Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Apr 24.
Published in final edited form as: Nurs Res. 2023 Apr 24;72(4):272–280. doi: 10.1097/NNR.0000000000000660

An Evaluation of the Multifactorial Model of Cancer-Related Cognitive Impairment

Kate R Oppegaard 1, Samantha J Mayo 2, Terri S Armstrong 3, Kord M Kober 1, Joaquin Anguera 4, Fay Wright 5, Jon D Levine 4, Yvette P Conley 6, Steven Paul 1, Bruce Cooper 1, Christine Miaskowski 1,4
PMCID: PMC10330009  NIHMSID: NIHMS1888086  PMID: 37104681

Abstract

Background:

Up to 45% of patients report cancer-related cognitive impairment (CRCI). A variety of characteristics are associated with the occurrence and/or severity of CRCI. However, an important gap in knowledge of risk factors for CRCI is the relative contribution of each factor. The Multifactorial Model of Cancer-Related Cognitive Impairment (MMCRCI) is a conceptual model of CRCI that can be used to evaluate the strength of relationships between various factors and CRCI.

Objectives:

The purpose of this study was to use structural regression methods to evaluate the MMCRCI using data from a large sample of outpatients receiving chemotherapy (n = 1,343). Specifically, the relationships between self-reported CRCI and four MMCRCI concepts (i.e., social determinants of health, patient-specific factors, treatment factors, and co-occurring symptoms) were examined. The goals were to determine how well the four concepts predicted CRCI and determine the relative contribution of each concept to deficits in perceived cognitive function.

Methods:

This study is part of a larger, longitudinal study that evaluated the symptom experience of oncology outpatients receiving chemotherapy. Adult patients were diagnosed with breast, gastrointestinal, gynecological, or lung cancer; had received chemotherapy within the preceding 4 weeks; were scheduled to receive at least two additional cycles of chemotherapy; were able to read, write, and understand English; and gave written informed consent. Self-reported CRCI was assessed using the Attentional Function Index. Available study data were used to define the latent variables.

Results:

On average, patients were 57 years of age, college educated, with a mean Karnofsky Performance Status score of 80. Of the four concepts evaluated, while co-occurring symptoms explained the largest amount of variance in CRCI, treatment factors explained the smallest amount of variance. A simultaneous structural regression model that estimated the joint effect of the four exogenous latent variables on the CRCI latent variable was not significant.

Discussion:

These findings suggest that testing individual components of the MMCRCI may provide useful information on the relationships among various risk factors, as well as refinements of the model. In terms of risk factors for CRCI, co-occurring symptoms may be more significant than treatment factors, patient-specific factors, and/or social determinants of health in patients receiving chemotherapy.

Keywords: cancer, chemotherapy, cognitive impairment, conceptual model, depression, fatigue, patient-reported outcomes, sleep disturbance, social determinants of health, structural equation model

Cancer-related cognitive impairment (CRCI) is reported by up to 45% of patients (Schmidt et al., 2016). While not fully understood, the causes of CRCI are thought to be multifactorial (e.g., treatment-related effects, patient-specific characteristics; Bai & Yu, 2021). A variety of cognitive domains are impacted (e.g., memory, attention, processing speed; Ren et al., 2019). Consequently, CRCI negatively impacts the everyday lives of those who experience it (Mayo et al., 2022). Prevention and mitigation strategies for CRCI remain limited (Bai & Yu, 2021), likely due to the lack of understanding of its underlying mechanism(s) and comprehensive evaluation of associated risk factors.

Treatment factors (e.g., hormonal changes, direct effects of chemotherapy) and co-occurring symptoms (e.g., anxiety, depression, fatigue) are among the most frequently identified risk factors for CRCI (Bai & Yu, 2021). In addition, various demographic and clinical characteristics are associated with the occurrence of CRCI (Janelsins et al., 2017). However, an important gap in our knowledge of the various risk factors is the relative contribution of each risk factor to CRCI. In other words, which risk factor significantly contributes to its occurrence, severity, and/or persistence? This knowledge is needed to begin prioritizing modifiable factors amenable to interventions.

One analytic approach that can be used to explore the strength of the relationships between/among variables is structural regression modeling (i.e., a type of structural equation modeling). Structural regression methods were developed to evaluate complex interrelationship patterns among variables (Maruyama, 1997). Therefore, these methods can be used to estimate the strength of the relationships between variables in a conceptual model (Maruyama, 1997). Using this analytical approach, indicator variables are selected to create an exogenous latent variable representing an otherwise unobserved independent variable (i.e., a hypothetical construct). For example, in the current study, the indicator variables of annual household income, years of education, cumulative lifetime stress, and resilience are used to create an exogenous latent variable that represented “social determinants of health.” In contrast, endogenous latent variables represent the dependent or outcome variable (e.g., self-reported CRCI). While structural regression methods were used to evaluate a number of outcomes in patients with cancer (e.g., post-traumatic growth; Zhang et al., 2021) and quality of life (Lee & Jeong, 2019), this analytic approach has not been used to evaluate risk factors for CRCI.

The Multifactorial Model of Cancer-Related Cognitive Impairment (MMCRCI) is a comprehensive conceptual model of CRCI that includes factors with known or hypothesized associations with CRCI (Oppegaard et al., 2023). Within the MMCRCI, these factors are organized into broader concepts: social determinants of health, patient-specific factors, treatment factors, co-occurring symptoms, and biological mechanisms. While the MMCRCI is based on an extensive review of the literature, it requires testing. Therefore, the purpose of this study was to use structural regression methods to evaluate the MMCRCI using data from a large sample of oncology outpatients receiving chemotherapy. Specifically, the relationships between CRCI and four of the MMCRCI concepts (i.e., social determinants of health, patient-specific factors, treatment factors, co-occurring symptoms) were examined; the joint effect of the four concepts on CRCI was evaluated; and the unique contribution of co-occurring symptoms on CRCI controlling statistically for the contributions of each of the other three concepts were determined. The overall goal was to verify how well the concepts in the MMCRCI predicted CRCI and determine the relative contribution of each concept to deficits in cognitive function.

