Chemotherapy-Induced Cognitive Decline: Moving from the Mechanistic Debate towards Prevention and Treatment—A Clinical Review

Patients receiving chemotherapy have reported cognitive challenges including short-term memory loss and reduced executive functioning. While cognitive decline can be multifactorial and related to aging, depression, surgery, and other medications, there has been a steadily increasing body of knowledge showing a significant association between cognitive decline and chemotherapy administration. This clinical review summarizes patient-reported cognitive changes, support from neuroimaging and neuropsychological testing. The mechanism of action of and patient susceptibilities to cognitive decline are reviewed. Current behavioral and pharmacologic interventions are discussed. There is a need to identify patients at risk for developing chemotherapy induced cognitive decline and to screen for early signs of cognitive deterioration. The risk of cognitive dysfunction and possible interventions should be included in the informed consent discussion with patients who are undergoing cytotoxic treatments. Controlled


Introduction
Recent advances in the diagnosis and treatment of cancer have led to an increased number of patients entering survivorship [1] [2]. Increased survival rates dictate that more resources should be directed towards diagnosing and managing the issues affecting patients' post-treatment quality of life. Among these, chemotherapy-induced cognitive decline (CICD) has emerged as an important sequelae of therapy.
The paramount challenge in any discussion of CICD is discerning the potential cognitive harms of cytotoxic drugs from an assortment of possible confounders, including age-related cognitive deterioration, the adverse effects of concurrent treatment modalities, and disease-related factors stemming from the cancer itself (Table 1) [3]- [19]. For instance, long-term adjuvant endocrine therapy (ET) is a standard treatment for hormone-positive breast cancer (BC) and is a plausible etiology for cognitive dysfunction in treated women [4] [5] [6]. While some studies have reported ET-related cognitive impairments over a short-term follow-up period, a recent longitudinal study which observed ET-treated patients for up to six years failed to demonstrate such an association [7]. Major oncological surgery is another candidate mechanism, with data showing post-surgical cognitive decline in a substantial percentage of patients, most notably in elderly populations [8]. Chronic opioid usage for management of cancer-related pain is similarly associated with cognitive deficits in a dose-related manner [9]. Van Dyke, et al. [7] Prospective longitudinal No association found between long-term endocrine therapy and neurocognitive performance Major oncologic surgery Plas M, et al. [8] Prospective longitudinal -12% of patients overall exhibit cognitive decline at 3-months following major oncologic surgery -18% of patients aged >75 years exhibit cognitive decline at 3 months following major oncologic surgery Chronic opioid usage Kurita GP, et al. [9] Prospective cross-sectional, multi-center -A third of opioid-treated patients exhibit possible (MMSE a score 24 -26)  and 92% reported persistent difficulties with cognitive function after five years [22]. Today, it is evident that the phenomenon, colloquially named "chemo-brain" or "chemo-fog", is experienced by patients across a variety of solid and hematologic malignancies [23]. Of note, the majority of the evidence is in breast cancer (BC) patients who tend to be young and highly functioning and may notice even mild perceived deficits which may limit the generalizability of these studies to cancer patients as a whole.
Estimates of the prevalence of SRCICD have varied substantially across studies (see Table 2) [24]- [33]. One systematic review of twenty-seven studies in BC patients, reported prevalence rates ranging from 21% to as high as 90%, with the most frequently reported deficits affecting the domains of memory, concentration and executive functioning [34]. One of the largest longitudinal studies to address SRCICD was published in 2016 by Janelsin and colleagues [30]. The study com-

