Worst Pain Severity Profiles of Oncology Patients Are Associated with Significant Stress and Multiple Co-Occurring Symptoms

Patients and settings

This study is part of a larger, longitudinal study of the symptom experience of oncology outpatients receiving chemotherapy.⁷⁰ 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. The major reason for refusal was being overwhelmed with their cancer treatment.

Study procedures

The study was approved by the Institutional Review Board at each of the study sites. Of the 2234 patients approached, 1343 consented to participate (60.1% response rate) and 1305 rated their worst pain intensity six times over two chemotherapy cycles (i.e., prior to chemotherapy administration (assessments 1 and 4), approximately 1 week after chemotherapy administration (assessments 2 and 5), and approximately 2 weeks after chemotherapy administration (assessments 3 and 6); Figure 1). The remaining questionnaires described in this paper were completed at enrollment (i.e., prior to their second or third cycle of chemotherapy), Medical records were reviewed for disease and treatment information.

Instruments

Data analysis

Descriptive statistics and frequency distributions were generated for sample characteristics at enrollment using the Statistical Package for the Social Sciences (SPSS) version 27.⁴⁷ As was done for other symptoms,^{4, 53, 75, 107, 110, 115, 116} LPA was used to identify unobserved subgroups of patients (i.e., latent classes) with distinct worst pain profiles over the six assessments, using the patients’ ratings of worst pain severity. Before performing the LPA, patients who reported the occurrence of pain for ≤1 of the six assessments were identified and labeled as the “None” class (n=371, 28.4%). Then, the LPA was performed on the remaining 934 patients using MPlus^™ Version 8.4.⁸⁰

Typically, for longitudinal data, growth mixture modeling (GMM) or latent class growth modeling (LCGM) of change trajectories would be used to identify latent classes of individuals who change differently over time. However, the current data demonstrated a complex pattern of change because the assessments included a pre-treatment assessment, an immediate post-treatment assessment, and a second post-treatment assessment -- for two cycles of treatment. One would expect the trajectory of change for pain scores over two treatment cycles, measured over only six occasions, to have a pattern that looks like two inverted “V’s”. GMM is ideally suited to discover change patterns that differ among latent classes such as linear change, or linear and quadratic change, when the number of assessments is as small as six. More complex patterns require more assessments for reliable identification of latent classes with different change trajectories.

Therefore, we identified latent classes of patients based on their profiles of means, where the means were estimated from the same symptoms (i.e., worst pain) measured on six occasions. In order to incorporate the expected correlations among the repeated measures, we included covariances among measures that were one or two occasions apart (a covariance structure with a lag of two). In this way, we retained the within person correlation among the measures, at the same time that we focused on the patterns of means that distinguished the latent classes. We limited the covariance structure to a lag of two to accommodate the expected reduction in correlation that would be introduced by two treatments within each set of three measurement occasions, and to reduce model complexity.

Estimation was carried out with full information maximum likelihood with standard error and a chi-square test that are robust to non-normality and non-independence of observations (“estimator=MLR”). Model fit was evaluated to identify the solution that best characterized the observed latent class structure with the Bayesian Information Criterion (BIC), Vuong-Lo-Mendell-Rubin likelihood ratio test (VLRM), entropy, and latent class percentages that were large enough to be reliable.⁷⁹ Missing data were accommodated for with the use of the Expectation-Maximization (EM) algorithm.⁷⁸ This algorithm is appropriate because the data in this study met the assumption of “missingness at random.

Differences among the latent classes in demographic and clinical characteristics, stress and resilience measures, and symptom severity scores at enrollment were evaluated using analysis of variance, Kruskal-Wallis or Chi Square tests. A p-value of <.05 was considered statistically significant. Post hoc contrasts were done using a Bonferroni corrected p-value of <.008 (.05/6 possible pairwise comparisons).

