• Open Access
    Original Article

    Lifetime stressors relate to invisible symptoms of multiple sclerosis

    Carri S. Polick 1,2*
    Tiffany J. Braley 3
    Robert Ploutz-Snyder 4
    Cathleen M. Connell 5
    Ali Watson 6
    Sarah A. Stoddard 7

    Explor Neuroprot Ther. 2024;4:158–171 DOI: https://doi.org/10.37349/ent.2024.00077

    Received: January 05, 2024 Accepted: March 28, 2024 Published: April 22, 2024

    Academic Editor: Manuel Zeitelhofer, Karolinska Institutet, Sweden



    Childhood stressors can increase adult stress perception and may accumulate over the lifespan to impact symptoms of multiple sclerosis (MS). Growing evidence links childhood stressors (e.g., abuse, neglect) to fatigue, pain, and psychiatric morbidity in adults with MS; yet literature in this area is lacking a comprehensive lifespan approach. The aim of this cross-sectional study was to examine contributions of childhood and adulthood stressor characteristics (i.e., count, severity), on three individual outcomes: fatigue, pain interference, and psychiatric morbidity in People with MS (PwMS).


    An online survey was distributed through the National MS Society. Hierarchical block regression modeling was used to sequentially assess baseline demographics, childhood stressors, and adult stressors per outcome. We hypothesized that child and adult stressors would significantly contribute to fatigue, pain interference, and psychiatric morbidity.


    Overall, 713 PwMS informed at least one final analytic model. Both childhood and adult stressors significantly contributed to pain interference and psychiatric morbidity. Adult stressor severity independently correlated with psychiatric morbidity (P < 0.0001). Childhood stressors significantly contributed to fatigue (LR test P < 0.0001). Childhood stressor severity independently significantly correlated with both fatigue likelihood (P = 0.03) and magnitude (P < 0.001).


    This work supports a relationship between stressors across the lifespan and fatigue, pain, and psychiatric morbidity in PwMS. Stressor severity may have an important role which may not be captured in count-based trauma measurement tools. Clinicians and researchers should consider lifetime stress when addressing fatigue, pain, and psychiatric morbidity among PwMS.


    Adverse childhood experiences, stressors, multiple sclerosis, fatigue, pain, psychiatric morbidity


    Traumatic childhood stressors (e.g., abuse, neglect) are associated with many negative biopsychosocial health outcomes including immune-mediated diseases and symptoms [13]. Research on adverse childhood experiences (ACEs) and multiple sclerosis (MS) is a quickly growing field, yet the focus has primarily been on the risk for developing MS and not disease burden or chronic symptoms which heavily impact the lives of people with MS (PwMS) [4, 5]. The few studies that have focused on common “invisible symptoms” of MS have found that childhood maltreatment was associated with adult MS fatigue [6, 7], pain catastrophizing [8], anxiety [9], and psychiatric morbidity (e.g., anxiety, depression) [10].

    Additionally, stressor measurement is inconsistent and still evolving. For example, most studies use count-based scales (e.g., ACEs), or tools that include severity of only a few stressors [e.g., Childhood Trauma Questionnaire (CTQ)] [5]. There is also a lack of consensus regarding what qualifies as a stressor. Indeed, with increased recognition of the importance of social determinants of health [11], additional factors such as unstable housing and discrimination have been incorporated into newer measures. For example, the CTQ only captures five core stressors of physical abuse/neglect, emotional abuse/neglect, and sexual abuse; while newer measures like the Stress and Adversity Inventory (STRAIN) capture experiences of unstable housing and being excluded because of personal factors like race or gender [12]. Recently, the STRAIN was used to evaluate only the stressors aligning with expanded ACE criteria, which revealed associations between emotional and physical stressors (e.g., abuse severity/duration) and invisible symptoms of fatigue, pain interference, and psychiatric morbidity in adults with MS [7]. However, limiting stressor measurement to only childhood provides limited insight, especially given that life stressors continue into adulthood, and that more childhood stress has been correlated with increased adult stress perception [13]. Consequently, studies that don’t include exposure to stress across the lifespan may miss predictive adult information and overestimate relationships with childhood stressors [14].

    A lifetime approach can help address the literature gap between child and adult stressor research, has the added value of potentially elucidating when stressors have the most impact on MS, and is more applicable to adult healthcare settings. The purpose of this study was to evaluate comprehensively measured lifetime stressors (e.g., cumulative child, adult, count, severity), to answer the research question of whether these stressors relate to three common invisible features of MS—fatigue, pain interference, and psychiatric morbidity, in a large national sample of PwMS. We used a hierarchical block modeling approach to highlight the importance of stressor timing (childhood vs. adulthood) and optimize future preventative and mitigation efforts. We hypothesized that successive models with a cumulative childhood stressor block and a cumulative adult stressor block would contribute significantly more predictive variance over the previous nested models, and thus, significantly associate to each outcome.

    Materials and methods

    The current study is a secondary data analysis of the Stress-MS dataset created by Polick et al., 2023 [7, 14]. Online surveys were distributed to US-based adults with MS in October 2021 via the National MS Society (NMSS) listserv including nearly 80,000 PwMS. STROBE guidelines were followed to strengthen reporting and transparency of observational studies [15]. Ethical approval was obtained from the University of Michigan and participants gave implied consent.



