-reported executive dysfunction and connectivity in the anterior cingulate and insula in survivor group. This supports previous findings of correlations between increased salience network connectivity and improved subjective cognitive function in response to a mindfulness intervention in breast cancer survivors [37]. Together, these findings suggest that salience network connectivity may represent a biomarker of subjective CRCI.

Executive function encompasses many self-regulating processes that manage cognitive functions, emotions, and behaviors [38]. Executive functioning related to goal and future-oriented behaviors has largely been designated to the prefrontal parietal cortices and the brain networks located within these regions [39, 40]. Accordingly, hypoactivation in these regions have been consistently identified in cancer survivor compared to controls [41], and correlations between subjective executive function and regions of the prefrontal cortex have been previously reported in breast cancer survivors [18, 19]. Executive functions are also associated with cingulate and insular brain regions, which interact with frontal and parietal regions through large, distributed brain networks [42]. Current evidence supports that at least six networks play integral roles in executive functioning, including the salience and cingulo-opercular networks, which both contain the ACC and anterior insula [40]. While these brain networks may be functionally distinct, they likely interact in flexible and dynamic ways, resulting in the array of executive functions that occur in the context of everyday life [40].

Emotional regulation falls under the umbrella of executive functioning and refers to the ability to monitor, evaluate, and control one’s internal and external emotional reactions, which can be automatic or effortful depending on context [43]. Emotional regulation mediates the impact of negative emotions on quality of life in breast cancer survivors [44]. The salience network, including the insula and anterior cingulate, is implicated in the complex processing and responding to emotional experiences [26, 28]. Several items on the BRIEF instrument interrogate this aspect of executive functioning (i.e., emotional control/regulation) which likely explains some of the correlations found in this study.

Self-monitoring, more broadly, is another component of executive functioning that can be negatively affected by breast cancer treatment [45]. Metacognition is a term that refers to self-monitoring of cognitive processes, or “thinking about thinking” [46, 47], that is relevant to subjective CRCI. While the neural architecture supporting metacognitive processes is still under investigation, the bilateral insula, together with the prefrontal cortex, has been identified as hubs of interoception, which is a critical component of metacognition [47]. The BRIEF instrument includes a metacognitive index that probes this aspect of executive functioning. It is possible that the nuances and dynamic neural processing of perceiving one’s cognitive dysfunction and/or self-regulation are coordinated in part by the salience network.

We found a pattern of lower connectivity in the hubs of the salience network in the breast cancer cohort compared to controls, which is consistent with a previous study [48]. Perrier and colleagues also found that hypoconnectivity in the salience network correlated with positive self-referential behaviors (i.e., processing of information relevant to oneself) [48]. Hypoconnectivity of the salience network may result in impaired coordination of external and internal stimuli and subsequent self-referential processing [49]. It is possible that hypoconnectivity in the salience network in our sample of survivors represents the self-referential process of evaluating executive functioning, but we did not measure self-referential behaviors so cannot interrogate this hypothesis.

Memory abilities did not correlate with the other self-report measures or salience network connectivity. It is likely that memory abilities correlate with connectivity in other brain regions not measured in this study. For instance, scores from this subjective memory test correlated with prefrontal and default mode network connectivity in our previous study [50]. In older, non-cancer adults with subjective cognitive decline, perceived memory complaints were associated with decreased connectivity between the posterior cingulate cortex and left parahippocampal gyrus, after controlling for depression and overall cognitive functioning [51]. Salience network hypoconnectivity may be a specific biomarker of elevated subjective executive dysfunction, not memory dysfunction. Alternatively, other measures of subjective memory problems or other cognitive domains may yield different results.

Subjective CRCI has often been explained as psychological distress given the strong correlation between the two [5, 6, 52]. We also noted a significant correlation between GEC and CAD scores in both survivors and controls and CAD was higher in the survivors, though not significantly (statistically or clinically). CAD scores did not correlate with salience network connectivity in either group, explaining less than 0.4% of the variance. These findings indicate that subjective CRCI and psychological distress have different neural mechanisms, and provide further evidence that subjective CRCI cannot be entirely explained by distress.

This study was limited by the retrospective, cross-sectional design. Future research should prospectively assess salience network connectivity changes in relation to subjective CRCI. We administered two valid and reliable measures for subjective CRCI; however, these measures covered only two cognitive domains. Similarly, we focused on the hub regions of the salience network as these are the most important for information processing and to limit the number of comparisons. Future studies with larger samples should include more regions of the salience network and compare other functional networks in relation to subjective CRCI. Our sample included survivors after primary treatment. Self-reported CRCI is more prevalent, and sometimes more severe, during active treatment; thus, this study should be replicated in patients earlier in the cancer trajectory [5, 53]. Our sample was highly educated and < 20% identified as an ethnic or racial minority, limiting the generalizability of the study findings. Future studies should include a more diverse sample in terms of sociodemographic characteristics.

Despite these limitations, our findings provide further support that subjective CRCI is associated with altered neural connectivity. Study findings could be used to inform future prospective studies investigating neural biomarkers of CRCI and/or to inform targeted intervention development to improve, or prevent, subjective CRCI in cancer populations.

Abbreviations

CAD

Clinical Assessment of Depression

CRCI

Cancer-related cognitive impairments

fMRI

Functional magnetic resonance imaging

GEC

Behavioral Rating Inventory of Executive Function General Executive Composite score

Data availability

The original MRI data underlying this article cannot be shared publicly due to data protection regulation, but connectome matrices are available upon request to the corresponding author (srkesler@austin.utexas.edu). All preprocessing and analysis codes are available at https://github.com/srkesler.