Diagnosing CICD Using Neuropsychological Testing
Given the high prevalence of patient-reported cognitive complaints, substantial effort has been put into empirically diagnosing and quantifying CICD, herein called "objectively-verified CICD" (OVCICD). Reports vary significantly in terms of the proportion of patients affected and the cognitive domains involved: some studies have shown marked impairments, most commonly in the domains of memory, attention, concentration, executive function and processing speed; others show only subtle impairments or no impairments at all [35].
The lack of a uniform research methodology seems to account for these inconsistencies. Researchers have employed different study designs (mostly crosssectional, some longitudinal), different control groups (healthy controls versus chemo-naive cancer patients), and different cut-off scores for diagnosing cognitive impairment [36] [37]. They have also timed the testing sessions differently relative to the time when chemotherapy was administered and have employed different testing batteries shown to differ in their respective sensitivity and specificity [38].
In 2011, the lack of consensus on how to investigate OVCICD compelled the International Cognitive and Cancer Task Force (ICCTF) to publish recommendations for standardizing CICD research [39]. In 2012, a meta-analysis of 17 studies looking at 807 patients evaluated neuro-psychologic testing of eight cognitive domains: attention, executive functioning, information processing, motor speed, verbal ability, verbal memory, visual memory, and visuospatial ability [40]. The analysis found that two cognitive domains: verbal and visuospatial ab-ilities were most impacted by chemotherapy; verbal ability was worse in BC patients treated with chemotherapy compared to healthy controls, and visuospatial ability was worse compared to chemo-naïve BC patients.
More recently, in 2017 the largest meta-analysis on the matter to date further reinforced the notion that the presence or absence of OVCICD heavily depends on the type of control group used [41]. The study incorporated 2939 BC patients from seventy-two prior studies, both cross-sectional and longitudinal. In an analysis of overall as well as domain-specific cognitive impairment, patients treated with chemotherapy had lower cognitive scores compared to healthy non-cancer controls. Crucially, however, chemotherapy-treated patients performed equally compared with cancer patients not treated with chemotherapy. The analysis strongly suggests that chemotherapy is not a driving factor for cognitive decline in cancer patients, at least as it pertains to its diagnosis using neuropsychological testing.
On the other hand, in a counter argument for the direct culpability of chemotherapy, Collins and colleagues demonstrated a significant dose-response relationship between chemotherapy and objective cognitive decline in a cohort of 60 BC patients, compared to a healthy control group, and after controlling for pre-treatment cognitive baseline [42]. Dose response effect of chemotherapy was also identified in a case cohort study [20]. At two years following chemotherapy completion, cognitive impairment was found in 32% of the patients treated with

Neuroimaging
In contrast to the variable results of neuropsychological testing, imagingbased studies have produced more consistent evidence for an obvious, measurable effect of cytotoxic drugs on the brain. These neuroimaging and the correlating anatomic brain changes have also been documented in early onset dementia syndromes, which present with similar alterations in memory and executive function [43]. Chemotherapy not only influences the brain's morphology but its hemody- Journal of Cancer Therapy namics as well. In a longitudinal follow-up of twenty-seven BC patients using pulsed arterial spin-labeling MRI, significantly increased cerebral perfusion was observed in the right precentral gyrus one month after completing cytotoxic treatment [52]. The authors postulated that this might reflect a compensatory hemodynamic response to treatment-induced neural damage. This increase in perfusion was negatively correlated with pre-treatment cognitive function, suggesting that lower cognitive reserve may be a risk factor for post-treatment cerebral perfusion dysregulation. Similarly, Chen and colleagues reported significant increases in cerebral blood flow across various brain regions following neoadjuvant therapy for BC which was significantly correlated with reduced performance on various attention tasks [53].
Finally, the apparent structural and hemodynamic changes induced by chemotherapy seem to coincide with functional alterations, as a growing number of functional MRI (fMRI) studies now show. For example, Miao et al. evaluated the long term chemotherapy-related functional changes to the anterior cingulate cortex (ACC) using fMRI in twenty-three chemotherapy-treated BC patients as compared to twenty-six healthy control subjects [54]. The results showed that functional connectivity was significantly lowered in the chemotherapy group and the observed changes were correlated with a reduction in executive function abilities as demonstrated in the Stroop Interference Test. Another study aimed to assess the long-term impact of cisplatin-based chemotherapy on whole-brain networks in testicular cancer patients who were recently orchiectomized [55].
Sixty-four patients underwent baseline and six-month follow-up fMRI imaging and neuro-cognitive testing. Of this cohort, twenty-two subjects were treated with cisplatin and forty-two were under surveillance only. Analysis showed that in patients who had received cisplatin, key connectivity properties of the brain were altered which affect distribution of information across the brain, both on the local as well as the global level [56]. Changes to these measures might reflect suboptimal cognitive abilities and reduced tolerability to local insult, and indeed the imaging findings correlated with poorer overall cognitive performance in the treated group.
A Chasm between SRCICD and OVCICD Taken as a whole, the data on OVCICD reveals an obvious discrepancy between patients' subjective experience of CICD, which is often substantial and crippling, and the underwhelming neuropsychological test results. In an illuminating systematic review, Hutchinson and colleagues analyzed 24 prior studies that used objective and subjective measures simultaneously to diagnose CICD in the same cohort. Of the included studies, only eight reported on a significant correlation between the two measures [57]. This finding suggests that SRCICD and OVCICD might be two independent phenomena: some patients have objective cognitive decline which is too subtle to interfere with their daily lives and thus goes unnoticed and unreported; in other patients, the burden of chemotherapy creates the subjective experience of impairment without any measurable