    The STRAIN encompasses 55 lifetime stressors including abuse, neglect, household dysfunction, housing instability, neighborhood safety, infertility, financial strain, and feeling excluded based on personal factors like race or gender [12]. If a participant endorsed a stressor, follow up questions captured the age at which it happened and stressor severity. Stressor severity items are scored on a 0–5 Likert scale from “very slightly or not at all” to “extremely”. Stressors were assessed as cumulative childhood count/severity and cumulative adult count/severity; higher scores represent higher stress.


    Patient Reported Outcome Information System (PROMIS) tools were used to measure pain interference and fatigue. PROMIS-Pain Interference is a validated 8-item questionnaire measuring the impact of pain on the mental, physical, and social aspects of life in the past week [16], which has been used with PwMS [17]. Likert scale scoring from 1 (not at all) to 5 (very much), indicates higher pain interference with higher scores. Reliability was very high in this study (Cronbach’s alpha = 0.98).

    The PROMIS-Fatigue MS Short Form is a validated 8-item questionnaire measuring fatigue in the last week specific to PwMS [18, 19]. Likert scale scoring from 1 (never) to 5 (always); indicates higher fatigue with higher scores. Reliability was very high in this study (Cronbach’s alpha = 0.95).

    Psychiatric morbidity is a composite count score including elements of self-reported diagnoses and symptoms, focused primarily on the most common challenges for PwMS (e.g., anxiety, depression). This approach aligns with how this concept has previously been measured in an MS sample in the child stress literature and used here to promote better comparisons to bolster the lifetime stressor literature [7]. This approach captures both PwMS who may be symptomatic but not diagnosed and those who may be diagnosed but no longer symptomatic. Four item PROMIS-Anxiety and PROMIS-Depression tools were used to measure symptoms. For parsimony in already complex modeling, PROMIS anxiety and depression scores were each dichotomized into symptomatic (1) (i.e., any positive score) and not symptomatic (0), and then summed with other dichotomous variables including an anxiety diagnosis (0/1), depression diagnosis (0/1), or presence of other diagnoses [e.g., bipolar, schizophrenia, post-traumatic stress disorder (PTSD), 0/1]. Summed scores ranged from 0–6; higher scores indicated higher psychiatric morbidity.


    Demographic and MS covariates were used in each analysis (i.e., age, gender, education, MS subtype). Treatments, such as disease modifying therapy (DMT) and a count of different types of medications that can impact pain (e.g., opiates, antidepressants), were used in pertinent analyses (Table 1).

    Predictors per hierarchical block modeling approach

    Sequential modelingPredictors per each model
    Base model 1: demographics and MS covariatesAge, gender, education, MS subtype, DMTa, pain medication countb
    Model 2 adds childhood stressorsBase model 1 + childhood stressor count, childhood stressor severity
    Model 3 adds adult stressorsModel 2 + adult stressor count, adult stressor severity
    Display full size

    a DMTs not included in psychiatric morbidity analysis; b pain medication count only included for pain interference and psychiatric morbidity analyses

    Data screening and pre-processing

    Raw PROMIS scores were transformed to normalized t-scores (https://www.healthmeasures.net/). Scores representing “no pain” or “no fatigue” were replaced with zeros to evaluate these outcomes appropriately using a two-part statistical model described below. Structural and social stressors disproportionately occur among minoritized populations, inhibiting our ability to disentangle race and racism. Thus, race and ethnicity variables were not included in the main analyses to not violate statistical principles (e.g., collinearity). These efforts align with our goal to thoughtfully reconsider attributing statistical onus on race, versus what social and health system structural factors contribute to outcomes for PwMS [20, 21].

    Analytic strategy

    Aligning with previous work that evaluated lifetime stressors and physical outcomes [14], our analytic approach aimed to assess fit of increasingly complex models that include blocks of related predictors (i.e., collinear variables representing latent constructs) to determine if their collective contributions improve model fit. Of note, this type of analysis focuses on establishing whether there is a relationship between the latent constructs (e.g., blocks) and the outcomes, and not necessarily the change or individual variable contributions because they could be underestimated. Successive models were compared to prior models using likelihood ratio (LR) testing, and Akaike Information Criterion (AIC) as an index of relative model fit, with lower AIC indicating better fit. If a block of predictors did not significantly improve the model, it was removed from final analytic modeling. Table 1 shows the predictors and covariates in each of the three blocks, with each model nested within the next higher-level model.

    The base model encompassed covariates to determine baseline contributions. Model fit of the base model was then compared to Model 2 which added childhood stressor predictors to assess if these contribute over and above the base model. For Model 3, adult stressor predictors were added to determine if they contributed additional predictive variance. After evaluating contributions of blocks of related predictors, contributions of individual predictors were examined. Yet, it must be noted that using related stressor variables is a strength of block modeling but may also cause an underestimation of individual stressor contributions.

    The specific types of hierarchical regressions included Poisson regression for the count of psychiatric morbidities outcome. The two PROMIS outcomes, pain interference, and fatigue, represented a mixed distribution therefore two-part modeling was used. Part 1 included a dichotomous (yes/no) component of experiencing any pain interference or fatigue utilizing logistic regression, followed by part 2 a normal distribution characterizing the magnitude of pain interference or fatigue utilizing OLS linear regression.