Methods

Study Sample and Procedures

This study is part of a larger, longitudinal study that evaluated the symptom experience of oncology outpatients receiving chemotherapy (Miaskowski et al., 2014). In brief, eligible patients were ≥ 18 years of age, had a diagnosis of breast, gastrointestinal, gynecological, or lung cancer; had received chemotherapy within the preceding four weeks, were scheduled to receive at least two additional cycles of chemotherapy, were able to read, write, and understand English, and gave written informed consent. Patients were recruited from two comprehensive cancer centers, one Veteran’s Affairs hospital, and four community-based oncology programs. A total of 2,234 patients were approached, and 1,343 consented to participate (60.1% response rate). The major reason for refusal was being overwhelmed with their cancer treatment. Data from the enrollment assessment (i.e., prior to receipt of second or third cycle of chemotherapy) were used in this analysis.

Conceptual Model

The structural regression models (SRMs) evaluated in this study are based on the MMCRCI (Oppegaard et al., 2023). Available study data were used to define observed indicators as latent variables (Kline, 2016) that mapped to each of the concepts in the MMCRCI (Figure 1).

Figure 1.

Figure 1

Open in a new tab

The Hypothetical Model to be Evaluated Based on the Multifactorial Model of Cancer-Related Cognitive Impairment

Variables

Demographic questionnaires obtained information on age, gender, ethnicity, education, employment status, and income. Medical records were reviewed for disease and treatment information.

Outcome Variable

Self-reported CRCI was assessed using the Attentional Function Index (AFI; Cimprich et al., 2011), a 16-item instrument designed to assess an individual’s perceived effectiveness in performing daily activities that are supported by attention, working memory, and executive functions (e.g., setting goals, planning, and carrying out tasks). A higher total mean score on a 0 to 10 numeric rating scale indicates greater capacity to direct attention (Cimprich et al., 2011). Clinically useful cut points for attentional function are: < 5.0 low function, 5.0 to 7.5 moderate function, and > 7.5 high function (Cimprich et al., 2005). Cronbach’s alpha for the total AFI score was 0.93.

Latent Variables

Estimation of the endogenous latent CRCI variable was carried out with a measurement model that used the individual AFI items as indicators of the latent score. When that measurement model was examined, because numerous correlated residuals were found among the items, the fit of the measurement model to the data was very poor. Therefore, the latent CRCI score was estimated following the recommendations of Jøreskog and Sørbom (as reported in (Raykov & Marcoulides, 2006) by estimating the measurement error and residual variance as (1 – reliability)* AFI computed variance. This value was defined as the CRCI “latent variable” residual variance for the subsequent SRMs.

Measurement models for each exogenous latent variable were created using MMCRCI concepts (i.e., social determinants of health, patient-specific factors, treatment factors, co-occurring symptoms). The indicator variables for each exogenous latent variable were selected from available study data. Specific information about each exogenous latent variable is described below (see Figure 1).

Social Determinants of Health.

Indicator variables used for this exogenous latent variable included: annual household income, years of education, cumulative lifetime stress, and resilience. Cumulative lifetime stress was assessed using the Life Stressor Checklist–Revised (LSC–R), an index of lifetime trauma exposure (e.g., being mugged, death of a loved one, sexual assault; Wolfe & Kimerling, 1997). Resilience was assessed using the Connor–Davidson Resilience Scale (CD–RISC), an instrument that evaluates a patient’s personal ability to handle adversity (Connor & Davidson, 2003).

Patient-Specific Factors.

Indicator variables for this exogenous latent variable included: age, functional status, comorbidity burden, the personality domain of neuroticism, global perceived stress, and cancer-specific stress. Functional status was assessed using the Karnofsky Performance Status (KPS) scale (Karnofsky et al., 1948). Comorbidity burden was assessed using the Self-Administered Comorbidity Questionnaire (SCQ) score (Sangha et al., 2003). The personality domain of neuroticism was assessed using the NEO–Five Factor Inventory (NEO–FFI; Costa & McCrae, 1992). Global perceived stress was assessed using the Perceived Stress Scale (PSS; Cohen et al., 1983), a measure of global perceived stress according to the degree that life circumstances are appraised as stressful over the course of the previous week. Cancer-specific stress was assessed using the Impact of Event Scale-Revised (IES–R; Weiss & Marmar, 1997).

Treatment Factors.

Indicator variables for this exogenous latent variable included: hemoglobin level, white blood cell count, toxicity of chemotherapy regimen, antiemetic regimen (i.e., number and type[s] of antiemetic medications), and chemotherapy cycle length. The toxicity of the chemotherapy regimen was determined using the MAX2 index (Extermann et al., 2004). Briefly, the MAX2 score is the average of the most frequent grade four hematologic toxicity and the most frequent grade three to four nonhematologic toxicity reported in publications of a regimen; it correlates well with that regimen’s average overall risk of severe toxicity.

Co-Occurring Symptoms.

Indicator variables for this exogenous latent variable included: severity of morning and evening fatigue, state anxiety, sleep disturbance, depressive symptoms, and severity of worst pain. The Lee Fatigue Scale (LFS; Lee et al., 1991) assessed morning and evening fatigue. State anxiety was assessed using the Spielberger State Anxiety Inventory (STAI–S; Spielberger et al., 1983). Sleep disturbance was assessed using the General Sleep Disturbance Scale (GSDS; Lee, 1992). Depressive symptoms were assessed using the Center for Epidemiological Studies Depression Scale (CES–D; Radloff, 1977). Severity of worst pain was assessed using the Brief Pain Inventory (Daut et al., 1983).