Latent profile analysis

The 371 patients (28.4%) who had ≤1 occurrence of pain over the six assessments were classified as the None class. For the remaining 934 patients whose data were entered into the LPA, a three-class solution was selected because the 3-class solution fit the data better than the 2-class solution (Table 1). The BIC for the 3-class solution was lower than the BIC for the 2-class solution. In addition, the VLMR was significant for the 3-class solution. Although the BIC was smaller for the 4-class than for the 3-class solution, the VLMR for 4-classes was not significant. Figure 2 displays the trajectories of worst pain profiles for the four classes. The latent classes were named based on clinically meaningful cutoff scores (1–3 = Mild, 4–6 = Moderate, ≥7 = Severe).⁹⁶ Of the 1305 patients in this study, 28.4% did not report pain (None), 12.5% were in the Mild class, 28.6% in the Moderate class, and 30.5% in the Severe class.

Differences in demographic and clinical characteristics at enrollment

Compared to the other three classes, patients in the Severe class had fewer years of education, a lower annual household income, were more likely to be unemployed, had a higher number of comorbid conditions, a higher SCQ score, a lower functional status, and were more likely to self-report diagnoses of anemia and back pain. Compared to the None and Moderate classes, patients in the Severe class had a higher BMI, were less likely to exercise in a regular basis, and were more likely to self-report diagnoses of ulcer or stomach disease and depression. Compared to the None class, patients in the Moderate and Severe classes reported a higher number of previous cancer treatments and were more likely to self-report a diagnosis of rheumatoid arthritis (Table 2).

Differences in stress and resilience measures at enrollment

Compared to the other three classes, patients in the Severe class reported higher PSS, intrusion, avoidance, hyperarousal, and total IES-R, as well as LSC-R affected sum scores. Compared to the None class, patients in the Moderate class reported higher PSS and IES-R intrusion and hyperarousal scores. Compared to the None class, patients in the other three classes reported higher LSC-R PTSD sum and total scores. Compared to the None class, patients in the Severe pain reported lower CDRS scores (Table 3).

Differences in symptom scores at enrollment

Compared to the other three classes, patients in the Severe class reported higher levels of depressive symptoms, trait anxiety, state anxiety, morning fatigue, evening fatigue, sleep disturbance, and cognitive dysfunction. Compared to the None class, patients in the Moderate class reported higher levels of depressive symptoms, trait anxiety, state anxiety, morning fatigue, sleep disturbance, cognitive dysfunction, and decrements in morning energy (Table 4).

Demographic characteristics associated with worst pain profiles

Table 5 provides a summary of the common and distinct characteristics associated with the Low, Moderate, and Severe pain classes, compared to the None class. The remainder of this discussion places these comparisons within the context of the extant literature. Patients in the Severe class were more likely to be female, single, unemployed, had a lower annual household income, and were less likely to exercise on a regular basis. While findings regarding gender differences in pain occurrence and intensity among oncology patients are inconsistent,^{1, 19, 34, 65} a growing body of preclinical evidence suggests that sexual dimorphism exists in the development of chemotherapy-induced peripheral neuropathy with female rats more likely to develop this adverse effect.^{37, 101} Equally important, several chronic pain conditions (e.g., osteoarthritis⁸⁵ rheumatoid arthritis⁷⁶) that occurred more frequently in our sample are more common in women. In addition, in one study of oncology outpatients,¹⁸ women reported more hesitancy in using analgesics and had lower levels of adherence with their analgesic regimen. These findings suggest that female oncology patients are a high risk group that warrant careful assessments of both cancer and non-cancer related pain.

As noted previously,^{15, 16} ongoing cancer treatments and unrelieved pain interfere with patients’ employment and contribute to financial toxicity. In addition, chronic pain conditions (e.g., back pain) may result in significant disability which interferes with one’s ability to work and remain employed. Finally, patients who are single may lack social support which is known to amplify patients’ pain experiences.⁴³ These factors are likely to add to the stress that patients in the Severe class were experiencing.

Clinical characteristics associated with worst pain profiles

Compared to the None class, patients in the Moderate and Severe classes reported a higher number of comorbidities, a higher comorbidity burden, a higher number of previous cancer treatments, and higher occurrence rates for osteoarthritis, back pain, and rheumatoid arthritis. Not surprisingly, these patients reported not only statistically significant but clinically meaningful decrements in functional status (Cohen’s d = 0.36 for the Moderate and 0.86 for the Severe classes).²⁶ Taken together and supported by previous findings from the general population,⁸⁶ multimorbidity is associated with more severe pain and significant decrements in functional performance. In addition, the receipt of chemotherapy may contribute to the development of comorbid conditions or make existing comorbidities worse.^{40, 94} These findings suggest that oncology clinicians need to work with primary care providers to achieve the optimal management of the patient’s cancer and co-occurring conditions, as well as their pain.