    Reflective of the conventional US MS research population, including previous studies using the NMSS listserv and other large studies, most participants were female (n = 597, 84%), White (n = 415, 88%), with relapsing remitting MS (RRMS, n = 559, 78%), and a college education (Table 2) [17, 22, 23]. Compared to the normalized t-scores of a healthy general population who’s mean (SD) is 50 (10) [24], this sample had higher mean fatigue 57 (9) and pain interference 53 (10.5). On average, participants experienced 2.6 (1.96) stressors in childhood with a severity of 9.8 (8.8), and 23.6 (14) stressors during adulthood with a severity of 55.3 (30.8).

    Sample characteristics

    CharacteristicsMean (SD)
    Age, mean (SD) (n = 713)49 (12.7) range: 21–85
    Length of time since MS onset, mean (SD) (n = 713)18 (12) range: 0–59
    Gender, n (%) (n = 712)
        Female597 (84%)
        Male100 (14%)
        Transgender, non-binary, gender non-conforming, or other15 (2%)
    MS subtype, n (%) (n = 712)
        Relapsing remitting MS (RRMS)559 (78%)
        Secondary progressive MS (SPMS)87 (12%)
        Primary progressive MS (PPMS)35 (5%)
        Progressive Relapsing MS (PRMS)9 (1%)
        Unsure23 (3%)
    DMT, n (%) (n = 709)
        None129 (18%)
        First line272 (38%)
        Second line308 (43%)
    Count of medication classes that can impact pain, mean (SD) (n = 705)1.63 (1.29) range: 0–5
    Race/ethnicity, n (%) (n = 471)
        White415 (88%)
        Bi-racial or mixed24 (5%)
        Black23 (5%)
        Asian4 (< 1%)
        Latinx2 (< 1%)
        American Indian or Alaska Native2 (< 1%)
        Native Hawaiian or Pacific Islander1 (< 1%)
    Smoking status, n (%) (n = 709)
        Never smoker465 (66%)
        Former smoker198 (28%)
        Current or social smoker46 (6%)
    Education, n (%) (n = 713)
        High school equivalency or below36 (5%)
        Associate degree or some college167 (23%)
        Bachelor’s degree259 (36%)
        Master’s degree or above251 (35%)
    Stressors, mean (SD) (n = 713)
        Childhood count 2.6 (1.96)
        Childhood severity9.8 (8.8)
        Adult count23.6 (14)
        Adult severity55.3 (30.8)
    Outcome variables (n = 713)
        Fatigue, median (IQR), mean (SD)58 (52–63), 57 (9)
        Pain interference, median (IQR), mean (SD)54 (41–62), 53 (10.5)
        Psychiatric morbidity count, mean (SD)2.2 (1.7) range: 0–6
    Display full size

    Pain interference

    Base model predictors contributed to a significant overall two-part model estimating the likelihood and magnitude of pain interference (logistic regression pseudo R2 = 0.2219, P < 0.0001; OLS regression R2 = 0.1831, P < 0.0001; model AIC = 3,751) (Table 3). The childhood stressor block of predictors in Model 2 improved predictions significantly over the base model (logistic regression pseudo R2 = 0.2448, P < 0.0001; OLS regression R2 = 0.2152, P < 0.0001; model AIC = 3,719, LR P < 0.0001). Similarly, the adult stressor predictors in Model 3 contributed significantly more information over the prior nested modes (logistic regression pseudo R2 = 0.2578, P < 0.0001; OLS regression R2 = 0.2667, P < 0.0001; model AIC = 3,685, LR P < 0.0001).

    Final analytic model of pain interference using two-part regression modeling

    Variables within hierarchical modelsFirst part—logistic regression
    (n = 701)
    Any pain interference (binary)
    Second part—OLS regression
    (n = 459)
    Magnitude of pain interference
    Overall model stats
    ORSE95% CIPPseudo R2bSE95% CIPR2AICLR test
    Base covariates< 0.00010.222< 0.00010.1833,751Base
    Age1.020.011.00–1.040.02–0.100.03–0.16––0.05< 0.001
    Gender (ref. female)
        Transgender, non-binary, gender non-conforming, or other1.210.880.29–5.060.80–2.121.91–5.86–1.630.27
    Education (ref. ≤ HS)
        Associates degree or some college0.130.110.03–0.650.01–0.861.23–3.27–1.550.47
        Bachelor’s degree0.090.070.02–0.44< 0.01–2.881.22–5.26––0.500.02
        Master’s degree or above0.080.070.02–0.40< 0.01–3.431.23–5.83––1.03< 0.01
    MS subtype (ref. RRMS)
        SPMS1.570.540.81––4.36< 0.01
        PRMS1.351.290.21–8.80.755.712.361.08–10.35< 0.02
    DMT (ref. no therapy)
        First line1.490.430.85–2.610.17–0.630.91–2.39–1.140.49
        Second line2.220.651.25–3.95< 0.01–1.320.86–3.01–0.380.13
    Pain med count2.310.231.90–2.80< 0.0011.020.240.56–1.48< 0.001
    Childhood stressors< 0.00010.245< 0.00010.2153,719< 0.0001
    Child stressor count1.190.220.83–1.710.34–0.580.54–1.63–0.470.28
    Child stressor severity0.990.040.91–1.070.730.140.12–0.10–0.380.24
    Adult stressors< 0.00010.258< 0.00010.2673,685< 0.0001
    Adult stressor count1.040.021.00––0.04–0.160.24
    Adult stressor severity1.000.010.98–1.020.920.040.02–0.01–0.080.10
    Display full size