Statistical Analysis

Statistical analyses were performed with Stata version 16.1 (StataCorp. 2019. College Station, TX: StataCorp LLC.). Means, standard deviations, and percentages were calculated for demographic and clinical characteristics. All variables were assessed for appropriateness for inclusion in the SRMs. Indicator variables that were included in the models were either continuous or ordinal. Given the large sample size, normality of the parameter estimates was assumed based on the central limit theorem (Kwak & Kim, 2017). Missing data were accommodated with the use of full information maximum likelihood (FIML) and the expectation–maximization (EM) algorithm (Muthén & Shedden, 1999). In Stata, this method of estimation is called maximum likelihood with missing values (mlmv).

Most of the models employed the usual Newton–Raphson (NR) algorithm for likelihood estimation. However, convergence with NR failed for the most complex models when mlmv was used. For these models, the Berndt–Hall–Hall–Hausman (BHHH) algorithm was used for several (e.g., 10) iterations, then estimation switched to NR for several iterations, then back to BHHH until convergence was achieved. Model fit for each measurement model and SRMs were evaluated using recommended fit indices. Absolute fit was evaluated using the chi-square test of goodness of fit (Kline, 2016). Model parsimony was evaluated using the root mean square error of approximation (RMSEA; Browne & Cudeck, 1992). Comparative fit was evaluated using the comparative fit index (CFI; Bentler, 1990). A chi-square close to nonsignificant, RMSEA of < .06, and CFI of > .95 were used as the desirable cut points for these fit indices.

Modification indices were examined to improve model fit by incorporating correlated residuals into some measurement models for exogenous latent variables. Standardized parameter estimates for the measurement model coefficients were used to interpret the relative importance of indicators measured on incongruent scales. A two-sided p-value of < 0.05 was considered statistically significant for hypothesis tests on the unstandardized coefficients. SRMs were built in the following order: an individual measurement model for each exogenous latent variable was estimated with significant coefficients, an individual SRM for the CRCI latent variable was regressed on each exogenous latent variable, and a simultaneous SRM was evaluated that regressed the CRCI latent variable on the four exogenous latent variables jointly. Finally, four hierarchical SRM was built to estimate the unique contribution of co-occurring symptoms on the CRCI latent variable, controlling for either the effect of social determinants of health, patient-specific factors, or treatment factors.

Results

Sample Characteristics

Demographic, clinical, symptom, stress, and resilience characteristics of the 1,343 patients are summarized in Table 1. On average, patients were 57 years of age and college educated, with a mean KPS score of 80. The majority were female, White, receiving only chemotherapy, and receiving chemotherapy on a 21-day cycle. Patients in this study had an average AFI score of 6.4, which suggests a moderate level of CRCI.

Table 1.

Demographic, Clinical, Stress, Resilience, and Symptom Characteristics (n = 1,343)

Demographic and Clinical Characteristics Mean (SD)
Age (years) 57.2 (12.4)
Education (years) 16.2 (3.0)
Neuroticism personality domain 15.1 (8.0)
Karnofsky Performance Status score 80.0 (12.5)
Self-administered Comorbidity Questionnaire score 5.5 (3.2)
MAX2 score 0.17 (0.08)
Hemoglobin 11.5 (1.4)
White blood cell count 7.3 (4.1)
% (n)
Gender (% female) 77.8 (1044)
Self-reported ethnicity
   American Indian/Alaskan Native 0.46 (6)
   Asian 11.8 (155)
   Black or African American 7.5 (98)
   Native Hawaiian or Other Pacific Islander 1.0 (13)
   White 72.9 (956)
   Mixed Ethnic Background 5.3 (69)
   Other 1.1 (15)
Annual household income
   Less than $30,000+ 18.4 (221)
   $30,000 to $70,000 21.2 (254)
   $70,000 to $100,000 16.9 (203)
   Greater than $100,000 43.6 (523)
Cancer diagnosis
   Breast cancer 40.2 (540)
   Gastrointestinal cancer 30.7 (412)
   Gynecological cancer 17.4 (233)
   Lung cancer 11.8 (158)
CTX regimen
   Only chemotherapy 70.1 (922)
   Only targeted therapy 3.0 (39)
   Both chemotherapy and targeted therapy 26.9 (354)
   Cycle length
   14-day cycle 42.1 (558)
   21-day cycle 50.6 (671)
   28-day cycle 7.3 (97)
Antiemetic regimen
   Nones 7.1 (92)
   Steroid alone or serotonin receptor antagonist alone 20.5 (265)
   Serotonin receptor antagonist and steroid 47.7 (618)
   Neurokinin–1 receptor antagonist and two other antiemetics 24.8 (321)
Stress and Resilience Measuresa Mean (SD)
Perceived Stress Scale total score 18.5 (8.2)
Impact of Event Scale-Revised total score (> 24) 18.8 (13.1)
Life Stressor Checklist-Revised total score (range 0–30) 6.1 (3.9)
Connor–Davidson Resilience Scale total score (range 0–40) 30.1 (6.4)
Symptomsa
Depressive symptoms (> 16.0) 12.8 (9.7)
State anxiety (> 32.2) 33.9 (12.4)
Morning fatigue (> 3.2) 3.5 (2.9)
Evening fatigue (> 5.6) 5.9 (2.7)
Sleep disturbance (> 43.0) 52.5 (20.2)
Attentional function (< 5.0 = Low, 5 to 7.5 = Moderate, > 7.5 = High) 6.4 (1.8)
Worst pain intensity score (range 0–10) 6.1 (2.5)
Open in a new tab

Note. SD = standard deviation

a

Clinically meaningful cutoff scores or range of scores are in parentheses.

