Inflammation, Attention, and Processing Speed in Patients With Breast Cancer Before and After Chemotherapy

Abstract

Background

Inflammation may contribute to cognitive difficulties in patients with breast cancer. We tested 2 hypotheses: inflammation is elevated in patients with breast cancer vs noncancer control participants and inflammation in patients is associated with worse attention and processing speed over the course of chemotherapy.

Methods

Serum cytokines (interleukin [IL]–4, 6, 8, 10; tumor necrosis factor [TNF]–α) and soluble receptors [sTNFRI, II]) were measured in 519 females with breast cancer before and after chemotherapy and 338 females without cancer serving as control participants. Attention and processing speed were measured by Rapid Visual Processing (RVP), Backward Counting (BCT), and Trail Making-A (TMT-A) tests. Linear regression models examined patient vs control cytokines and receptor levels, adjusting for covariates. Linear regression models also examined relationships between patient cytokines and receptor levels and test performance, adjusting for age, body mass index, anxiety, depression, cognitive reserve, and chemotherapy duration. Statistical tests were 2-sided (α = .05).

Results

sTNFRI and sTNFRII increased over time in patients relative to controls, whereas IL-4, IL-6, and IL-10 decreased. Prechemotherapy, higher IL-8 associated with worse BCT (β = 0.610, SE = 0.241, P = .01); higher IL-4 (β = −1.098, SE = 0.516, P = .03) and IL-10 (β = −0.835, SE = 0.414, P = .04) associated with better TMT-A. Postchemotherapy, higher IL-8 (β = 0.841, SE = 0.260, P = .001), sTNFRI (β = 6.638, SE = 2.208, P = .003), and sTNFRII (β = 0.913, SE = 0.455, P = .045) associated with worse BCT; higher sTNFRII also associated with worse RVP (β = −1.316, SE = 0.587, P = .03). At prechemotherapy, higher IL-4 predicted RVP improvement over time (β = 0.820, SE = 0.336, P = .02); higher sTNFRI predicted worse BCT over time (β = 5.566, SE = 2.367, P = .02). Longitudinally, increases in IL-4 associated with BCT improvement (β = −0.564, SE = 0.253, P = .03).

Conclusions

Generally, worse attention and processing speed were associated with higher inflammatory cytokines and receptors and lower anti-inflammatory cytokines in patients; future confirmatory studies are needed.

Cancer- and chemotherapy-related cognitive decline (CRCD) impacts up to 70% of patients during and after chemotherapy for breast cancer and approximately 30% of patients before treatment (1-7). The etiology of CRCD is unknown, but inflammation may play a role (3,8,9). Inflammation is implicated in the development of cognitive difficulties related to aging, infection, neurodegenerative disorders, chronic illness, and postsurgical recovery (10–23). Risk factors for cancer, the presence of tumors, and cancer treatments are associated with elevations in several inflammatory markers, warranting an investigation of the link between inflammation and cognitive function in patients with cancer (24,25).

Most research on inflammation and cognitive function in patients with breast cancer has focused on the time during and after chemotherapy (2,3,9,24–34). With some exceptions (33,34), these studies demonstrate detrimental relationships of pro-inflammatory cytokines and beneficial relationships of anti-inflammatory cytokines with cognitive function. For example, higher pro-inflammatory cytokine interleukin (IL)-1β was associated with worse response speed, whereas anti-inflammatory IL-4 was associated with better response speed in patients with breast cancer receiving chemotherapy (27). In another study of patients with breast cancer receiving chemotherapy, worse visual memory performance was associated with higher levels of soluble receptors for pro-inflammatory tumor necrosis factor alpha (TNFα) type I and II (sTNFRI and sTNFRII) (30). However, the relationship between inflammation and cognition is complex; cytokines and receptors exhibit statistically significant relationships with some cognitive tests but not others, and the strength and direction of association is not consistent across studies (15,24,25,35,36). Additional studies in large cohorts are necessary to elucidate the relationships of inflammatory markers with cognitive function.

Few studies have investigated cognitive function and cytokines and receptors in patients with breast cancer prior to chemotherapy (2,24). In 174 patients, higher sTNFRII was associated with worse memory function prior to treatment (2). In 75 chemotherapy-naive patients, lower levels of IL-17 and higher granulocyte colony-stimulating factor related to better executive functioning, psychomotor speed, and response time (24). Although these studies show a link between prechemotherapy inflammation and cognitive function, additional studies are needed to longitudinally assess relationships between inflammation and cognitive change.