    ≤ HS: high school equivalency or below; PPMS: primary progressive MS; PRMS: progressive-relapsing MS; ref.: reference; SPMS: secondary progressive MS. Blank cells indicate not applicable to individual variables

    Regarding individual predictors, childhood stress severity was significantly associated with the higher magnitude of pain interference (b = 0.33, P = 0.005) in Model 2 but lost significance when adult stress was added for Model 3, suggesting shared variance among child and adult stressors. In the final model, age impacted both the likelihood (OR = 1.02, P < 0.03) and magnitude (b = –0.10, P < 0.001) of pain.


    The base model of predictors contributed to a significant overall two-part model estimating the likelihood and magnitude of having fatigue (logistic regression pseudo R2 = 0.074, P < 0.04; OLS regression R2 = 0.086, P < 0.0001; model AIC = 4,192) (Table 4). The childhood stressor predictors in Model 2 contributed a significant amount of variance over and above the base model (logistic regression pseudo R2 = 0.11, P < 0.01; OLS regression R2 = 0.14, P < 0.0001; model AIC = 4,160, LR P < 0.0001). While adult stressor severity was independently significant for the magnitude of fatigue (b = 0.073, P = 0.001), overall, the adult stressor predictors in Model 3 did not significantly contribute and reduced model fit (logistic regression pseudo R2 = 0.136, P = 0.004; OLS regression R2 = 0.242, P < 0.0001; model AIC = 4,844, LR P = 1.0). Thus, adult stressors were removed from the final analytic model, as only base covariates and childhood stressors correlated with fatigue.

    Final analytic model of fatigue using two-part regression modeling

    Variables within hierarchical modelsFirst part—logistic regression
    (n = 600)
    Any fatigue (binary)
    Second part—OLS regression
    (n = 576)
    Magnitude of fatigue
    Overall model stats
    ORSE95% CIPPseudo R2bSE95% CIPR2AICLR test
    Base covariates< 0.040.074< 0.00010.0864,192Base
    Gender (ref. female)
    Education (ref. ≤ HS)
        Bachelor’s degree0.140.140.02–1.070.06–2.480.82–4.09–0.880.002
        Master’s degree or above0.160.170.02–1.250.08–4.510.81–6.10––2.92< 0.001
    DMT (ref. no therapy)
        First line0.600.410.16–2.270.45–1.340.97–3.23–0.560.17
        Second line1.841.450.39–8.580.440.250.97–1.65–2.150.80
    MS subtype (ref. RRMS)
    Childhood stressors0.010.105< 0.00010.1384,160< 0.0001
    Child stressor count0.510.190.24–1.070.07–1.230.59–2.38––0.08< 0.04
    Child stressor severity1.240.121.02–1.510.030.470.130.21–0.74< 0.001
    Display full size

    Categories within variables dropped from the model based on collinearity: 1) transgender, non-binary, gender non-conforming, or other, 2) associate degree or some college, 3) PPMS, 4) PRMS, 5) unsure. ≤ HS: high school equivalency or below; PPMS: primary progressive MS; PRMS: progressive-relapsing MS; ref.: reference; SPMS: secondary progressive MS. Blank cells indicate not applicable to individual variables

    In the final model, childhood stress severity was significantly associated with 24% higher odds of experiencing any fatigue for each increase 1-unit increase in severity rating (OR = 1.24, P = 0.03), and with the magnitude of fatigue (b = 0.47, P < 0.001). Interpreting this in context of the average childhood stress severity (9.8), this translates to the average PwMS in this sample being 235% more likely to experience fatigue. Childhood stressor count (b = –1.23, P < 0.04) and age (b = –0.07, P = 0.019) were both negatively associated with the magnitude of fatigue.

    Psychiatric morbidity

    The base model contributed significantly to estimating the risk of accumulating psychiatric morbidity (R2 = 0.061, P < 0.0001; model AIC = 2,560) (Table 5). The childhood stressor predictors in Model 2 significantly improved over the base model (R2 = 0.09, P < 0.0001; AIC = 2,485, LR P < 0.0001). Similarly, the adult stressor predictors in Model 3 contributed significantly more information over the prior nested model (R2 = 0.116, P < 0.0001, AIC = 2,420, LR P < 0.0001). Therefore, childhood and adult stressors both correlate with psychiatric morbidity for PwMS.

