According to the International Cognition and Cancer Task Force (ICCTF), cancer-related cognitive impairment (CRCI) is characterized as changes to mainly the cognitive domains of memory, processing speed, and executive function [1]. In the literature, survivors of cancer have reported experiencing CRCI, even among patients diagnosed with non-central nervous system (CNS) cancer. The prevalence of CRCI in non-CNS cancer survivors can be wide, ranging from 30% before treatment to as high as 85% in patients on active treatment [2-6]. Although the effects of CRCI may be transient or long term, cognitive decline can adversely affect patients’ quality of life [7]. Clinicians have continued to identify CRCI as an ongoing unmet need in survivorship care [8].

It must be acknowledged that the cause of CRCI is likely to be multifactorial. There are several proposed mechanisms of CRCI [9]. These include but are not limited to oxidative stress, dysregulation of inflammatory cytokines, genetic susceptibility, hormone deficiencies, and/or psychological distress. Chemotherapeutic drugs, which may not be able to cross the blood brain barrier (BBB), have been postulated to induce inflammatory effects that may alter the structural integrity of brain. Therein lies the potential value for biological factors that may correlate with changes in brain activity, structure, and volume, which are factors associated with cognitive impairment (Fig. 1).

Figure 1. Treatment- and patient-related factors underlying CRCI, adapted from Mayo et al. [9].

Over the past 15 years, our research team based in Singapore and the United States has conducted numerous human studies that were dedicated to evaluating the underlying mechanisms of CRCI. In a number of these studies, potential biomarkers that may explain the pathophysiology of CRCI were discovered and validated. In this review, we describe the evidence that has been generated in our collection of studies (Table 1).

Table 1 . Biomarkers associated with cancer-related cognitive impairment.

Biomarker	Mechanism	Findings in observational studies and challenges
Brain-derived neurotrophic factors (BDNF)	Impact on neuronal survival, proliferation, and plasticity	• Reduction over time during treatment is associated with CRCI, suggesting augmentation of BDNF may be a therapeutic angle. • Findings not consistent across studies.
Dehydroepiandrosterone (DHEA), its sulfated form (DHEAS) and estradiol	Hormone suppression can lead to cognitive process interruption and overexpression has shown to improve memory outcomes	• Pre-chemotherapy and during treatment levels may potentially explain CRCI in breast cancer patients. • Findings are limited to hormone expressing breast cancer patients.
Extracellular vesicles/exosomes	Impact on synaptic transduction function and other neurological processes	• Dynamin-1, zonula occludens-2 (ZO-2), junctional adhesion molecule C (JAM-C), and claudin hold great potential. • Limited data available; needs validation.
Inflammatory cytokines	Known to mediate the neuronal and glial cell functioning and affect neuronal repair	• Pro-inflammatory cytokines (such as IL-1b, IL-6, TNF-α) are induced during treatment, and are associated with CRCI. • Although cytokines are consistently observed to be associated with CRCI among studies, specific cytokine targets are not consistent.
Mitochondrial DNA	Involved in energy production. Mitochondrial dysfunction and resultant alterations can affect energy metabolism	• Lack of association with CRCI observed.
Vascular endothelial growth factor (VEGF)	Modulation of neuronal plasticity	• Lack of association with CRCI observed.