The goal of the present study was to evaluate serum cytokine and receptor levels in patients with breast cancer before and after chemotherapy compared with participants without cancer (serving as study controls) and to assess relationships of cytokine and receptor levels with tests of cognitive function in patients. Because of the nuanced relationships between inflammatory factors and various aspects of cognitive function, we chose to focus specifically on tests of attention and processing speed, because deficits in these areas can undermine performance in other cognitive domains and impair everyday function (37–41) and data from aging research show a relationship between inflammation and these domains (42). We hypothesized that patients would have greater levels of pro-inflammatory markers and lower levels of anti-inflammatory markers than noncancer control participants before and after chemotherapy. We also hypothesized that, in patients, greater inflammation would be associated with worse test performance before and after chemotherapy. Lastly, we evaluated whether prechemotherapy cytokines and receptors predicted decreases in test performance over chemotherapy and whether changes in cytokines and receptors co-varied with changes in test performance over time.

Methods

Study Participants

This analysis is part of a nationwide, prospective longitudinal cohort study designed to study cognitive changes before and after chemotherapy in patient participants with cancer compared with participants without cancer assessed at equivalent times (1) (URCC10055; ClinicalTrials.gov identifier: NCT01382082). Participants providing blood samples were recruited from 17 National Cancer Institute Community Oncology Research Program (NCORP) locations across the United States from May 2011 to October 2013. Eligibility criteria for patients included women who were diagnosed with stage I-IIIC breast cancer (43), scheduled to receive a standard course of curative-intent chemotherapy, chemotherapy-naïve, able to read and speak English, aged 21 years or older, and able to give written informed consent and who had a life expectancy of 10 or more months, no central nervous system disease, no neurodegenerative disease, and no recent psychiatric illness requiring hospitalization. Patients with planned radiation therapy were excluded given the goal that the parent study was to evaluate cognitive function over chemotherapy specifically. Noncancer control participants were allowed to be family members or friends of patient participants or were recruited from the community. Control participants had no cancer history and met all relevant eligibility criteria. All participants provided informed consent. The institutional review boards of each NCORP site and the University of Rochester approved this study.

Timing of Assessments

Cognitive assessments and blood draws occurred at the prechemotherapy assessment (A1) and postchemotherapy assessment (A2). The A1 blood draw was on average 2.7 (SD = 5.0) days before patients started their first chemotherapy cycle. Eight patients who had “prechemotherapy” blood drawn after starting chemotherapy were excluded. The A2 blood draw was on average 30.9 (24.5) days after the last chemotherapy administration. Cytokines and receptors for noncancer control participants were assessed at the equivalent time points. The average latency between A1 and A2 blood draws was 19.5  (5.2) weeks for patients and 19.0 weeks (SD = 7.7) for controls.

Cytokine Assays

A panel of cytokines (interleukin [IL]-4, IL-6, IL-8, IL-10, TNFα) and soluble receptors (sTNFRI, sTNFRII) was examined based on their relationships to cognitive function in prior studies (2,3,9,24,27,29–32). Cytokines and receptors were analyzed using standardized Milliplex xMAP human cytokine (high sensitivity) and cytokine receptor immunoassay kits (HSCYTOMAG-28SK, HSCRMAG-32K; Luminex Corp, Austin, TX) following the manufacturer’s instructions. Additional details are in the Supplementary Methods (available online).

Cognitive Assessments

The Rapid Visual Processing Task (RVP) of the Cambridge Neuropsychological Test Automated Battery (Cambridge Cognition, Cambridge, UK) tests sustained attention and processing speed (44–49). The backward counting task (BCT) tests processing speed and attention (50). The Comprehensive Trail-Making Test Trail 1 (denoted as TMT-A herein) tests visual attention, visual scanning, and processing speed (51–55). To ensure uniformity across sites, research personnel were trained by the study team to administer the assessments and followed a standardized script and manual. Additional details are in the Supplementary Methods (available online).

Covariates

Demographic information (eg, race, menopausal status) was self-reported at A1. Collection of data on race and ethnicity was required by the funding agency. Race categories on the participant self-report form were American Indian or Alaskan Native, Asian or Asian American, Black or African American, Native Hawaiian or Pacific Islander, White, or unknown. Patient medical information was abstracted from medical records. Reading ability, a proxy for cognitive reserve, was measured with the Wide Range Assessment Test Fourth Edition reading subscale (56) at A1. Trait anxiety was assessed by the Spielberger Trait Anxiety Inventory and depressive symptoms by item 21 of the Multidimensional Fatigue Symptom Inventory at A1 (57,58). Body mass index (BMI) was calculated ([weight in kg]/[height in m]²) (59).

Statistical Analyses

Pearson χ² tests (categorical values) and 2-sided t tests (continuous variables) were used to compare patient vs control characteristics at A1.