    Final analytic model of psychiatric morbidity using Poisson regression (n = 705)

    Variables within hierarchical modelsIRRSE95% CIPOverall model statistics
    Pseudo R2AICLR test
    Base covariates< 0.00010.0612,560Base
    Age0.980.0020.98–0.99< 0.001
    Gender (ref. female)
        Transgender, non-binary, gender non-conforming, or other0.970.150.71–1.320.86
    Education (ref. ≤ HS)
        Associate degree or some college0.930.110.75–1.160.54
        Bachelor’s degree0.960.110.77–1.190.70
        Master’s degree or above0.930.110.75–1.170.54
    MS subtype (ref. RRMS)
    Pain med count1.100.021.06–1.15< 0.001
    Childhood stressors< 0.00010.0902,485< 0.0001
    Child stressor count0.970.050.88–1.070.57
    Child stressor severity1.020.011.00–1.040.11
    Adult stressors< 0.00010.1162,420< 0.0001
    Adult stressor count1.000.0040.99–1.0040.27
    Adult stressor severity1.010.0021.006–1.014< 0.0001
    Display full size

    ≤ HS: high school equivalency or below; IRR: incident rate ratio; PPMS: primary progressive MS; PRMS: progressive-relapsing MS; ref.: reference; SPMS: secondary progressive MS. Blank cells indicate not applicable to individual variables

    Regarding individual predictors, adult stress severity was significantly associated with psychiatric morbidity (IRR = 1.01, P < 0.0001). As adult stressor severity increased by 1-unit, the risk of having an additional psychiatric diagnosis or symptom increased by 1%. Interpreting that within the context of the average adult stressor severity in this sample, 55.3 (30.8) this translates to a 55% increased risk for the average PwMS, with nearly 31% more risk just one standard deviation away. In Model 2, childhood stressor severity carried nearly five times that risk, with a 4.8% increased risk of psychiatric morbidity for each 1-unit increase in severity rating, however lost significance when adding the adult stressors for Model 3, again suggesting shared variance. In the final model, the risk of psychiatric morbidity decreased by 2% for each year since MS onset (IRR = 0.98, P < 0.001).



    To our knowledge, this is the first study to use a lifetime approach to comprehensively measure the effect of child and adult stressors on three invisible issues in MS, fatigue, pain interference, and psychiatric morbidity. Use of hierarchical block modeling allowed us to determine the overall contribution of similar stressor variables, count and severity, to better assess the latent concept of stress. Cumulative childhood stressors correlated to all three outcomes, while adult stressors additionally associated with pain interference and psychiatric morbidity (Table 6). This work aligns with the few previous studies which associated only ACE-focused childhood stressors with adult fatigue [6], pain catastrophizing [8], and mental health outcomes [7, 9, 10] in PwMS or immune-mediated inflammatory diseases. Expanding beyond childhood stressor literature, this current study also aligns with evidence suggesting increased adult and lifetime stressors relate to worsening MS outcomes more broadly; although, this literature is largely focused on physical clinical outcomes (e.g., disease onset, relapses, progression) [14, 25, 26]. Specifically, since only childhood stressors related to fatigue in our study, our findings do not align with evidence that adult adversity (i.e., adverse life events in the last 60 days) is associated with MS fatigue [27]. This divergence may stem from measurement differences (e.g., adult vs. lifetime). The current findings suggest that stressor severity may individually carry a more significant impact relative to stressor count for some outcomes; therefore, relying solely on count-based measures (e.g., ACEs) is not ideal. Similarly, evidenced by multiple instances of shared variance or contributions in our analyses, examination of childhood stressors without consideration of adult stressors is not ideal; therefore, a lifetime approach should be used when possible.

    Summary of stressor correlations to MS outcomes

    MS clinical feature outcomesPredictor blocks included in final modelAdditional significant individual stressor contributions to MS outcomes
    Pain interferenceChild & adult stressors
    FatigueChild stressorsChildhood stress severity related to:
    reporting any fatigue & magnitude of fatigue
    Child stress count related to the magnitude of fatigue
    Psychiatric morbidityChild & adult stressorsAdult stress severity
    Display full size

    Blank cell indicates model level significance and no additional individual level contributions

    Omitting race/ethnicity from analyses is not optimal for discerning unique racial/ethnic experiences or differences across outcomes, yet may be necessary due to included stressors (e.g., discrimination) to avoid statistical issues. Since MS samples are largely White, even when studies do include race/ethnicity in analyses, small individual cell sizes typically lead to collapsing multiple categories into a dichotomous variable to abide by ethical/IRB reporting standards to protect participant identity and to create less statistical error variance. A recent review revealed that race/ethnicity was not accounted for in a third of the studies assessing childhood stressors and MS risk or features [5], leaving much room for improvement. More diverse samples are needed to address health disparities and inequities in MS research and treatment; and race and ethnicity should be included in future analyses when possible [28].

    Clinical relevance and future directions

    Our finding that each incremental increase in childhood stressor severity increased the odds of experiencing fatigue by 24% is noteworthy. Fatigue is the most common symptom experienced by PwMS yet remains one of the most challenging symptoms to treat, in part because of its personalized nature that can be influenced by stress. Meta-analyses highlight that mindfulness-based stress reduction approaches may be effective for fatigue, with mixed results for pain for PwMS [29], however, symptoms often co-occur thus broader outcomes should be considered. A recent study by Braley and colleagues found that telephone-based version cognitive behavioral therapy (CBT) for fatigue performed similarly to pharmacological treatment with modafinil [30, 31] in terms of fatigue impact reduction; however, combination therapy with both was associated with more global benefits based on the Patient Global Impression of Change (PGIC) score. Although the PGIC was a secondary outcome in this trial, the findings suggest that an interdisciplinary approach may offer the most benefit when considering a person’s perception of global function. Further, benefits of interventions for invisible symptoms may be best captured by multifaceted instruments that capture overall activity, symptoms, mood, physical, and social function, which themselves can be influenced by stress, and targeted with psychotherapy. Future work is needed to implement and evaluate the benefits of adjunct therapies, in tandem with standard MS care, on invisible symptoms and broader measures.

