Pain in long-term breast cancer survivors: the role of body mass index, physical activity, and sedentary behavior.

Participants in this study were women enrolled in the Health, Eating, Activity, and Lifestyle (HEAL) Study, a population-based, multicenter, multiethnic, prospective study of women newly diagnosed with stages 0–IIIA breast cancer. HEAL participants are being followed to determine the impact of weight, physical activity, diet, hormones, and other exposures on breast cancer prognosis.[ 43 ] Written or documented verbal informed consent was obtained from each participant for all assessments. All study protocols were approved by the Institutional Review Board of each participating center, in accordance with an assurance filed with and approved by the United States Department of Health and Human Services. Eligibility, recruitment, and retention of HEAL participants are described in detail elsewhere.[ 43 ] Briefly, patients diagnosed with their first primary breast cancer (N=1,183) were recruited from three Surveillance Epidemiology and End Results (SEER) registries in New Mexico, Western Washington, and Los Angeles County, California. In New Mexico, we recruited 615 women aged 18 years or older diagnosed with in situ to regional breast cancer between 1996 and 1999, living in Bernalillo, Santa Fe, Sandoval, Valencia, or Taos counties. In Western Washington, we recruited 202 women aged 40–64 years diagnosed with in situ to regional breast cancer between 1997 and 1998, living in King, Pierce, or Snohomish counties. The age range for the Washington patients was restricted to avoid overlap with eligibility requirements of other accruing studies. In Los Angeles County, we recruited 366 black women diagnosed with in situ to regional breast cancer between May 1995 and May 1998 who had participated in the Los Angeles portion of the Women’s Contraceptive and Reproductive Experiences (CARE) Study[ 44 , 51 ] or who had participated in a parallel case-control study of in situ breast cancer.[ 37 – 38 ] These two studies were limited to women ages 35–64 years. Data were drawn from 5 assessments ( Figure 1 ): (1) baseline/6 months post-diagnosis (in-person interview or self-administered questionnaire; stable demographics and clinical variables); (2) 30 months post-diagnosis (in-person interview or self-administered questionnaire; time-dependent demographics); (3) 40 months post-diagnosis (telephone or self-administered questionnaire; time-dependent demographics, pain); (4) 5 years post-diagnosis (telephone or self-administered questionnaire; BMI, physical activity, sedentary behavior); and (5) 10 years post-diagnosis (telephone or self-administered questionnaire; pain, BMI, physical activity, sedentary behavior). We used the Bodily Pain subscale of the Medical Outcomes Study short form 36 (SF-36),[ 22 , 68 ] which is commonly used to measure quality of life (QOL).[ 24 ] Considerable psychometric analyses have indicated high internal consistency for the pain subscale.[ 68 ] Pain scores are standardized on a T-score metric (overall U.S. general population mean=50, standard deviation=10); higher scores indicate less pain. Because pain interference increases with age,[ 64 ] we utilized age-stratified population norms.[ 68 ] In the absence of a case definition for pain in cancer survivors, we defined survivors with above-average pain as those with pain scores ≥½ of the age-specific standard deviation below the age-specific population mean. One-half standard deviation is considered the minimally important difference across QOL studies.[ 49 ] Based on a normal distribution, 30.9% of survivors would be expected to meet criteria for above-average pain. Change in pain (40-month to the 10-year assessment) was defined by two dimensions of pain: severity (whether pain was considered low versus above-average) and magnitude of change (≥½ standard deviation versus no change). Four distinct longitudinal patterns emerged: (1) maintained low pain (no pain change, low pain at both times), (2) pain improved (change with increased pain scores), (3) pain worsened (change with decreased pain scores), and (4) maintained above-average pain (no pain change, above-average pain at both times). Survivors with pain scores that fluctuated around the cut-point for above-average pain with no change in pain were not classified into a pain change group. Body mass index (BMI, kg/m 2 ) was computed from clinic-measured height (baseline) and self-reported weight (5-year and 10-year follow-up). Self-reported height was substituted for 144 participants missing clinic-measured height, including all participants from Los Angeles County for whom baseline clinic measures were not collected (r=0.94 between clinic-measured and self-reported height among participants with both measures). Change in BMI was also calculated from these height and weight values (increased ≥;5%, decreased ≥5%, maintained within <5%). Physical activity was evaluated (5 and 10 years post-diagnosis) using the Modifiable Activity Questionnaire, which has been shown to be reliable and valid.[ 28 , 54 ] This questionnaire assessed the type, duration, and frequency of activities performed in the prior year. Given observed associations with breast cancer prognosis and QOL,[ 1 , 23 ] we focused on moderate-vigorous sports/recreational activity. Hours/week spent in each activity were estimated by multiplying frequency by duration. A variable representing activity relative to current physical activity guidelines for US adults (≥150 minutes/week moderate activity or ≥75 minutes/week vigorous activity)[ 66 ] was created: no activity (0 minutes/week), activity not meeting guidelines (>0 and <150 minutes/week moderate activity) or meeting guidelines. Change in physical activity was defined by four distinct longitudinal patterns: (1) stayed inactive (not meeting physical activity guidelines at either assessment), (2) active to inactive (meeting guidelines at 5-year follow-up but not at 10-year follow-up), (3) inactive to active (not meeting guidelines at 5-year follow-up, meeting guidelines at 10-year follow-up), and (4) stayed active (meeting guidelines at both assessments). Sedentary behavior (5- and 10-year follow-ups) was operationally defined by television time(self-reported hours spent sitting while watching television/videos during a typical 24-hour period on weekdays and weekends) in the past year.