Measurement Models for Each Exogenous Latent Variable

Fit indices for the measurement models for each exogenous latent variables are listed in Table 2. All models’ fit indices met the established cut points (i.e., chi-square close to nonsignificant, RMSEA of < .06, and CFI of > .95). Details on each measurement model for the four exogenous latent variables are provided in Appendix A.

Table 2.

Fit Indices for the Measurement Models for Each Exogenous Latent Variable

Latent variable chi-square (df) p-value RMSA CFI
Social determinants of health 6.35 (2) .042 0.040 0.982
Patient-specific factors 21.78 (7) .003 0.040 0.992
Treatment factors 6.97 (3) 073 0.031 0.986
Co-occurring symptoms 22.61 (7) .002 0.041 0.994
Open in a new tab

Note. CFI = comparative fit index; df = degrees of freedom; RMSEA = root mean squared error of approximation

SRM for the CRCI Latent Variable Regressed on Each Exogenous Latent Variable

Results of the individual SRM for the CRCI latent variable regressed on each of the exogenous latent variables are listed in Table 3. As indicated by the standardized path coefficients, co-occurring symptoms (−0.762), patient-specific factors (−0.658), and social determinants of health (0.653) had the largest effects on the CRCI latent variable. In contrast, treatment factors (0.092) had the smallest effect. Details on each of the SRM for the CRCI latent variable regressed on each exogenous latent variable are provided in Appendix A.

Table 3.

Results of Individual Structural Regression Models for the Cancer-Related Cognitive Impairment Latent Variable Regressed on Each of the Exogenous Latent Variables

Exogenous Latent variable p-value Path coefficient Standardized path coefficient Model R2
Social determinants of health .001 0.863 0.653 0.427
Patient-specific factors < .001 −0.873 −0.658 0.433
Treatment factors .028 0.092 0.092 0.008
Co-occurring symptoms < .001 −1.177 −0.762 0.581
Open in a new tab

Simultaneous and Hierarchical SRM

A simultaneous SRM that estimated the joint effect of the four exogenous latent variables on the CRCI latent variable was not significant (data not shown). The results of each hierarchical SRM that estimated the effect of co-occurring symptoms on the CRCI latent variable, controlling for each exogenous latent variable, are listed in Table 4. For each SRM, pairwise comparisons were done that evaluated the unique contribution of co-occurring symptoms using the difference in R2 between a model for one of the other three exogenous variables alone, followed by a model with co-occurring symptoms added. The unique variance contributions of co-occurring symptoms on CRCI, after controlling for social determinants of health, patient-specific factors, or treatment factors, were 0.203, 0.144, and 0.574, respectively.

Table 4.

Hierarchical Structural Regression Models that Estimate Unique Contribution of Co-occurring Symptoms on the Cancer-Related Cognitive Impairment Latent Variable for Either Social Determinants of Health, Patient Specific Factors, or Treatment Factors

Exogenous Latent Variables Models Path coefficient Z-statistic p-value Model R2 Change in R2* 95% Confidence interval
Pairwise comparison model testing for social determinants of health
SDOHa Model 1 0.863 8.27 < .001 0.427 n/a 0.66, 1.07
SDOH
CoOccSymb 		Model 2 0.443
−0.951		 3.37
−9.11		 .001
< .001		 0.630 0.203 0.19, 0.70
−1.16, −0.75
Pairwise comparison model testing for patient-specific factors
PtSpecFxc Model 1 −0.873 −19.50 < .001 0.433 n/a −0.96, −0.79
PtSpecFx
CoOccSym		 Model 2 0.611
−1.729		 1.92
−4.92		 .055
< .001		 0.577 0.144 −0.01, 1.23
−2.42, −1.04
Pairwise comparison model testing for treatment factors
TxFxd Model 1 0.092 2.20 .028 0.008 n/a 0.01, 0.17
TxFx
CoOccSym		 Model 2 −0.059
−1.189		 −1.18
−18.23		 .237
< .001		 0.582 .574 −0.16, 0.04
−1.32, −1.06
Open in a new tab

Note. CFI = Comparative fit index; CoOccSym = co-occurring symptoms; PtSpecFx = patient specific factors; RMSEA = root mean squared of approximation; SDOH = social determinants of health; SRM = structural regression model; TxFx = treatment factors.

*

Change in R2 between SRM of latent variable and outcome variable versus SRM of latent variable, outcome variable, and co-occurring symptoms

a

Indicator variables for social determinants of health included: years of education, annual income, cumulative lifetime stress, resilience levels

b

Indicator variables for co-occurring symptoms included: morning and evening fatigue, state anxiety, depressive symptoms, sleep disturbance, occurrence of pain

c

Indicator variables for patient-specific factors included: age, functional status, comorbidity burden, the personality domain of neuroticism, perceived stress, cancer-specific stress

d

Indicator variables for treatment factors included: white blood cell count, hemoglobin level, a rating of the toxicity of the chemotherapy regimen, number and type(s) of medications in the antiemetic regimen, chemotherapy cycle length

Discussion

In a large sample of patients receiving chemotherapy, this study is the first to use structural regression methods to examine the relationships between self-reported CRCI and four of the concepts in the MMCRCI (i.e., social determinants of health, patient-specific factors, treatment factors, co-occurring symptoms). Specifically, CRCI was operationalized as perceived changes in the domains of attention, working memory, and executive functions as measured by the AFI. This evaluation included an examination of the joint effect of the four concepts on predicting CRCI. In addition, in three separate SRMs, the unique contribution of co-occurring symptoms on CRCI was estimated after controlling for each of the other concepts.