    Machine learning research has been burgeoning, especially for healthcare applications such as clinical decision making and treatment optimization [32, 33]. Our findings may inform future work such as using stressor history as one of the parameters in supervised machine learning to help determine whether those with high childhood adversity may respond better to cognitive or pharmacological treatment of fatigue. Similarly, integrating stress informed machine learning decision tools may help optimize treatment for other invisible and physical symptoms. Evidence suggests that interventions including stress reduction, coping/resilience skills, and smoking cessation, are useful for symptom management (e.g., pain, fatigue, depression, disability) [3437]. However, such research progress and clinician acceptance of these strategies hinge on health system parameters and infrastructure [38]. Additional implementation studies focused on provision of new services (e.g., stress reduction, coping, smoking cessation clinics), and/or protocols that facilitate increased referrals (e.g., screening, therapy, smoking clinics) are sorely needed to demonstrate feasibility and acceptability throughout various neurological settings to promote clinical buy-in and adoption of translational change.

    As PwMS aged, the risk of psychiatric morbidity decreased by 2%, suggesting that PwMS may be most vulnerable at diagnosis but may learn to cope or feel more in control of their disease over time. This is somewhat supported by the finding that while the presence of pain increased with age, the magnitude of interference in daily life decreased. Interestingly, as both the count of childhood stressors and age increased, the fatigue magnitude decreased, which may suggest that PwMS who experienced more stressors may have already received mental health support and similarly learned to cope better over time. Alternatively, this may also indicate that PwMS who experienced high childhood stressors and potentially high coping skills, may be faring better and participating in research more than their counterparts. Since fatigue was the only outcome to which adult stressors did not have a significant relationship, yet evidence of this relationship has previously been shown to be mediated by resilience [27], there may be additional factors like coping and resilience that may have different mediating impacts across the three outcomes. Other complex considerations may be differing genetic contributions and intergenerational transmission of trauma. As this emerging area of research grows, more prospective and mechanistic work is needed to determine how resilience, coping, and other complex factors may mediate, moderate, or otherwise impact relationships between stressors and health outcomes in PwMS.


    Causal inference cannot be determined with cross sectional data. Yet, as the first study to assess many of these lifetime relationships with an MS focused sample, it fills an important gap. Those who responded may have increased ability to take an online survey (e.g., technology access, less disability). A response bias may be present due to the high number of PwMS on the NMSS listserv (approximately 80,000). Self-reported retrospective data has potential for recall bias, therefore, it is recommended to collect self-reported data using measures which have been validated using a test-retest approach and perform well over time (e.g., STRAIN) [12]. Sensitive information such as stressors could have a social desirability bias and be under-reported. However, the online format and anonymity may have facilitated more accurate reporting compared to other formats. Our sample was highly educated and may not represent all groups. However, this sample aligns with traditional US MS research samples including other studies that used the NMSS listserv and captures the widest geographical range and largest sample size in this emerging area and thus bolsters the internal and external validity [17, 22]. Additional strengths include a wide range of covariates compared to other work in this area [5]. While the PROMIS measures allowed us to compare two outcomes against a healthy group, we did not compare stressor experience. Future studies would be more robust by using a design that allows for a true comparison against other populations of interest (e.g., healthy controls, similar chronic diseases).


    These findings support an association between childhood stressors and pain interference, fatigue, and psychiatric morbidity; as well as an association between adult stressors and pain interference, and psychiatric morbidity for PwMS. Additional studies are needed to assist clinical efforts of trauma informed precision medicine and intervention efforts to mitigate stressor impact on PwMS. While pediatric MS is far less common, investigating stressor experience across different developmental stages may be helpful in evaluating MS outcomes in this sub-population. Future research should replicate this work with more diverse MS samples and expand to include other positive aspects (e.g., coping, resiliency, social support) with mediation analyses and additional clinical features (e.g., sleep, cognition, substance use).



    adverse childhood experiences


    Akaike Information Criterion


    disease modifying therapy


    incident rate ratio


    likelihood ratio


    multiple sclerosis


    National Multiple Sclerosis Society


    Patient Reported Outcome Information System


    people with multiple sclerosis


    relapsing remitting multiple sclerosis


    Stress and Adversity Inventory



    Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of Duke CTSI or the VA.

    Author contributions

    CSP: Conceptualization, Formal analysis, Writing—original draft. TJB: Conceptualization, Resources, Writing—review & editing. RPS: Conceptualization, Formal analysis, Writing—review & editing. CMC: Conceptualization, Writing—review & editing. AW: Writing—review & editing. SAS: Conceptualization, Supervision, Writing—review & editing.

    Conflicts of interest

    The authors have no conflicts of interest to declare.

    Ethical approval

    This study was approved by the University of Michigan IRB (HUM00200716).

    Consent to participate

    Participants read informed consent materials during the screening process and agreed to an implied consent statement to proceed with the study.

    Consent to publication

    Not applicable.