Overview

Brain-derived neurotrophic factor (BDNF) is among the best investigated proteins of the neurotrophin superfamily for its role in pathophysiology of neurocognitive disorders. BDNF is highly expressed in the central nervous system (CNS), concentrating in the prefrontal cortex and hippocampus, and functions to support neuronal survival, proliferation, and plasticity by binding to tropomyocin receptor kinase B (TrkB) [10,11]. Although the action and concentration of BDNF is the highest in the CNS, in vivo quantification of BDNF in the brain is impossible. Thus, human studies largely utilized circulating BDNF levels, in serum or plasma, as a proxy of CNS levels and neurogenesis activity [12]. Numerous studies have revealed significant correlations between circulating BDNF and neurocognition, whereby lower serum levels were found among those with Alzheimer’s disease compared to non-disease controls [13]. Plasma BDNF levels were reportedly higher in interventions that concurrently enhanced cognition [14-17]. Following this evidence, polymorphisms in the BDNF gene has garnered increased interest for explaining baseline risks in neurodegenerative diseases, with the Val66Met (rs6265) single nucleotide polymorphism (SNP) being the most investigated SNP for the gene. The single nucleotide substitution (G→A) at nucleotide 196 in the coding sequence causes a missense mutation, from valine (Val) to methionine (Met), at codon 66 of precursor BDNF protein which has been implicated in the abnormal sorting of BDNF into secretory vesicles at the Golgi apparatus and activity-dependent BDNF secretion [18]. However, the association between Val66Met and neurocognitive outcomes in non-cancer human studies have been inconsistent, with some meta-analyses reporting poorer outcomes among Met allele carriers [19-21], while other reviews finding no significant relationship [22-24]. It may be explained that the risk associated to Val66Met and cognitive impairment is disease-specific, and hence studies regarding Val66Met and CRCI are required.

Evidence Summary: Circulating BDNF and CRCI

Our working hypothesis is that lower circulating BDNF levels are related to CRCI occurrence among cancer patients and survivors. This has been demonstrated consistently in the research on plasma BDNF levels from cohort studies. In a multicentered, prospective, longitudinal study of 174 non-metastatic breast cancer patients, a smaller magnitude of reduction in BDNF levels was significantly associated with lower odds of developing self-perceived cognitive impairment (odds ratio [OR] 0.88, 95% confidence interval [CI] 0.79–0.99) during chemotherapy [25]. In the same cohort, among participants with sustained CRCI from between end of chemotherapy and up to 24 months post-chemotherapy, the decrease in BDNF levels due to chemotherapy exposure was more sensitive to increases in plasma IL-6 levels compared to other participants. This suggests that the pro-inflammatory state of the body during chemotherapy administration is a probable mechanism underlying lower neurogenesis activity in the brain [26].

Serum BDNF were more commonly analyzed in CRCI-focused studies compared to plasma BDNF based on a recent systematic review [27]. In the same review, it was found that greater consistency in positive BDNF-cognition correlation was observed among studies looking at BDNF levels in plasma compared to serum, albeit that more plasma BDNF studies did not report correlation findings [27]. Between serum and plasma, BDNF levels quantified from plasma may have greater predictive power of cognitive function as it is more reflective of the amount of free and unbound BDNF (BDNF physiologically binds to platelets) which equilibrates with CNS BDNF levels and, hence, greater correlation with BDNF activity in the brain [12]. Nevertheless, the review found no negative correlation between circulating BDNF (serum or plasma) and cognition among cancer populations, thus providing reproducible evidence for BDNF as a key mediator and biomarker in CRCI pathology [27].

Evidence Summary: Genetic BDNF Markers and CRCI

In the CRCI literature, the relationship between Val66Met and cognition has never been consistent. A meta-analysis of a temporally separate breast cancer cohort found lower odds of self-perceived CRCI among Met allele carriers compared to Val/Val genotypes (OR 0.52, 95% CI 0.29–0.94) [28]. This was facilitated by a lower reduction in plasma BDNF levels during chemotherapy among Met carriers [25]. Our systematic review in 2022, however, found largely inconsistent findings, with two studies reporting better cognition among Val/Val, one finding improved cognition among Met carriers, and the remaining ten studies reporting no relationship between any of the Val66Met genotypes and CRCI [27]. Other BDNF polymorphisms were rarely investigated, although one study had found preliminary evidence for rs10767664, rs10835210, rs11030104, and rs2030324 [29]. This suggests that a polygenic risk score looking at multiple SNPs might be necessary to understand how BDNF SNPs affect BDNF expression and function, and explain phenotypic changes in cognition.