Linear mixed models were used to assess between-group differences in cytokines and receptors at prechemotherapy (A1) and postchemotherapy (A2) and the differential pre-to-postchemotherapy change (change score [A2-A1]) between groups (patients vs controls). In these models, cytokine and receptor levels at A1, A2, and the A2-A1 change scores were separately modeled as dependent variables, and group was an independent variable. Models were adjusted for age (continuous), BMI (continuous), race (White vs non-White), and education (high school or below vs some college or above), which were selected as covariates a priori because of their association with inflammation (60,61). Plate was modeled as a random effect to account for variability between batches. Within-group changes in adjusted cytokine and receptor levels were assessed by testing the null hypothesis (A2-A1 change score = 0) for each group.

Relationships between cytokines and receptors and test performance were analyzed for patients only. Linear regression models were used to evaluate cross-sectional relationships between 1) cytokines and receptors and cognitive test scores at A1 and 2) cytokines and receptors and test scores at A2. Linear regression models were also used to evaluate longitudinal relationships between 1) cytokines and receptors at A1 and the change in test scores (A2-A1) and 2) the change (A2-A1) in cytokines and receptors and the change (A2-A1) in test scores. Models where the change (A2-A1) in test score was the dependent variable were adjusted for the A1 test score. Age, BMI, chemotherapy duration (date of first chemotherapy to last chemotherapy administration), cognitive reserve, anxiety, and a depression item score (all continuous variables) were selected as covariates a priori for all models given their relationships to both inflammation and cognitive function (1,61–64).

For cytokines and receptor levels, results are reported as adjusted means and standard errors estimated from the regression models. Analyses with unadjusted cytokine levels are presented in the Supplement (available online). All P values are 2-sided. A P value less than .05 was considered statistically significant. Given the exploratory nature of these analyses and focus on biological importance at this stage of research, no multiplicity adjustments were performed.

Results

Participant Characteristics

Of the 580 patient participants with breast cancer and the 363 noncancer control participants in the parent study (1), serum was available for 519 (89.5%) patients and 338 (93.1%) controls for at least 1 time point. Patients (controls) were on average aged 53.3 (52.8) years, 97.7% (97.0%) not Hispanic or Latino, 53.4% (49.7%) postmenopausal, 71.9% (76.3%) married or in a long-term relationship, and 75.1% (88.2%) had at least some college (Table 1). Race differed between the groups, as 1.2% (0.0%) of patients (controls) were American Indian or Alaska Native, 1.7% (0.9%) Asian or Asian American, 8.3% (4.1%) Black or African American, and 88.8% (95.0%) White. On average, patients had less education, more anxiety and depressive symptoms, lower cognitive reserve, and higher BMI than noncancer control participants.

Table 1.

Participant characteristics

Characteristic	Breast cancer	Control	P
	(N = 519)	(N = 338)
Age, y			.48a
Mean (SD)	53.3 (10.6)	52.8 (10.3)
Range	22-81	27-81
Race, No. (%)			.01b
American Indian or Alaska Native	6 (1.2)	0 (0.0)
Asian or Asian American	9 (1.7)	3 (0.9)
Black or African American	43 (8.3)	14 (4.1)
White	461 (88.8)	321 (95.0)
Ethnicity, No. (%)			.84b
Hispanic or Latino	6 (1.2)	5 (1.5)
Not Hispanic or Latino	507 (97.7)	328 (97.0)
Unknown	6 (1.6)	5 (1.5)
Education, No. (%)			<.001b
<High school	11 (2.1)	0 (0.0)
High school or GED	118 (22.7)	40 (11.8)
≥Some college	390 (75.1)	298 (88.2)
Marital status, No. (%)			.23b
Single	41 (7.9)	30 (8.9)
Widowed	28 (5.4)	15 (4.4)
Divorced and separated	77 (14.8)	35 (10.4)
Married and long-term relationship	373 (71.9)	258 (76.3)
Menopausal status, No. (%)			.05b
Premenopausal	161 (31.0)	94 (27.8)
Perimenopausal	38 (7.3)	42 (12.4)
Postmenopausal	277 (53.4)	168 (49.7)
Medically induced	43 (8.3)	34 (10.1)
Body mass index, kg/m2			.03a
Mean (SD)	30.2 (7.4)	29 (6.9)
WRAT-4 reading			<.001a
Mean (SD)	62.6 (6.2)	64.0 (4.3)
Anxiety (STAI)			<.001a
Mean (SD)	35.9 (12.2)	28.2 (9.2)
Depression item (MFSI21)			< .001a
Mean (SD)	0.66 (0.90)	0.38 (0.76)
Cancer stage, No. (%)
I	139 (26.8)	N/A	N/A
II	259 (49.9)	N/A
III	98 (18.9)	NA
Unknown	23 (4.4)	NA

P values are from 2-sided t tests. GED = General Education Development; MFSI21 = Multidimensional Fatigue Inventory item 21; N/A = not applicable; STAI = Spielberger Trait Anxiety Inventory; WRAT-4 = Wide Range Assessment Test-Fourth Edition.