    Availability of data and materials

    Data is available upon request to corresponding author or Harvard Dataverse repository.


    CSP was supported by NIH/NINR grant [T32NR016914], “Complexity: Innovations for Promoting Health and Safety”, Rackham Graduate School, Duke Clinical and Translational Science Institute (CTSI), and Durham VA. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


    © The Author(s) 2024.


    Felitti VJ, Anda RF, Nordenberg D, Williamson DF, Spitz AM, Edwards V, et al. Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults. The Adverse Childhood Experiences (ACE) Study. Am J Prev Med. 1998;14:24558. [DOI] [PubMed]
    Dube SR, Fairweather D, Pearson WS, Felitti VJ, Anda RF, Croft JB. Cumulative childhood stress and autoimmune diseases in adults. Psychosom Med. 2009;71:24350. [DOI] [PubMed] [PMC]
    Grummitt LR, Kreski NT, Kim SG, Platt J, Keyes KM, McLaughlin KA. Association of childhood adversity with morbidity and mortality in US adults: a systematic review. JAMA Pediatr. 2021;175:126978. [DOI] [PubMed] [PMC]
    Eid K, Torkildsen Ø, Aarseth J, Aalstad M, Bhan A, Celius EG, et al. Association of adverse childhood experiences with the development of multiple sclerosis. J Neurol Neurosurg Psychiatry. 2022;93:64550. [DOI] [PubMed] [PMC]
    Polick CS, Polick SR, Stoddard SA. Relationships between childhood trauma and multiple sclerosis: a systematic review. J Psychosom Res. 2022;160:110981. [DOI] [PubMed]
    Pust GEA, Dettmers C, Randerath J, Rahn AC, Heesen C, Schmidt R, et al. Fatigue in multiple sclerosis is associated with childhood adversities. Front Psychiatry. 2020;11:811. [DOI] [PubMed] [PMC]
    Polick CS, Ploutz-Snyder R, Braley TJ, Connell CM, Stoddard SA. Fatigue, pain interference, and psychiatric morbidity in multiple sclerosis: the role of childhood stress. PLoS One. 2023;18:e0292233. [DOI] [PubMed] [PMC]
    MacDonald TM, Fisk JD, Bernstein CN, El-Gabalawy R, Hitchon CA, Kornelsen J, et al. The association between childhood maltreatment and pain catastrophizing in individuals with immune-mediated inflammatory diseases. J Psychosom Res. 2021;145:110479. [DOI] [PubMed]
    Eilam-Stock T, Links J, Khan NZ, Bacon TE, Zuniga G, Laing L, et al. Adverse childhood experiences predict reaction to multiple sclerosis diagnosis. Health Psychol Open. 2021;8:20551029211052830. [DOI] [PubMed] [PMC]
    Wan A, Bernstein CN, Graff LA, Patten SB, Sareen J, Fisk JD, et al.; CIHR Team in Defining the Burden and Managing the Effects of Immune-mediated Inflammatory Disease. Childhood maltreatment and psychiatric comorbidity in immune-mediated inflammatory disorders. Psychosom Med. 2022;84:109. [DOI] [PubMed]
    The effects of adverse childhood experiences and childhood socioeconomic disadvantage on adolescent depression and suicidality: a scoping review protocol [Internet]. Ann Arbor (MI): University of Michigan; c2024 [cited 2022 May 24]. Available from: https://deepblue.lib.umich.edu/handle/2027.42/172501
    Slavich GM, Shields GS. Assessing lifetime stress exposure using the Stress and Adversity Inventory for Adults (Adult STRAIN): an overview and initial validation. Psychosom Med. 2018;80:1727. [DOI] [PubMed] [PMC]
    Albott CS, Forbes MK, Anker JJ. Association of childhood adversity with differential susceptibility of transdiagnostic psychopathology to environmental stress in adulthood. JAMA Netw Open. 2018;1:e185354. [DOI] [PubMed] [PMC]
    Polick CS, Ploutz-Snyder R, Braley TJ, Connell CM, Stoddard SA. Associations among stressors across the lifespan, disability, and relapses in adults with multiple sclerosis. Brain Behav. 2023;13:e3073. [DOI] [PubMed] [PMC]
    von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: guidelines for reporting observational studies. Int J Surg. 2014;12:14959. [DOI] [PubMed]
    Amtmann D, Cook KF, Jensen MP, Chen WH, Choi S, Revicki D, et al. Development of a PROMIS item bank to measure pain interference. Pain. 2010;150:17382. [DOI] [PubMed] [PMC]
    Braley TJ, Whibley D, Alschuler KN, Ehde DM, Chervin RD, Clauw DJ, et al. Cannabinoid use among Americans with MS: current trends and gaps in knowledge. Mult Scler J Exp Transl Clin. 2020;6:2055217320959816. [DOI] [PubMed] [PMC]
    Cook KF, Bamer AM, Roddey TS, Kraft GH, Kim J, Amtmann D. A PROMIS fatigue short form for use by individuals who have multiple sclerosis. Qual Life Res. 2012;21:102130. [DOI] [PubMed] [PMC]
    Senders A, Hanes D, Bourdette D, Whitham R, Shinto L. Reducing survey burden: feasibility and validity of PROMIS measures in multiple sclerosis. Mult Scler. 2014;20:110211. [DOI] [PubMed] [PMC]