Overview

Inflammatory cytokines play crucial roles in the CNS. On the physiological level, cytokines are known to mediate the neuronal and glial cell functioning and affect neuronal repair. Given that most of the current chemotherapeutic drugs are not known to penetrate the BBB due to their large molecular sizes, it has been postulated that tumor biology and/or cancer treatment itself alter cytokine expression and trigger downstream pro-inflammatory pathways that lead to neurological changes. Several characteristics, such as patients’ psychological distress alongside chemotherapy can also initiate the inflammation cascade, which may lead cytokine dysregulation. Considering the abundant cytokine receptors within the brain, processes that enhance cytokine expression or increase cytokine levels in the CNS may cause local inflammation via oxidative and nitrosative processes in the hippocampus, affecting learning and memory consolidation. In the literature, cytokines that have been investigated for their influence on cognition include interleukins (IL), chemokines, lymphokines, interferons (IFN), transforming growth factors (TGF), and tumor necrosis factors (TNF) [5]. Cytokines can be pro-inflammatory (e.g., IL-6, IL-1β, TNF-α, and IFN-γ) or anti-inflammatory (e.g., IL-4, IL-10 and IFN-α) in nature [5]. We speculate that the initiation of the proinflammatory cascade, due to cancer or treatment, may increase the risk for cognitive changes in cancer patients.

Evidence Summary: Circulating Plasma Cytokines and CRCI

Our systematic review conducted in 2013 has shown a weak to moderate correlation between cytokines (specifically IL-1β, IL-6, TNF-α) and different levels of cognitive impairment. Furthermore, we have observed that different types of chemotherapy may be associated with varying severity and presentation of cytokines-induced CRCI [30]. We also conducted a multicentered, cohort study that recruited stages I–III breast cancer patients receiving chemotherapy to evaluate the influence of cytokines on cognitive changes with prospective data. In the study, poorer response speed and self-perceived cognitive disturbances were associated with elevated plasma concentrations of pro-inflammatory IL-6 and IL-1β, but anti-inflammatory IL-4 may be protective against CRCI [5]. Therein lies evidence that CRCI pathogenesis is linked to pro-inflammatory processes and could be alleviated by anti-inflammatory mediators.

Given that the severity of CRCI had been found to differ over time and during patients’ treatment trajectories, our research group had previously characterized self-perceived cognitive trajectories into five distinctive types: no CRCI, as well as acute, intermittent, delayed, and persistent CRCI [31]. The existence of heterogenous cognitive trajectories seemed to suggest that the biomechanisms underlying different CRCI trajectories may differ. To evaluate how cytokines may directly or indirectly influence cognitive trajectories, we conducted a follow up study on the same cohort of patients to identify distinct cytokine profiles across the varying self-perceived cognitive trajectories. Interestingly, we have observed that plasma TNF-α was associated with acute CRCI (within the first 12 weeks of chemotherapy), whereas IL-6 and IL-8 were more persistently associated with persistent CRCI (cognitive impairment that occurred approximately 12 weeks after chemotherapy and at post 1-year follow up) [32].

Lastly, as both cytokines and BDNF are implicated in CRCI, we were interested to explore the relationship between these two groups of biomarkers. Hence, we have conducted a study to evaluate the associations of cytokines and BDNF among early-stage breast cancer patients with different CRCI trajectories. Interestingly, all cytokines analyzed showed inverse associations with BDNF levels. There was a significant interaction between IL-6 and persistent CRCI, which would impact on BDNF levels (p=0.026). The inverse associations with BDNF were more pronounced for IFN-γ, IL-1β, IL-4, IL-8, and GM-CSF in patients with persistent CRCI. The coefficient values for IL-2, IL-4, and TNF-α also indicate that there was a greater magnitude of decrease in BDNF levels for every unit of cytokine increase in patients with acute and persistent CRCI, compared to patients without CRCI. The differential relationships between cytokines and BDNF suggest that individuals could differ in their vulnerability to BDNF reduction in response to increased cytokine levels during chemotherapy, which may explain differential risk of CRCI observed in longitudinal cohort studies [32].