A strength of this study is the evaluation of the unobservable influence of the broader MMCRCI concepts on CRCI through the creation of exogenous latent variables. Good model fit was achieved for each of the measurement models that represented the MMCRCI concepts (i.e., the exogenous latent variables; Table 2). As noted in one review (Ahles & Hurria, 2018), specific groups of risk factors, rather than individual risk factors, may increase patients’ risk for CRCI. Our results support this hypothesis and provide initial information on groups of risk factors that warrant further evaluation.

Each exogenous latent variable was significantly associated with the CRCI latent variable. These findings suggest that these four MMCRCI concepts are valid predictors of CRCI and support the multifactorial nature of CRCI. The majority of the indicator variables selected for each exogenous latent variable were supported by available evidence (Oppegaard et al., 2023). However, some of the indicator variables are relatively novel. For example, some personality domains (e.g., neuroticism, openness) are associated with an increased risk for other types of cognitive impairment (Curtis et al., 2015). However, in the only study of patients with cancer (Hermelink et al., 2010), trait negative affectivity was associated with decrements in self-reported cognition and attention. The specific domain of neuroticism from the NEO–FFI was selected as one of the indicator variables in the patient-specific latent variable because of its association with CRCI in our sample. However, other personality domains warrant evaluation in future studies.

In terms of other novel indicator variables, cumulative life stress and resilience were included as part of the social determinants of health latent variable. Cumulative life stress was included because it is associated with other social determinants of health (e.g., lower income, discrimination; Mikhail et al., 2018). In terms of resilience, as noted in one review (Lopez et al., 2021), individuals vary considerably in their ability to adapt to various life stressors, as well as in the development of resilience. Therefore, resilience is an important factor to consider as part of a more comprehensive evaluation of cumulative life stress and other social determinants of health. It is worth noting that resilience was included in the patient-specific factors concept in the original MMCRCI. However, the authors describe potential overlap among the model concepts, which supports the inclusion and evaluation of resilience as part of the social determinants of health latent variable.

Annual income and years of education were the other indicator variables included in the social determinants of health latent variable. In a study that evaluated associations between formal education, income, and cognitive function across 22 countries with varying income levels (Rodriguez et al., 2021), findings suggest that education had the dominant effect on cognitive functioning. Of note, this effect was large enough that it may offset the adverse impact of living in poverty on cognitive function. While the current study evaluated one set of factors to represent the concept of “social determinants of health,” additional research is warranted to determine which social determinants are the most significant risk factors for CRCI.

In terms of the other types of stress, indicator variables representing global and cancer-specific stress were included in the patient-specific factors latent variable. As Ahles and Hurria (2018) noted, studies are needed to evaluate stress as a risk factor for CRCI. While this study aimed not to examine the effect of the individual indicator variables, global stress was the variable within the patient-specific factors latent variable that had the largest association with CRCI (Appendix A). This finding is consistent with previous research that found that higher levels of perceived stress were an independent predictor for self-reported CRCI (Kim & Abraham, 2021). In addition, this finding supports the need to evaluate various types of stress as risk factors for CRCI.

Interestingly, the simultaneous SRM that evaluated the joint effect of the latent variables that represented the four MMCRCI concepts on CRCI was not significant. This finding was unexpected for two reasons. First, each exogenous latent variable was independently and significantly associated with the CRCI latent variable. Second, based on conservative estimates of observations to predictor ratios for SRM (e.g., 15:1), the large sample size in the current study allowed for evaluation of a complex SRM (Babyak, 2004). However, it is possible that the level of multicollinearity among the variables and/or small effect sizes contributed to this result (Babyak, 2004). Taken together, the joint effect of the four MMCRCI concepts may be difficult to parse out when evaluated in a complex SRM despite an adequate sample size. Rather than a complex SRM, future studies using the MMCRCI can test smaller and/or individual parts of the model.

Treatment factors of the four MMCRCI concepts evaluated explained the smallest amount of variance in CRCI. Data on a relatively comprehensive list of treatment-related factors (i.e., white blood cell count, hemoglobin level, a rating of the toxicity of the chemotherapy regimen, number and type[s] of medications in the antiemetic regimen, chemotherapy cycle length) were included in this exogenous latent variable. This finding supports previous research that found that CRCI occurs independent of treatment factors (e.g., it happens prior to treatment; Bai & Yu, 2021), months to years after completion of treatment (Lv et al., 2020), across various cancer types (Vannorsdall, 2017), independent of treatment regimen (Janelsins et al., 2017). While not evaluated routinely, the inclusion of the type of antiemetic regimen was justified because of the potential adverse effects associated with these medications (e.g., mood changes, fatigue) that may impact cognitive function (Adel, 2017). However, some treatment factors that were not included in this exogenous latent variable but are associated with CRCI (i.e., dose intensity (Bai & Yu, 2021), higher number of chemotherapy cycles (Hodgson et al., 2013) need to be evaluated in future studies of the MMCRCI.

In contrast, co-occurring symptoms explained the largest amount of variance in CRCI (Table 4). This exogenous latent variable included some of the most common symptoms associated with cancer and its treatments (i.e., morning and evening fatigue, state anxiety, sleep disturbance, depression, pain; Miaskowski et al., 2014). Our findings are consistent with previous research that found that decrements in cognitive function were associated with each of these co-occurring symptoms (i.e., fatigue; Abd-Elfattah et al., 2015), anxiety (Smith et al., 2021), sleep disturbance (Song et al., 2021), depression (Rock et al., 2014), and pain (Zis et al., 2017).