P values are from Pearson χ² tests (2-sided).

Cytokine and Receptor Levels

Cytokine and receptor levels were neither associated with the latency from the prechemotherapy blood draw to the first chemotherapy administration nor the latency from the last chemotherapy administration to the postchemotherapy blood draw (Supplementary Table 1, available online). NCORP site was added as a factor to all models in sensitivity analyses and was statistically significant only for IL-8. Inclusion of NCORP site did not impact the interpretation of any results.

Assessment 1 (Prechemotherapy or Time-Equivalent). At A1, mean [SE] sTNFRI (0.35 [0.02] vs 0.32 [0.02] ng/ml, P = .002) and sTNFRII (1.59 [0.07] vs 1.51 [0.07] ng/ml, P = .03) were higher in patients than controls (Figure 1 and Table 2). TNFα was lower in patients than controls (2.00 [0.14] vs 2.27 [0.15] log(pg/ml), P < .001).

Figure 1.

Adjusted cytokine and receptor values at prechemotherapy and postchemotherapy for patients with breast cancer (and the equivalent times for noncancer control participants). Error bars indicate standard error. Means, 95% confidence intervals, P values, and number of participants are reported in Table 2. Cytokine and receptor values are adjusted for age (continuous, fixed effect), race (binary, fixed effect), education (binary, fixed effect), body mass index (continuous, fixed effect), and plate (categorical, random effect). IL = interleukin; sTNFRI and sTNFRII = soluble tumor necrosis factor receptor I and II; TNFα = tumor necrosis factor alpha.

Table 2.

Adjusted cytokine and receptor values at prechemotherapy, postchemotherapy, and change from prechemotherapy to postchemotherapy (and the equivalent times for controls)^a

Cytokine/receptor and group	Assessment 1 (A1, prechemotherapy or time-equivalent for controls)			Assessment 2 (A2, postchemotherapy or time-equivalent for controls)			Pre- to postchemotherapy (or time-equivalent) change (A2-A1)
	No.	Mean (95% CI)	Between groups Pb	No.	Mean (95% CI)	Between groups Pb	No.	Mean change (95% CI)	Within groups Pc	Between groups Pd
IL-4, log (pg/ml)
Breast cancer	515	4.08 (3.77 to 4.38)	.59	494	3.34 (3.01 to 3.66)	<.001	490	−0.71 (−0.82 to −0.59)	<.001	<.001
Control	333	4.12 (3.81 to 4.44)		333	4.11 (3.77 to 4.45)		328	−0.02 (−0.15 to 0.12)	.80
IL-6, log (pg/ml)
Breast cancer	515	0.34 (0.01 to 0.67)	.68	494	0.10 (−0.24 to 0.44)	.12	490	−0.23 (−0.33 to −0.14)	<.001	.02
Control	333	0.31 (−0.03 to 0.65)		333	0.25 (−0.10 to 0.60)		328	−0.05 (−0.17 to 0.07)	.38
IL-8, log (pg/ml)
Breast cancer	515	1.38 (1.00 to 1.76)	.44	494	1.34 (0.96 to 1.72)	.52	490	−0.04 (−0.17 to 0.10)	.57	.97
Control	333	1.30 (0.90 to 1.70)		333	1.27 (0.88 to 1.67)		328	−0.04 (−0.19 to 0.12)	.65
IL-10, log (pg/ml)
Breast cancer	515	2.85 (2.47 to 3.23)	.34	494	2.11 (1.72 to 2.50)	<.001	490	−0.72 (−0.83 to −0.62)	<.001	<.001
Control	333	2.95 (2.56 to 3.35)		333	2.91 (2.51 to 3.32)		328	−0.06 (−0.19 to 0.06)	.34
TNFα, log (pg/ml)
Breast cancer	515	2.00 (1.72 to 2.28)	<.001	494	2.02 (1.75 to 2.29)	<.001	490	0.03 (−0.02 to 0.07)	.27	.15
Control	333	2.27 (1.99 to 2.56)		333	2.24 (1.97 to 2.51)		328	−0.03 (−0.08 to 0.03)	.32
sTNFRI, ng/ml
Breast cancer	507	0.35 (0.31 to 0.38)	.002	493	0.45 (0.41 to 0.49)	<.001	481	0.10 (0.08 to 0.11)	<.001	<.001
Control	332	0.32 (0.29 to 0.36)		334	0.33 (0.29 to 0.37)		328	0.01 (0.00 to 0.02)	.20
sTNFRII, ng/ml
Breast cancer	507	1.59 (1.46 to 1.73)	.03	493	2.23 (2.07 to 2.40)	<.001	481	0.62 (0.56 to 0.68)	<.001	<.001
Control	332	1.51 (1.37 to 1.65)		334	1.56 (1.39 to 1.73)		328	0.04 (−0.03 to 0.11)	.29