Evidence Summary: Cytokine Genes and CRCI

Expression of pro-inflammatory cytokines may be influenced by SNPs in the promoter regions of the pro-inflammatory cytokine genes. Hence, we have conducted a study [33] to evaluate the associations between two common pro-inflammatory cytokine gene polymorphisms namely, IL6-174 G>C (rs1800795) and TNF-308 G>A (rs1800629), and CRCI among Asian early-stage breast cancer patients. However, these SNPs did not play a major role to the cytokine fluctuations as well as CRCI in this cohort. Additionally, the differential effect of these SNPs on plasma IL-6 and TNF-α levels, and the associations of plasma IL-6 and TNF-α levels with CRCI were investigated. Consistent with our other studies, higher plasma concentrations of IL-6 were associated with greater severity of self-perceived cognitive impairment (p=0.001).

Overview

Although the inflammation cascade as well as the BDNF pathways are highly likely to explain the pathophysiology of CRCI, results in the literature have not been consistent. For example, different inflammatory cytokines have been observed to be associated with CRCI, and BDNF biomarkers are not consistently associated with the same types of cognitive tests. Hence, investigations are still worthy to explore other biochemical pathways that could be implicated in CRCI. In our research group, we have conducted studies to evaluate several potential CRCI biomarkers including extracellular vesicles (EVs) and exosomes, mitochondrial DNA (mtDNA), precursor hormones such as dehydroepiandrosterone (DHEA), sulfate form of DHEA [DHEA(S)], estradiol, and vascular endothelial growth factor (VEGF).

Evidence Summary: Extracellular Vesicles/Exosomes and CRCI

EVs are small vesicles secreted from eukaryotic cells sized between 30–1000 nm [34]. They are critical for CRCI due to their influence on cell-cell communications in neurodegenerative diseases [35]. A previous study on animal models infused with stem cell-derived microvesicles manifested that neuroplasticity related to cognition impairment can be improved thereby supporting the idea that there is a bridge existing between EVs and CRCI [36]. In an exploratory study of EVs and CRCI in breast cancer patients [37], the proteome of EVs was identified using mass spectrometry firstly from longitudinal pooled patients’ plasma samples. Differential expression in dynamin-1, tight-junction proteins, galactosylceramidase, p2X purinoceptor, cofilin-1, nexilin, and ADAM10 from pre-treatment samples were observed between CRCI and non-CRCI groups. These results confirm the hypothesis that EV cargoes play a role in anthracycline-induced CRCI and provided evidence for the modulation of CNS mechanisms and BBB in CRCI pathogenesis. For example, longitudinal expressions of dynamin-1, which is known to have an impact on synaptic transduction function in vitro [38], were found to have more greatly decreased trends over time when comparing CRCI and non-CRCI groups. These results emphasize the potential role of dynamin-1 downregulation in CRCI and set the groundwork for future validation research.

Evidence Summary: Mitochondrial DNA and CRCI

Chemotherapeutics in breast cancer trigger mitochondria dysfunction and reactive oxygen species release, which further induce neurodegeneration [38] and chronic fatigue syndrome [39]. Since mtDNA is essential in the regulation of mitochondrial function and can be measured via peripheral blood [40], a cohort study with 108 patients participating published in 2017 indicated that cancer-related fatigue (CRF) was associated with the reduction the mtDNA [41]. The mtDNA content was found to be lower in CRF patients at baseline (p<0.001). Both the proportion of CRF-worsening patients and inner scores of MFSI-SF are associated with the reduction of mtDNA. However, self-perceived CRCI was not associated with mtDNA volume.

Evidence Summary: DHEA/DHEA(S)/Estradiol and CRCI

Previous studies suggest that hormones like estradiol have a connection to human cognitive function, as suppression leads to cognitive process interruption [42] and overexpression causes better memory outcomes [43]. Both enhancement of metabolism precursors, pre-chemotherapy DHEA and DHEA(S) levels are responsible for the synthesis of sex hormones reported by our team to result in lower odds of CRCI [44]. In our longitudinal study, we have identified DHEA(S) and estradiol downregulation after initiating cancer chemotherapy treatment and this is our first study that observed an association between declining DHEA(S) levels and acute onset of CRCI at 6 weeks from baseline [45]. Although patients reporting CRCI showed a greater magnitude of decline in estradiol compared to non-CRCI patients, the differences in decline was not found to be statistically significant.