CICD Mechanisms
The clinical findings, as well as the functional and morphologic brain findings associated with CICD, are most likely the endpoint of multiple mechanisms acting on the brain in synergy. The three most thoroughly developed hypotheses as to the underlying causes are described below.

CICD in the Context of Accelerated Aging
It has been argued that CICD can best be understood in the broader context of chemotherapy-induced accelerated aging [67]. Evidence pointing to a hastening of the aging process brought about by cancer and/or its treatment is abundant across studies ranging from animal models to epidemiological analyses.
From a clinical standpoint, physiological frailty, which includes cognitive deterioration, can be considered a "physical phenotype" of aging [68]. Interestingly, a study conducted by Ness and colleagues in 2005 has shown frailty to be prevalent in young adult survivors of childhood cancer at a similar rate to that encountered among adults sixty-five years old and above [69]. On the molecular level, chemotherapeutic treatment has been shown to negatively affect several well-identified biological markers of aging [70]. shown to lead to a decline in cognitive functions, and is associated with some neurodegenerative disorders and age-dependent decay of neuroplasticity [86].
This mechanism might be shared by other common ROS-generating anti-neoplastic drugs such as cyclophosphamide and methotrexate [87]. It also opens the door to several potential therapeutic targets to prevent ROS-mediated CICD, for example, co-administering anti-TNF antibodies [85]. Animal experiments have shown that doxorubicin-induced oxidative stress is ameliorated by administering 2-mercaptoethan sulfonate sodium (MESNA), an anti-oxidant commonly given to patients as part of drug regimens containing cyclophosphamide or ifosfamide to prevent the occurrence of hemorrhagic cystitis [88].

Susceptibility to CICD
It is important to know if there are predisposing risk factors for CICD as determination of baseline susceptibility will allow better risk stratification and patient counseling and may alter decisions about treatment choices.  [90]. In a more recent study, testicular cancer survivors who were heterozygous or homozygous for the ε4 allele and had been treated with BEP (bleomycin, etoposide, cisplatin) performed worse on cognitive tests compared to chemotherapy-treated patients who did not carry the allele [47].