In addition, the hierarchical regression models demonstrated the unique contribution of co-occurring symptoms on CRCI even after controlling for social determinants of health, patient-specific factors, and treatment factors (Table 4). Across these three models, co-occurring symptoms accounted for a large amount of variance in CRCI. These findings showcase several critical directions for future research. First, common mechanism(s) may be driving multiple co-occurring symptoms in patients with cancer. Importantly, research that focuses on identifying common mechanism(s) for co-occurring symptoms is sparse (Harris et al., 2022). Second, future studies need to consider evaluating other common symptoms that co-occur with CRCI. As noted by Lacourt et al. (2018), a need exists to identify different phenotypes of CRCI based on the presence of other co-occurring symptoms. Finally, our findings support previous research that suggests that intervention strategies that can effectively target more than one symptom may significantly improve cognitive function (Vega et al., 2022).

While this study has numerous strengths (e.g., first study to evaluate the MMCRCI, use of a large sample of patients receiving chemotherapy, and inclusion of a variety of factors known or hypothesized to be associated with CRCI), some limitations are worth noting. First, operationalizing the concepts and outcomes for evaluating the MMCRCI were limited to the available data and/or appropriateness for use in SRM. Other indicator variables can be used to define and test this model and may yield different findings. For example, testing this model based on an objective measure of CRCI may provide different insights into the relationships between/among the various concepts in the MMCRCI. In addition, because other potentially important risk factors for CRCI (e.g., gender, type of cancer) were represented by nominal variables in this study, they could not be evaluated as part of a latent variable. Finally, these data represent one time point in the treatment trajectory. Longitudinal evaluation of these findings is warranted in future studies.

Conclusion

Our findings suggest that testing individual components of the MMCRCI may provide useful information on the relationships among various risk factors for CRCI, as well as refinements of the model. In terms of risk factors for CRCI, co-occurring symptoms may be more essential than treatment factors, patient-specific factors, and/or social determinants of health in patients receiving chemotherapy. This knowledge can be used to design future studies as well as prioritize interventions to prevent and/or improve CRCI.

Supplementary Material

Supplemental Data File (doc, pdf, etc.)

Acknowledgments

This study was funded by a grant from the National Cancer Institute (CA134900). Ms. Oppegaard was supported by a grant from the National Institute of Nursing Research (T32NR016920) and the Oncology Nursing Foundation. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Dr. Miaskowski is an American Cancer Society clinical research professor.

The study was approved by the Committee on Human Research at the University of California, San Francisco, and the institutional review boards at each study site. All patients gave written informed consent.

Footnotes

The authors have no conflicts of interest to report.