Evidence Summary: VEGF and CRCI

VEGF is famous for mediating cancer angiogenesis as one of the most important hallmarks of cancer. Not surprisingly, VEGF also proved to have an influence on neuron function and cognition [46]. A number of case reports and case series have reported the association between cognitive impairment and VEGF inhibitors use [47]. However, in our observation study of breast cancer patients receiving anthracyclines, although there was a change of median plasma VEGF levels over time as compared to the baseline, plasma VEGF levels were not associated with CRCI [48].

Despite the widespread use of biomarkers to predict the risk of cancer development, the use of biomarkers in cancer supportive care has been relatively unexplored. The term “supportive care” encompasses managing the symptoms and toxicities of cancer treatment through treatment and survivorship, and CRCI is a perfect case study because its presentation can occur during and beyond treatment. Furthermore, CRCI compromises patients’ functional capabilities, reduces social role functioning and their ability to work as they try to resume their pre-cancer roles and lifestyle. Their impact on cancer patient’s quality of life has gained recognition and is known to greatly affected both patients and survivors alike, in the current healthcare landscape. Besides adding on to emotional and psychological distress, they may also have economic consequences that may extend to caregivers and family members [49].

This premise leads to our research motivation to identify biomarkers that correlate well with the extent of symptom burden in post-cancer treatment side effects such as CRCI. Biomarkers are often used to provide the link between measurement and prediction of clinical outcome assessment [50], and they could serve as susceptibility biomarkers evaluating the risk of patients for developing certain condition, and/or prognostic biomarkers which are quantified in an individual who currently have the condition for disease progression evaluations. In addition, there is potential for serial measurement to serve a monitoring role to assess the treatment-related toxicities. Given that CRCI can be measured both subjectively using patient-reported outcome tools and objectively through neuropsychological batteries [51], biomarkers are able to provide another avenue to assess patients’ cognitive symptoms with additional objectivity.

With advances in analytical methods and bioinformatics, there are more measurable biological processes that may be explainable beyond physiological model systems to changes at cellular levels. Our evaluation of the various biomarkers of CRCI have progressed from simple biometric measurement to more quantitative analyses of genomic or proteomic analytes and to even a collective panel or index. These technological advances will allow us to continue to investigate the underlying mechanisms of CRCI in the most efficient manner as the condition is still lacking effective management at this time.

Although much of our effort in the past 15 years was devoted in the discovery of potential biomarkers in CRCI, much work is still needed in the translations into clinical applications. We have recently evaluated the role of BDNF augmentation in an animal model, with the goal of repurposing medications that may be helpful for managing CRCI [52]. Additionally, besides finding specific agents that can target or mitigate these biochemical changes, there is a need for us to draw upon the capacity of biomarkers to distinguish between biological states. To ensure the clinical utility of biomarkers, we need to take into context how biomarker levels may be altered over time due to patient-related factors such as those with inflammatory or autoimmune disease and psychosocial factors. In addition, biomarkers that are identified must be (1) validated in independent patient cohorts and evaluated in its performance of sensitivity and specificity and (2) validated in preclinical settings to evaluate these biomarkers in much controlled settings. Lastly, majority of our CRCI biomarker research were conducted in breast cancer patients as well as adolescent and young adult cancer patients. Additional validation must be conducted in other disease populations where patients are at risk for CRCI.

Management of CRCI continues to remain as a challenge within the cancer supportive care community. Research performed by our team over the past 15 years has focused on identifying biomarkers that may explain the pathophysiology of this debilitating condition. Promising findings related to BDNF, inflammatory cytokines, and dynamin-1 warrant further investigations on their potential to translate observational results into experimental therapeutics. The quest for effective therapeutics will continue to unfold as we continue to learn more about the pathophysiology and associated biochemical process behind CRCI.

None.

None.

No potential conflict of interest relevant to this article was reported.