^a Adjusted means and confidence intervals for cytokine and receptor levels are estimated from the linear regression models, adjusting for age (continuous, fixed effect), race (binary, fixed effect), education (binary, fixed effect), body mass index (continuous, fixed), and plate (categorical, random effect). CI = confidence interval; IL = interleukin; sTNFRI and sTNFRII = soluble tumor necrosis factor receptor I and II; TNFα = tumor necrosis factor alpha.

^b Between-group P values are from 2-sided t tests cross-sectionally assessing patients vs controls at A1 and A2.

^c Within-group P values are from 2-sided t tests separately assessing A1 to A2 changes in cytokines and receptors levels in patients and in controls.

^d Between-group P values are from 2-sided t tests assessing the differential A1 to A2 change in cytokines and receptors levels between patients and controls.

Assessment 2 (Postchemotherapy or Time-Equivalent). At A2, mean [SE] sTNFRI (0.45 [0.02] vs 0.33 [0.02] ng/ml, P < .001) and sTNFRII (2.23 [0.08] vs 1.56 [0.09] ng/ml, P < .001) were higher in patients than controls (Figure 1 and Table 2), and IL-4 (3.34 [0.17] vs 4.11 [0.17] log(pg/ml), P < .001) and IL-10 (2.11 [0.20] vs 2.91 [0.21] log(pg/ml), P < .001) were lower. TNFα remained lower in patients vs controls (2.02 [0.14] vs 2.24 [0.14] log(pg/ml), P < .001).

Change From Assessment 1 to Assessment 2. There were statistically significant between-group differences (patients vs controls) in the change in cytokine and receptor levels from A1 to A2 for IL-4, IL-6, IL-10, sTNFRI, and sTNFRII (Figure 1 and Table 2). Within patients, IL-4, IL-6, and IL-10 decreased from A1 to A2 (mean [SE]: change in IL-4 = −0.71 [0.06] log(pg/ml), P < .001; change in IL-6 = −0.23 [0.05] log(pg/ml), P < .001; change in IL-10 = −0.72 [0.05] log(pg/ml), P < .001), whereas sTNFRI and sTNFRII increased (change in sTNFRI = 0.10 [0.01] ng/ml, P < .001; change in sTNRII = 0.62 [0.03] ng/ml, P < .001). Within controls, there were no changes from A1 to A2 in any of the inflammatory cytokines and receptors measured.

Cognitive Test Performance

Results of the cognitive tests for the parent cohort were reported previously (1). Results for the present smaller cohort show the same pattern as the parent cohort and are summarized in Supplementary Table 3 and Supplementary Figure 1 (available online). Because cytokines and receptors did not change over time in the noncancer control participants and we were specifically interested in the relationship between inflammation and cognitive function in the context of cancer and chemotherapy, the analyses evaluating relationships between inflammation and cognitive test scores were conducted on patient data only.

Relationships Between Inflammation and Test Performance in Patients

Rapid Visual Processing. At A1, there were no statistically significant relationships between cytokine and receptor levels and RVP score. At A2, higher sTNFRII was associated with worse RVP score (β = −1.316, SE = 0.587, P = .03; Table 3). In longitudinal analyses, higher A1 IL-4 was associated with improvement in RVP performance over time from A1 to A2 (β = 0.820, SE = 0.336, P = .02). There were no statistically significant relationships between changes in cytokines and receptors and changes in RVP score from A1 to A2.

Table 3.

Association of cytokines and receptors with attention and processing speed tests in patients with breast cancer^a