References

  1. Abd-Elfattah HM, Abdelazeim FH, & Elshennawy S (2015). Physical and cognitive consequences of fatigue: A review. Journal of Advanced Research, 6, 351–358. 10.1016/j.jare.2015.01.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Adel N (2017). Overview of chemotherapy-induced nausea and vomiting and evidence-based therapies. American Journal of Managed Care, 23, S259–S265. [PubMed] [Google Scholar]
  3. Ahles TA, & Hurria A (2018). New Challenges in Psycho‐Oncology Research IV: Cognition and cancer: Conceptual and methodological issues and future directions. Psycho-Oncology, 27, 3–9. 10.1002/pon.4564 [DOI] [PubMed] [Google Scholar]
  4. Babyak MA (2004). What you see may not be what you get: A brief, nontechnical introduction to overfitting in regression-type models. Psychosomatic Medicine, 66, 411–421. [DOI] [PubMed] [Google Scholar]
  5. Bai L, & Yu E (2021). A narrative review of risk factors and interventions for cancer-related cognitive impairment. Annals of Translational Medicine, 9, 72. 10.21037/atm-20-6443 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bentler PM (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107, 238–246. 10.1037/0033-2909.107.2.238 [DOI] [PubMed] [Google Scholar]
  7. Browne MW, & Cudeck R (1992). Alternative ways of assessing model fit. Sociological Methods & Research, 21, 230–258. 10.1177/0049124192021002005 [DOI] [Google Scholar]
  8. Cimprich B, So H, Ronis DL, & Trask C (2005). Pre-treatment factors related to cognitive functioning in women newly diagnosed with breast cancer. Psycho–Oncology, 14, 70–78. 10.1002/pon.821 [DOI] [PubMed] [Google Scholar]
  9. Cimprich B, Visovatti M, & Ronis DL (2011). The Attentional Function Index—A self-report cognitive measure. Psycho–Oncology, 20, 194–202. 10.1002/pon.1729 [DOI] [PubMed] [Google Scholar]
  10. Cohen S, Kamarck T, & Mermelstein R (1983). A global measure of perceived stress. Journal of Health and Social Behavior, 24, 385–396. https://www.ncbi.nlm.nih.gov/pubmed/6668417 [PubMed] [Google Scholar]
  11. Connor KM, & Davidson JRT (2003). Development of a new resilience scale: The Connor–Davidson Resilience Scale (CD–RISC). Depression and Anxiety, 18, 76–82. 10.1002/da.10113 [DOI] [PubMed] [Google Scholar]
  12. Costa PT, & McCrae RR (1992). Revised NEO Personality Inventory (NEOPI–R) and NEO Five-Fator Inventory (NEO–FFI). Psychological Assessment Resources. [Google Scholar]
  13. Curtis RG, Windsor TD, & Soubelet A (2015). The relationship between Big-5 personality traits and cognitive ability in older adults—A review. Aging, Neuropsychology, and Cognition, 22, 42–71. 10.1080/13825585.2014.888392 [DOI] [PubMed] [Google Scholar]
  14. Daut RL, Cleeland CS, & Flanery RC (1983). Development of the Wisconsin Brief Pain Questionnaire to assess pain in cancer and other diseases. Pain, 17, 197–210. 10.1016/0304-3959(83)90143-4 [DOI] [PubMed] [Google Scholar]
  15. Extermann M, Bonetti M, Sledge GW, O’Dwyer PJ, Bonomi P, & Benson AB III. (2004). MAX2—A convenient index to estimate the average per patient risk for chemotherapy toxicity: Validation in ECOG trials. European Journal of Cancer, 40, 1193–1198. 10.1016/j.ejca.2004.01.028 [DOI] [PubMed] [Google Scholar]
  16. Harris CS, Kober KM, Conley YP, Dhruva AA, Hammer MJ, & Miaskowski CA (2022). Symptom clusters in patients receiving chemotherapy: A systematic review. BMJ Supportive & Palliative Care, 12, 10–21. 10.1136/bmjspcare-2021-003325 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Hermelink K, Küchenhoff H, Untch M, Bauerfeind I, Lux MP, Bühner M, Manitz J, Fensterer V, & Münzel K (2010). Two different sides of ‘chemobrain’: Determinants and nondeterminants of self-perceived cognitive dysfunction in a prospective, randomized, multicenter study. Psycho–Oncology, 19, 1321–1328. 10.1002/pon.1695 [DOI] [PubMed] [Google Scholar]
  18. Hodgson KD, Hutchinson AD, Wilson CJ, & Nettelbeck T (2013). A meta-analysis of the effects of chemotherapy on cognition in patients with cancer. Cancer Treatment Reviews, 39, 297–304. 10.1016/j.ctrv.2012.11.001 [DOI] [PubMed] [Google Scholar]
  19. Janelsins MC, Heckler CE, Peppone LJ, Kamen C, Mustian KM, Mohile SG, Magnuson A, Kleckner IR, Guido JJ, Young KL, Conlin AK, Weiselberg LR, Mitchell JW, Ambrosone CA, Ahles TA, & Morrow GR (2017). Cognitive complaints in survivors of breast cancer after chemotherapy compared with age-matched controls: An analysis from a nationwide, multicenter, prospective longitudinal study. Journal of Clinical Oncology, 35, 506–514. 10.1200/JCO.2016.68.5826 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Karnofsky DA, Abelmann WH, Craver LF, & Burchenal JH (1948). The use of nitrogen mustards in the palliative treatment of carcinoma. Cancer, 1, 634–656. [Google Scholar]
  21. Kim H-J, & Abraham I (2021). Determinants of the higher prevalence and severity of subjective cognitive impairment in cancer patients compared to healthy subjects: Fatigue and stress. Clinical Nursing Research, 30, 809–817. 10.1177/1054773820957474 [DOI] [PubMed] [Google Scholar]
  22. Kline RB (2016). Principles and practice of structural equation modeling (4th ed.). Guilford Press. [Google Scholar]
  23. Kwak SG, & Kim JH (2017). Central limit theorem: The cornerstone of modern statistics. Korean Journal of Anesthesiology, 70, 144–156. 10.4097/kjae.2017.70.2.144 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Lacourt TE, De La Garza R II, & Dantzer R (2018). Can cancer-related cognitive impairment be considered in isolation from other cancer-related symptoms? Psycho–Oncology, 27, 2511–2512. 10.1002/pon.4826 [DOI] [PubMed] [Google Scholar]
  25. Lee JL, & Jeong Y (2019). Quality of life in patients with non-small cell lung cancer: Structural equation modeling. Cancer Nursing, 42, 475–483. 10.1097/ncc.0000000000000645 [DOI] [PubMed] [Google Scholar]
  26. Lee KA (1992). Self-reported sleep disturbances in employed women. Sleep, 15, 493–498. 10.1093/sleep/15.6.493 [DOI] [PubMed] [Google Scholar]
  27. Lee KA, Hicks G, & Nino-Murcia G (1991). Validity and reliability of a scale to assess fatigue. Psychiatry Research, 36, 291–298. 10.1016/0165-1781(91)90027-m [DOI] [PubMed] [Google Scholar]