Pharmaco-Genetic Determinants
Another avenue of investigation concerns genetic polymorphism which might render the CNS more vulnerable to penetration by intra-venously administered neurotoxic drugs which normally cannot cross into the CNS due to BBB impenetrability [91]. This impermeability to cytotoxic drugs is dependent upon the functioning of local drug transporters, chief among them the P-glycoprotein (p-gp) efflux transporter and the OATP1A2 influx pump [92] [93]. Since many commonly used anti-neoplastic drugs are known substrates of P-gp and OATP1A2, it has been posited that polymorphisms in the genes encoding for these transporters (ABCB1 and SLCO1A2, respectively) might play a role in an individual's susceptibility to CICD [89]. Studies exploring the association between ABCB1 single-nucleotide polymorphisms (SNPs) and chemotherapy induced toxicity have been conflicting, and few studies included CNS toxicity and/or cognitive impairment in their assessments [94]. Of note is a single study by Erdilyi et al.
who retrospectively assessed the correlation between chemotherapy-associated encephalopathy and ABCB1 genotypes in 291 acute lymphoblastic leukemia patients [95]. The genotypes of an additional gene in the same ATP-binding cassette transporter family of genes, ABCG2, were also examined. The authors showed that carrying the ABCB1 3435 TT genotype confers a higher risk of treatment-induced encephalopathy, while carrying both a mutated ABCB1 and a mutated ABCG2 allele results in an even higher risk, suggesting a synergistic effect of the two polymorphisms together. Whether this finding is generalizable to the realm of CICD remains debatable, as there are multiple mutations and genetic polymorphisms in this gene family and further investigation is necessary [96]. With regards to a similar interaction between cognitive reserve and SRCICD, the opposite could be expected, namely that well-educated patients with occupations requiring high-level functioning might be more sensitive to even the slightest cognitive changes and thus more likely to report on cognitive symptoms.

Baseline Cognition as a Potential Moderator of CICD
Interestingly, however, limited evidence suggests that increased cognitive reserve attenuates the subjective perception of CICD [30]. In a longitudinal study using the Functional Assessment of Cancer Therapy-Cognitive Function (FACT-Cog) questionnaire, Janelsins and colleagues found that decreased reserve prior to chemotherapy significantly correlates with lower FACT-Cog scores in BC patients [30]. Similar to the previous study by Ahles et al., baseline reserve was assessed through the proxy of reading ability, using the Wide Range Achievement Test (WRAT) [99].

Future Areas of Investigation
With the development of targeted chemotherapeutic agents, the potential for novel drug classes to contribute to or worsen chemotherapy induced cognitive dysfunction is real. Emerging therapies such as immunotherapies and targeted therapies will need to be studied to identify any cognitive side-effects in treated patients.
Additionally, several pharmacologic and nonpharmacologic interventions have been evaluated as potential therapies for cognitive decline and CICD which may be used as adjuncts during conventional chemotherapeutic treatment.

Emerging CICD Culprits
Of novel cancer therapies, perhaps the most attention should be paid to im-  [105]. Performance in two neuropsychological tests (MoCA and TNI-93) remained stable or improved at the three-month mark, with no evidence of cognitive dysfunction due to the treatment in any of the patients. However, further larger-scale prospective studies will be required in order to determine whether or not ICIs exert any influence on cognition. One methodological challenge in future studies will be controlling for the effects of chemotherapy, since many patients exposed to immune checkpoint blockade would have already been treated with often multiple lines of cytotoxic chemotherapy.

Interventions for CICD Prevention and Treatment
There are currently no FDA-approved drugs for the prevention or treatment of CICD, and no quality data to support the endorsement of any such interventions. However, preliminary clinical and pre-clinic studies have shown promising results, which, at a minimum, should encourage further investigation (see Table 3). Anti-TNF antibody Mice treated with systemic doxorubicin (intraperitoneal injection) with/without anti-TNF antibody -TNF levels in brain tissue were significantly elevated following doxorubicin treatment (p < 0.01) -Measures of brain mitochondrial function were significantly reduced following doxorubicin treatment (p < 0.05) -Anti-TNF antibody administration prevented the increase of central TNF levels, as well as the decline in mitochondrial function Keeney et al. [88] 2-mercaptoethanesulfo nate sodium (MESNA) Mice treated with systemic doxorubicin (intraperitoneal injection) with/without MESNA -Indicators of oxidative stress (protein carbonyl, protein-bound 4-hydroxynonenal) were significantly elevated in the sera and brain tissue of mice following doxorubicin administration (brain: p < 0.01; sera: p < 0.0001 for protein carbonyl, p < 0.001 for protein-bound 4-hydroxynonenal) -Novel Object Recognition (NOR) was significantly reduced in doxorubicin-treated mice (p < 0.05) -MESNA administration before and after doxorubicin ameliorated the rise of oxidative stress measures in brain (p < 0.01) and sera (p < 0.01 for protein carbonyl, p < 0.05 for protein-bound 4-hydroxynonenal) -MESNA administration prevented the doxorubicininduced deterioration in NOR Zhou et al. [106] Metformin Mice intra-peritoneally treated with cisplatin with/without metformin -Exposure to cisplatin significantly reduced performance in the Novel Object and Place Recognition Test (NOPRT) (p < 0.05), an effect that was not exhibited in subjects treated concurrently with metformin co-administering the acetylcholine esterase inhibitors donepezil and galantamine together with chemotherapy [108]. Interestingly, the administration of the same drugs after completion of chemotherapy did not prevent the development of these deficits in a separate mice cohort.