Cognitive test and cytokine/receptor	Prechemo (A1) inflammatory factor and prechemo (A1) test score		Postchemo (A2) inflammatory factor and postchemo (A2) test score		Prechemo (A1) inflammatory factor and change in test score (A2-A1)b		Change in inflammatory factor (A2-A1) and change in test score (A2-A1)b
	β (SE)	P c	β (SE)	P c	β (SE)	P c	β (SE)	P c
RVPd
IL-4	0.674 (0.346)	.05	0.362 (0.319)	.26	0.819 (0.336)	.02	−0.413 (0.360)	.25
IL-6	0.117 (0.335)	.73	0.106 (0.386)	.78	0.330 (0.322)	.31	−0.285 (0.364)	.44
IL-8	−0.327 (0.267)	.22	−0.188 (0.328)	.57	−0.071 (0.259)	.78	0.177 (0.353)	.62
IL-10	0.247 (0.278)	.37	0.065 (0.291)	.82	0.366 (0.268)	.17	−0.391 (0.338)	.25
TNFα	0.667 (0.551)	.23	0.539 (0.677)	.43	0.244 (0.532)	.65	−0.411 (0.849)	.63
sTNFRI	−2.684 (3.334)	.42	−3.132 (2.885)	.28	1.591 (3.227)	.62	−3.460 (3.161)	.27
sTNFRII	−1.414 (0.760)	.06	−1.316 (0.587)	.03	−0.319 (0.752)	.67	−0.615 (0.591)	.30
BCTe
IL-4	0.254 (0.317)	.42	0.049 (0.249)	.84	0.437 (0.241)	.07	−0.564 (0.253)	.03
IL-6	0.226 (0.312)	.47	0.426 (0.301)	.16	0.047 (0.231)	.84	0.228 (0.259)	.38
IL-8	0.610 (0.241)	.01	0.841 (0.260)	.001	0.272 (0.186)	.15	−0.003 (0.278)	.99
IL-10	0.122 (0.255)	.63	−0.050 (0.228)	.83	0.028 (0.193)	.88	−0.134 (0.245)	.58
TNFα	0.682 (0.501)	.18	0.833 (0.533)	.12	0.742 (0.377)	.05	−0.431 (0.578)	.46
sTNFRI	3.189 (3.027)	.29	6.638 (2.208)	.003	5.566 (2.367)	.02	1.880 (2.256)	.41
sTNFRII	0.325 (0.687)	.64	0.913 (0.455)	.045	0.425 (0.548)	.44	0.528 (0.424)	.21
TMT-Ae
IL-4	−1.098 (0.516)	.03	−0.167 (0.346)	.63	−0.561 (0.388)	.15	0.595 (0.415)	.15
IL-6	−0.971 (0.499)	.05	0.175 (0.421)	.68	0.090 (0.374)	.81	−0.080 (0.425)	.85
IL-8	0.066 (0.398)	.87	0.104 (0.359)	.77	−0.388 (0.295)	.18	0.580 (0.408)	.16
IL-10	−0.835 (0.414)	.04	0.235 (0.315)	.46	0.123 (0.310)	.69	0.725 (0.392)	.07
TNFα	−0.695 (0.823)	.40	0.124 (0.734)	.87	−0.101 (0.611)	.87	0.278 (0.985)	.78
sTNFRI	6.253 (5.036)	.22	−1.330 (3.124)	.67	−4.039 (3.736)	.28	3.063 (3.650)	.40
sTNFRII	1.922 (1.151)	.10	−0.103 (0.641)	.87	−0.410 (0.862)	.64	0.374 (0.686)	.59

^a Linear regression models are adjusted for age, body mass index, duration of chemotherapy, cognitive reserve, anxiety, and a depression item score (all continuous). A1 = assessment 1; A2 = assessment 2; β = beta parameter estimate; BCT = backward counting task; IL = interleukin; post chemo = postchemotherapy; pre chemo = prechemotherapy; RVP = Rapid Visual Processing Task; sTNFRI and sTNFRII = soluble tumor necrosis factor receptor I and II; TMT-A = Trail Making Test part A; TNFα = tumor necrosis factor alpha.

^b Indicates models are also adjusted for baseline (A1) test performance.

^c P values are from 2-sided t tests of linear regression coefficients.

^d Note that a higher score is better for RVP.

^e Note that a lower score is better for BCT and TMT-A.

Backward Counting. At A1, higher IL-8 was associated with higher (worse) BCT score (β = 0.610, SE = 0.241, P = .01; Table 3). At A2, higher IL-8 (β = 0.841, SE = 0.260, P = .001), sTNFRI (β = 6.638, SE = 2.208, P = .003), and sTNFRII (β = 0.913, SE = 0.455, P = .045) were associated with worse scores. In longitudinal analyses, higher sTNFRI at A1 was associated with worse performance in BCT over time (β = 5.566, SE = 2.367, P = .02; Table 3). Greater increase in IL-4 from A1 to A2 was associated with better performance on BCT over the same time period (β = −0.564, SE = 0.253, P = .03).

Trail Making Test (Part A). Higher IL-4 was associated with lower (better) TMT-A score at A1 (β = −1.098, SE = 0.516, P = .03; Table 3). Higher IL-10 was also associated with better TMT-A score at A1 (β = −0.835, SE = 0.414, P = .04).