    Amezcua L, Rivera VM, Vazquez TC, Baezconde-Garbanati L, Langer-Gould A. Health disparities, inequities, and social determinants of health in multiple sclerosis and related disorders in the US: a review. JAMA Neurol. 2021;78:151524. [DOI] [PubMed]
    Amezcua L, McCauley JL. Race and ethnicity on MS presentation and disease course. Mult Scler. 2020;26:5617. [DOI] [PubMed] [PMC]
    Marrie RA, Leung S, Tyry T, Cutter GR, Fox R, Salter A. Use of eHealth and mHealth technology by persons with multiple sclerosis. Mult Scler Relat Disord. 2019;27:139. [DOI] [PubMed]
    Who gets MS? [Internet]. Waltham (MA): National Multiple Sclerosis Society; c2024 [cited 2023 Nov 6]. Available from: https://www.nationalmssociety.org/What-is-MS/Who-Gets-MS
    Cella D, Riley W, Stone A, Rothrock N, Reeve B, Yount S, et al.; PROMIS Cooperative Group. The Patient-Reported Outcomes Measurement Information System (PROMIS) developed and tested its first wave of adult self-reported health outcome item banks: 2005–2008. J Clin Epidemiol. 2010;63:117994. [DOI] [PubMed] [PMC]
    Jiang J, Abduljabbar S, Zhang C, Osier N. The relationship between stress and disease onset and relapse in multiple sclerosis: a systematic review. Mult Scler Relat Disord. 2022;67:104142. [DOI] [PubMed]
    Artemiadis AK, Anagnostouli MC, Alexopoulos EC. Stress as a risk factor for multiple sclerosis onset or relapse: a systematic review. Neuroepidemiology. 2011;36:10920. [DOI] [PubMed]
    Swanepoel I, van Staden W, Fletcher L. Psychological resilience and vulnerability as mediators between adverse life events and fatigue, motor dysfunction, and paresthesia in multiple sclerosis. Psychosom Med. 2020;82:13846. [DOI] [PubMed]
    Okai AF, Howard AM, Williams MJ, Brink JD, Chen C, Stuchiner TL, et al. Advancing care and outcomes for African American patients with multiple sclerosis. Neurology. 2022;98:101520. [DOI] [PubMed] [PMC]
    Simpson R, Simpson S, Ramparsad N, Lawrence M, Booth J, Mercer SW. Mindfulness-based interventions for mental well-being among people with multiple sclerosis: a systematic review and meta-analysis of randomised controlled trials. J Neurol Neurosurg Psychiatry. 2019;90:10518. [DOI] [PubMed]
    Braley TJ, Alschuler K, Ehde D, Little R, von Geldern G, Chervin RD, et al. A comparative effectiveness trial of telephone-delivered cognitive behavioral therapy, modafinil, and combination therapy of both interventions for fatigue in multiple sclerosis: primary results from the COMBO-MS trial. ACTRIMS 2023: Proceedings of the Americas Committee for Treatment and Research in Multiple Sclerosis (ACTRIMS) Forum; 2023 Feb 23–25; San Diego, CA, US. SAGE Publications; 2023. pp. 3–17.
    Kratz AL, Alschuler KN, Ehde DM, von Geldern G, Little R, Kulkarni S, et al. A randomized pragmatic trial of telephone-delivered cognitive behavioral-therapy, modafinil, and combination therapy of both for fatigue in multiple sclerosis: the design of the “COMBO-MS” trial. Contemp Clin Trials. 2019;84:105821. [DOI] [PubMed]
    Adlung L, Cohen Y, Mor U, Elinav E. Machine learning in clinical decision making. Med. 2021;2:64265. [DOI] [PubMed]
    Li T, Wang Z, Lu W, Zhang Q, Li D. Electronic health records based reinforcement learning for treatment optimizing. Inf Syst. 2022;104:101878. [DOI]
    Nag N, Yang X, Jelinek G, Neate S, Simpson-Yap S. Undertaking specific stress-reducing activities are associated with reduced fatigue and depression, and increased mastery, in people with multiple sclerosis. Mult Scler Relat Disord. 2022;62:103804. [DOI] [PubMed]
    Bradson ML, Cadden MH, Guty ET, Riegler KE, Thomas GA, Graham-Engeland JE, et al. Coping style moderates the effect of pain on depression symptoms in multiple sclerosis. Arch Clin Neuropsychol. 2022;37:151526. [DOI] [PubMed]
    Montazeri S, Aghaei A, Golparvar M. Mindfulness-integrated cognitive behavioral therapy: Does it effective on stress coping strategies in women with multiple sclerosis? Caspian J Intern Med. 2022;13:6348. [DOI] [PubMed] [PMC]
    Rodgers J, Friede T, Vonberg FW, Constantinescu CS, Coles A, Chataway J, et al. The impact of smoking cessation on multiple sclerosis disease progression. Brain. 2022;145:136878. [DOI] [PubMed] [PMC]
    Polick CS, Polick SR, Stoddard SA, Braley TJ, Slavich GM. The importance of assessing life stress exposure in multiple sclerosis: a case report. Mult Scler Relat Disord. 2021;54:103145. [DOI] [PubMed] [PMC]