Clinical Studies
Extrapolating from experience with Alzheimer's disease, donepezil has also been evaluated in the clinical setting [109]. In Phase 3, randomized placebo-controlled trial among brain tumor survivors who received radiation, administration of donepezil did not improve patients' overall cognitive scores but did elicit a modest benefit in the domains of memory, dexterity and motor speed [110]. In a pilot study of BC survivors who received prior chemotherapy, daily 5 -10 mg of oral donepezil improved patients' performance on two memory tests as compared to controls. However, no improvement was demonstrated in other cognitive domains or in self-reporting of cognitive functions [111]. Given this conflicting evidence, further phase 3 trials are required to elucidate the potential role of anti-cholinergic medications in countering CICD symptoms.
Another agent under investigation is modafinil, routinely used as a first-line pharmacologic treatment for narcolepsy-associated daytime sleepiness. In cancer patients, there is some limited evidence for cognitive improvement with modafinil therapy [112] [113]. Kohli  phase of the study, modafinil was found to significantly improve memory and attention skills compared to placebo [112]. A separate trial evaluated modafinil in the palliative setting for 28 patients with advanced cancer and a high tiredness score. The drug evoked superior performance compared with placebo in two cognitive tests: the Finger Tapping Test (FTT) for evaluation of psychomotor speed, and the Trail Making Test (TST) for assessing visual attention and task switching [113]. Conversely, evaluations of modafinil for cognitive dysfunction as a secondary outcome in two studies of cancer-related fatigue (CRF) found no improvement in any of the administered cognitive tests after treatment [114] [115]. A recent meta-analysis of 19 placebo-controlled trials in non-sleep-deprived adults showed only limited ability of modafinil to improve cognition outside the already established setting of sleep-deprivation [116].
Finally, in 2014, a randomized placebo-controlled crossover trial evaluated the effect of methylphenidate on cognitive performance as a secondary outcome among 33 women with BC undergoing chemotherapy [117]. CRF, the primary endpoint of the study, was not improved by the intervention, however, treated patients performed better on tests of memory, scanning speed, verbal learning and visual perception, suggesting a potential role for methylphenidate in alleviating CICD.

Non-Pharmacological Interventions
Given the dearth of effective pharmacologic treatments, there has been significant interest in developing behavioral interventions to treat CICD. Efforts to apply mindfulness-based interventions to the problem of CICD has produced some evidence of benefit, however, results across studies are conflicting [118] [119] [120]. Cognitive rehabilitation strategies, aimed at restoring damaged cognitive skills through re-training and the development of compensatory mechanisms, have also been examined, beginning with a pilot study in 2007 which showed initial promise for reversing changes in attention and memory [121].

Conclusion
Taken together, the evidence confirms the existence of CICD as a substantial, albeit subtle, clinical issue for cancer patients. However, it is important to rule out other potential causes of cognitive dysfunction. While formal neurocognitive testing might not be sensitive enough to detect CICD, the impact on patients' quality of life is unmistakable and we believe that this merits the inclusion of CICD in any pre-treatment consent discussion in the same way as other better-established risks of chemotherapy. In aiming to improve the quality of life for cancer survivors, we see a need for better measures in two key fields: identifying who is most at risk of developing CICD, and screening for early signs of cognitive deterioration. A third field-prompt intervention is still lacking in actionable data and requires further rigorous investigation.

Conflicts of Interest
The authors have no disclosures or conflicts of interest.