Discussion

To our knowledge, this is the largest study to date evaluating relationships between inflammation and attention and processing speed in patients with breast cancer at prechemotherapy and postchemotherapy. This study also provides information on the levels of inflammatory and anti-inflammatory cytokines and receptors before and after chemotherapy in patients vs control participants without cancer. Inflammation was greater in patients than noncancer control participants even before chemotherapy, evidenced by higher levels of pro-inflammatory sTNFRI and sTNFRII. Chemotherapy exacerbated the inflammatory phenotype: sTNFRI and sTNFRII increased in patients from prechemotherapy to postchemotherapy, whereas anti-inflammatory IL-4 and IL-10 decreased. With some exceptions, inflammatory cytokines and receptors were associated with worse attention and processing speed performance and anti-inflammatory cytokines were associated with better performance. However, these associations are considered exploratory and future confirmatory studies are needed.

TNF receptor levels were statistically significantly higher in patients than noncancer control participants before chemotherapy and rose during chemotherapy, consistent with cross-sectional analyses demonstrating higher sTNFRI and sTNFRII in patients with various cancers relative to noncancer controls and higher sTNFRII in patients postchemotherapy compared with chemotherapy-naïve patients (31,65,66). Both TNF receptors were related to worse test performance in multiple analyses. In cross-sectional analyses, higher prechemotherapy sTNFRII was associated (non-statistically significantly; Table 3) with worse RVP performance; at postchemotherapy, higher sTNFRI and sTNFRII were statistically significantly associated with worse BCT score and RVP (sTNFRII only). In longitudinal analyses, higher prechemotherapy sTNFRI predicted statistically significantly worse BCT performance over time. Previous studies demonstrate relationships between higher sTNFRI and sTNFRII and worse cognitive test performance, frailty, and fatigue in breast cancer patients after chemotherapy (2,30,31,66,67). Together, these data suggest that TNF activity may have multiple behavioral effects related to cancer and chemotherapy.

Anti-inflammatory cytokines such as IL-4 and IL-10 can skew central nervous system immune cells toward a protective phenotype, delay neurodegeneration (68), and may protect cognitive function in patients with cancer or other disorders (24,27,69–72). To our knowledge, this is 1 of only 2 studies to evaluate anti-inflammatory cytokines and cognitive outcomes in breast cancer patients before chemotherapy (24). In the present study, higher IL-4 and IL-10 were statistically significantly associated with better TMT-A scores, and IL-4 was non-statistically significantly associated with better RVP score at prechemotherapy. In longitudinal analyses, higher prechemotherapy IL-4 predicted statistically significantly greater improvement in RVP over time. An increase in IL-4 from prechemotherapy to postchemotherapy was statistically significantly associated with improvement in BCT score from prechemotherapy to postchemotherapy. Although in general IL-4 and IL-10 seemed to have a protective relationship with attention and processing speed performance, more research is needed evaluating these relationships in longitudinal models and understanding relationships between anti-inflammatory factors and specific domains of cognition.

Higher levels of pro-inflammatory IL-8 were associated with worse BCT score cross-sectionally at prechemotherapy and postchemotherapy. Pro-inflammatory TNFα at prechemotherapy predicted (non-statistically significantly) worsened BCT performance over time. Interestingly, patients had statistically significantly lower TNFα than controls at prechemotherapy and postchemotherapy. Previous studies have reported higher serum TNFα in patients with breast cancer (32,73–75), no group differences (76,77), or higher TNFα only in patients with late stages of disease (78). Interpretation of these studies is limited by small sample sizes, nonmatched controls, retrospective design, inclusion of nonchemotherapy-naïve patients, and/or high percentages of values below the limit of detection for TNFα. The present study used a larger sample size, age- and sex-matched controls, a prospective design, patients that were chemotherapy-naïve at baseline, and high-sensitivity TNFα assays. Also, our results may in part be explained by patients having high levels of sTNFRI and sTNFRII, which can sequester TNFα, reducing free TNFα in serum (65,79). Further study is needed to understand how these inflammatory factors impact cognition in the context of breast cancer and chemotherapy.

Surprisingly, pro-inflammatory IL-6 decreased statistically significantly over time in patients relative to controls, and lower IL-6 was associated with worse TMT-A performance (non-statistically significantly) in patients at prechemotherapy. Some studies have reported increases in IL-6 during chemotherapy (24,27,28,32). However, IL-6 levels can rise and fall during chemotherapy and differ based on regimen type and treatment response (24,28,76). IL-6 can also enact pro-inflammatory or anti-inflammatory responses depending on its modality of receptor binding (80). Additional studies following patients with different clinical characteristics over multiple time points are needed to clarify the trajectory of IL-6 levels and its relationship to cognitive function.