  28. Lopez M, Ruiz MO, Rovnaghi CR, Tam GK-Y, Hiscox J, Gotlib IH, Barr DA, Carrion VG, & Anand KJS (2021). The social ecology of childhood and early life adversity. Pediatric Research, 89, 353–367. 10.1038/s41390-020-01264-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Lv L, Mao S, Dong H, Hu P, & Dong R (2020). Pathogenesis, assessments, and management of chemotherapy-related cognitive impairment (CRCI): An updated literature review. Journal of Oncology, 2020, 3942439. 10.1155/2020/3942439 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Maruyama GM (1997). Basics of structural equation modeling. SAGE Publications. [Google Scholar]
  31. Mayo SJ, Wozniczka I, Edwards B, Rourke SB, Howell D, Metcalfe KA, Haghayegh AT, & Lipton JH (2022). A qualitative study of the everyday impacts of cognitive difficulties after stem cell transplantation. Oncology Nursing Forum, 49, 315–325. 10.1188/22.Onf.315-325 [DOI] [PubMed] [Google Scholar]
  32. Miaskowski C, Cooper BA, Melisko M, Chen L-M, Mastick J, West C, Paul SM, Dunn LB, Schmidt BL, Hammer M, Cartwright F, Wright F, Langford DJ, Lee K, & Aouizerat BE (2014). Disease and treatment characteristics do not predict symptom occurrence profiles in oncology outpatients receiving chemotherapy. Cancer, 120, 2371–2378. 10.1002/cncr.28699 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Mikhail JN, Nemeth LS, Mueller M, Pope C, & NeSmith EG (2018). The social determinants of trauma: A trauma disparities scoping review and framework. Journal of Trauma Nursing, 25, 266–281. 10.1097/jtn.0000000000000388 [DOI] [PubMed] [Google Scholar]
  34. Muthén B, & Shedden K (1999). Finite mixture modeling with mixture outcomes using the EM algorithm. Biometrics, 55, 463–469. 10.1111/j.0006-341X.1999.00463.x [DOI] [PubMed] [Google Scholar]
  35. Oppegaard KR, Mayo SJ, Armstrong TS, Anguera JA, Kober K, & Miaskowski C (2023). The multifactorial model of cancer-related cognitive impairment. Oncology Nursing Forum, 50, 135–147. [DOI] [PubMed] [Google Scholar]
  36. Radloff LS (1977). The CES–D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385–401, 10.1177/014662167700100306. [DOI] [Google Scholar]
  37. Raykov T, & Marcoulides GA (2006). A first course in structural equation modeling (2nd ed.). Lawrence Erlbaum Associates, Publishers. [Google Scholar]
  38. Ren X, Boriero D, Chaiswing L, Bondada S, St. Clair DK, & Butterfield DA (2019). Plausible biochemical mechanisms of chemotherapy-induced cognitive impairment (“chemobrain”), a condition that significantly impairs the quality of life of many cancer survivors. Biochimica et Biophysica Acta (BBA) – Molecular Basis of Disease, 1865, 1088–1097. 10.1016/j.bbadis.2019.02.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Rock PL, Roiser JP, Riedel WJ, & Blackwell AD (2014). Cognitive impairment in depression: A systematic review and meta-analysis. Psychological Medicine, 44, 2029–2040. 10.1017/s0033291713002535 [DOI] [PubMed] [Google Scholar]
  40. Rodriguez FS, Hofbauer LM, & Röhr S (2021). The role of education and income for cognitive functioning in old age: A cross-country comparison. International Journal of Geriatric Psychiatry, 36, 1908–1921. 10.1002/gps.5613 [DOI] [PubMed] [Google Scholar]
  41. Sangha O, Stucki G, Liang MH, Fossel AH, & Katz JN (2003). The Self-Administered Comorbidity Questionnaire: A new method to assess comorbidity for clinical and health services research. Arthritis & Rheumatism, 49, 156–163. 10.1002/art.10993 [DOI] [PubMed] [Google Scholar]
  42. Schmidt JE, Beckjord E, Bovbjerg DH, Low CA, Posluszny DM, Lowery AE, Dew MA, Nutt S, Arvey SR, & Rechis R (2016). Prevalence of perceived cognitive dysfunction in survivors of a wide range of cancers: Results from the 2010 LIVESTRONG survey. Journal of Cancer Survivorship, 10, 302–311. 10.1007/s11764-015-0476-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Smith L, Jacob L, López-Sánchez GF, Butler L, Barnett Y, Veronese N, Soysal P, Yang L, Grabovac I, Tully MA, Shin JI, & Koyanagi A (2021). Anxiety symptoms and mild cognitive impairment among community-dwelling older adults from low- and middle-income countries. Journal of Affective Disorders, 291, 57–64. 10.1016/j.jad.2021.04.076 [DOI] [PubMed] [Google Scholar]
  44. Song D, Zhou J, Ma J, Chang J, Qiu Y, Zhuang Z, Xiao H, & Zeng L (2021). Sleep disturbance mediates the relationship between depressive symptoms and cognitive function in older adults with mild cognitive impairment. Geriatric Nursing, 42, 1019–1023. 10.1016/j.gerinurse.2021.06.004 [DOI] [PubMed] [Google Scholar]
  45. Spielberger CD, Gorsuch RL, Luchene RE, Vagg PR, & Jacobs GA (1983). Manual for the State-Trait Anxiety Inventory. Consulting Psychologists Press. [Google Scholar]
  46. Vannorsdall TD (2017). Cognitive changes related to cancer therapy. Medical Clinics of North America, 101, 1115–1134. 10.1016/j.mcna.2017.06.006 [DOI] [PubMed] [Google Scholar]
  47. Vega JN, Albert KM, Mayer IA, Taylor WD, & Newhouse PA (2022). Subjective cognition and mood in persistent chemotherapy-related cognitive impairment. Journal of Cancer Survivorship, 16, 614–623. 10.1007/s11764-021-01055-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Weiss DS, & Marmar CR (1997). The Impact of Event Scale – Revised. In Wilson JP & Keane TM (Eds.), Assessing psychological trauma and PTSD (pp. 399–411). Guilford Press. [Google Scholar]
  49. Wolfe J, & Kimerling R (1997). Gender issues in the assessment of posttraumatic stress disorder. In Wilson JP & Keane TM (Eds.), Assessing psychological trauma and PTSD (pp. 192–238). Guilford Press. [Google Scholar]
  50. Zhang H, Ma W, Wang G, Wang S, & Jiang X (2021). Effects of psychosocial factors on posttraumatic growth among lung cancer patients: A structural equation model analysis. European Journal of Cancer Care, 30, e13450. 10.1111/ecc.13450 [DOI] [PubMed] [Google Scholar]
  51. Zis P, Daskalaki A, Bountouni I, Sykioti P, Varrassi G, & Paladini A (2017). Depression and chronic pain in the elderly: Links and management challenges. Clinical Interventions in Aging, 2017, 709–720. 10.2147/cia.S113576 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Data File (doc, pdf, etc.)
NIHMS1888086-supplement-Supplemental_Data_File__doc__pdf__etc__.docx (4.6MB, docx)

RESOURCES