A higher BMI and lack of regular exercise were two modifiable characteristics associated with membership in the Severe pain class. While findings regarding the relationship between BMI and pain are inconsistent,^{35, 106, 109} it is reasonable to hypothesize that in addition to unrelieved pain, a mean BMI in the overweight range; the higher occurrence rates for a number of painful comorbid conditions; and the lack of regular exercise are associated with a lower function status in the patients in the Severe class. In addition to pain management interventions, these patients warrant referrals to dieticians and physical therapists to improve their level of physical functioning.

Stress and resilience characteristics associated with worst pain profiles

While patients in the Moderate and Severe classes had significantly higher general stress scores than those in the None class, all four classes had PSS scores that were above the clinically meaningful cutpoint. This finding is consistent with previous studies that noted that a cancer diagnosis and associated treatments are stressful experiences for most patients.³

In addition, compared to the None class, patients in the other three classes had significantly higher levels of disease-specific stress and cumulative life stress. While the mean IES-R total score in the Severe class did not exceed the clinically meaningful cutpoint, 42% of the patients in this class had scores above the cutpoint of 24 (i.e., partial PTSD) and 17% met the criteria for PTSD. While the LSC-R has not been used to assess cumulative life stress in oncology patients, the total score for the LSC-R is comparable to a community-based samples of Columbian (7.2 ± 3.8)⁴⁶ and Mexican (9.5 ± 4.2)⁸² women and individuals who were dependent on prescription opioids (7.7 ± 0.6).⁵⁹ Our findings are consistent with previous reports that demonstrated positive associations between chronic pain and the extent of lifetime trauma exposures.^{45, 58} As noted in two reviews,^{17, 33} stress can decrease pain thresholds due to a heavy allostatic load and dysregulation of the neuroendocrine stress axes. In addition, early life stress is known to alter the responsiveness of the hypothalamic pituitary axis, as well as the functioning of the opioidergic, monoaminergic, endocannabinoid, and immune systems.¹⁰

In terms of resilience, while none of the latent classes had CDRS scores above the normative adult mean score in the United States of 31.8 (±5.4),^{11, 12} their scores were consistent with a previous study of oncology patients⁶⁸ and higher than those reported by primary care patients with depressive symptoms.⁸⁷ While a behavioral response to stress, recent evidence suggests that a large amount of inter-individual variability exists in resilience. In addition, the active and unique biological mechanisms associated with resilience appear to buffer an individual’s response to stress, not simply reverse pathophysiologic mechanisms.^{14, 36, 91}

Multiple co-occurring symptoms associated with worst pain profiles

In terms of multiple co-occurring symptoms, it should be noted that all four classes reported severity ratings for decrements in morning and evening energy, as well as sleep disturbance that were above the clinically meaningful cutoff scores. Given that between 30% to 88% of oncology patient report sleep disturbance during treatment,⁹⁵ it is not surprising that these symptoms clustered together in our sample regardless of pain class.

Compared to the None class, patients in the Moderate and Severe classes reported a higher overall symptom burden (i.e., higher scores for depressive symptoms, state and trait anxiety, morning fatigue and sleep disturbance, as well as significant decrements in morning energy and cognitive function; Table 5). These findings suggest that dynamic and synergistic interactions occur between pain and multiple co-occurring symptoms. This hypothesis is consistent with a previous review that suggested that patients with multiple co-occurring symptoms experience more severe pain due to the complex interactions among a variety of physiological, psychological, behavioral, and sociocultural factors.⁸³.

Perhaps most important, the Moderate and Severe classes reported higher levels of all three types of stress, in addition to multiple co-occurring symptoms. While these initial findings warrant confirmation, a growing body of evidence suggests that stress, pain, and other common symptoms associated with cancer and its treatment may share common underlying mechanisms including alterations in inflammatory cascades,^{54, 55} disruptions in the hypothalamic-pituitary-adrenal axis,⁶⁴ and alterations in the gut brain axis.^{92, 103} This hypothesis warrants investigation in future studies.