Strengths of this study include a large sample size, the use of objective cognitive tests, prospective longitudinal design, and a novel investigation of prechemotherapy inflammation as a predictor of cognitive test performance change from prechemotherapy to postchemotherapy. The use of multiple community clinic sites across the United States improves representation and increases generalizability.

There are a few limitations to this study. A validated multidimensional depression measure (eg, Center for Epidemiologic Studies Depression Scale) was not used, so scoring was limited to a single question that does not capture nuanced aspects of depression. Data on smoking and comorbidities were not included for this analysis. Because depression, smoking, and comorbidities may relate positively to inflammation and cognitive dysfunction, residual confounding may be biasing the results away from the null. Racial diversity was limited, although the sample was geographically diverse. Patients had heterogeneous chemotherapy regimens, but assigning risk of inflammation or cognitive difficulties to particular variations of treatment would require analyses outside the scope of this analysis. Lastly, given the discovery nature of this research, we opted to report P values without multiplicity adjustments to lower the chance of false negatives (81–83). Thus, these findings should be considered preliminary and be validated by future studies.

RVP has been used to evaluate patients with a variety of medical conditions and distinguishes cognitively healthy adults from those with mild cognitive impairment (45). However, clinically meaningful changes in RVP score have not been established. Therefore, RVP was used in this study specifically to investigate mechanistic links between inflammation and attention and processing speed. Additionally, although numerous studies evaluate cytokine and receptor levels in various patient populations relative to healthy controls, a consensus on clinically meaningful changes in cytokine and receptor levels does not exist (84). Thus, we focused our interpretation on the potential mechanistic links between inflammation and CRCD, highlighting biological significance more than the clinical significance of the observed changes in cytokines and receptors and cognitive test scores. Future studies will evaluate the relationship between inflammation and patient self-reported cognitive outcomes and other domains of cognitive function. Lastly, because of the relative lack of change in the cytokines and receptors and cognitive outcomes in noncancer control participants in this study, we chose to only analyze relationships between these variables in patients in the present analyses. It would be interesting to further explore these relationships in studies assessing patients relative to other comparator groups.

In conclusion, the present data suggest that patients with breast cancer have baseline inflammation that is exacerbated by chemotherapy and that pro-inflammatory factors may be detrimental and anti-inflammatory factors beneficial to cognitive function, specifically attention and processing speed. However, because of the exploratory nature of this work, further research is needed. Ultimately, better understanding of the relationships between inflammatory markers and cognition may improve our ability to predict a patient’s risk for developing cognitive difficulties before and after chemotherapy.

Funding

This work was supported by the National Cancer Institute grant numbers R01CA231014, DP2195765, K07CA1688, K07CA221931, T32CA102618, U10CA037420, UG1CA189961, and the University of Rochester Clinical and Translational Science Award number KL2TR001999.

Notes

Role of the funders: The funders had no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication.

Disclosures: There are no conflicts of interest to report for any author.

Author contributions: Authors contributed to the conceptualization (EB, EC, MJ), data curation (EC, LL, CH, JG, JH, SC), formal analysis (EB, EC, LL, CH, MS, MJ), funding acquisition (MJ), investigation and methodology (EB, NG, SH, AK, IK, EC, BT, LL, ZW, JG, JH LL, SC CH, MS, MJ), project administration and resources (MJ), supervision (MJ), validation (EB, EC, LL, CH, MJ), visualization (EB), writing—original draft (EB), writing—review and editing (all authors) of this work.

Acknowledgements: We thank the participants in this study and all staff at the University of Rochester Cancer Control NCORP Base and our NCORP affiliates who recruited and followed participants. We thank the National Cancer Institute NCORP programs for their funding and support of this project. NCORP was formerly known as the Community Clinical Oncology Program (CCOP) when this study started in 2011. The following CCOP/NCORPs participated in this study: Heartland (formerly Central Illinois), Columbus, Cancer Research Consortium of West Michigan (formerly Grand Rapids and Kalamazoo), Dayton, Delaware, Greenville, Hematology-Oncology Associates of Central New York, Kansas City, Metro-Minnesota Community Oncology Research Consortium, Nevada, Northwell (formerly North Shore), Pacific Cancer Research Consortium (formerly Western Oncology Research Consortium), Southeast Clinical Oncology Research Consortium (formerly Southeast Cancer Control Consortium), Upstate Carolina, Virginia Mason, Wichita, and Wisconsin (formerly Marshfield). We also thank the staff of the Supportive Care in Cancer Research Program, the Cancer Control and Psychoneuroimmunology Laboratory, and the Human Biophysiology Shared Resource for their assistance with this project.

Prior presentations: This work was presented in part at the 2020 Biennial Meeting of the International Cognition and Cancer Task Force as an abstract and poster presentation.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.