[ 28 ] Similar measures of television watching have been used in public health research,[ 11 ] but we did not have direct evidence of validity or reliability of the questions asked in this study. We calculated continuous television time/day ((weekday median daily time spent watching television×5/7) + (weekend median time spent watching television×2/7)). In the absence of reliable groupings for television time associated with mortality or QOL,[ 18 ] we used a median split (high versus low television time). Change in television time was defined by four distinct longitudinal patterns: (1) maintained low television time (median at 5-year follow-up, <median at 10-year follow-up), (3) increased television time (median at 10-year follow-up), and (4) maintained high television time (>median at both assessments). Change patterns were based on the median at 5 and 10 years to identify those who changed relative to their peers (more than typical age-related changes). Baseline demographic and clinical characteristics included self-reported measures of education and race/ethnicity/study site (combined due to colinearity); furthermore, stage of disease, estrogen receptor and progesterone receptor status , b reast cancer treatment and number of nodes examined were abstracted from medical records and/or SEER. Age, lymphedema (ever versus never experienced), and general health status were self-reported at the 40-month and 10-year follow-ups. Menstruation (still menstruating versus stopped menstruating) and smoking status were self-reported at the 30-month and 10-year follow-ups. Comorbidity burden (0, 1, or ≥2 self-reported medical conditions that limited activities) was assessed in the 30-month follow-up only. Tamoxifen use (ever versus never) was assessed at 40 months. Because new chemoprevention agents became available, selective Estrogen Receptor Modulator use (ever versus never used tamoxifen, raloxifene, or toremifene) and a romatase inhibitor use (ever versus never) were self-reported at the 10-year follow-up. All analyses were conducted using Stata 11.2. Changes in mean pain severity and in the proportion of women reporting above-average pain were tested using a paired-samples t-test and McNemar’s test, respectively. We tested associations between BMI, physical activity, and sedentary behavior with pain cross-sectionally at 10 years two ways. First, linear regression models were fit with continuous pain scores as the dependent variable and BMI, physical activity, and sedentary behavior each individually fit as the independent variable. Next, BMI, physical activity, and sedentary behavior were modeled simultaneously. Second, logistic regression models were fit with the same independent variables but using the dichotomous above-average pain variable as the dependent variable. We evaluated associations between changes in BMI, physical activity, and sedentary behavior (5 to 10 years post-diagnosis) and pain at 10 years with pain treated both continuously and dichotimously. These analyses were conducted using linear regression and logistic regression approaches with BMI, physical activity, and sedentary behavior each fit separately and with a combined model that included all three factors. For each model described above, potential confounders (age, education, race/ethnicity, hormone receptor status, breast cancer stage, cancer treatment type, nodes examined, time since cancer diagnosis, SERM use, AI use, and smoking status) were modeled using backward elimination, with a ≥10% change in the coefficient for BMI, physical activity, or sedentary behavior indicating confounding.[ 45 , 65 ] We adjusted for comorbidities after identifying other confounders because of probable reciprocal associations between BMI, physical activity, and sedentary behavior with comorbidities. We tested associations of changes in BMI, physical activity, and sedentary behavior with changes in pain. Because we anticipated small cell sizes, we planned separate unadjusted logistic regression models comparing those who increased versus decreased BMI, physical activity, or sedentary behavior on likelihood of being classified in each pain change group (maintained above-average pain, pain improved, pain worsened) in separate models relative to maintaining low pain.

Most HEAL participants were Non-Hispanic White (61%)( Table 1 ). The average age was 58.9 years at the 40-month assessment. More than half of the women had locally–staged breast cancer; approximately 22% had in situ cancers. Most women were treated with surgery and radiation (37%) or surgery, radiation, and chemotherapy (22%). The sample available for analyses at 40-months, 10 years, and changes in pain includes 68%, 48%, and 44% of women who completed the baseline assessment, respectively ( Figure 1 ). Compared to the entire HEAL sample, survivors who completed sufficient follow-up assessments to be included in any of the current analyses were younger, achieved higher education, had less advanced cancer at diagnosis, and were less likely to be ER- or PR- (all p0.05). Mean pain scores and above-average pain stratified by age are shown in Table 2 . At 40 months and 10 years, the overall mean pain score was 50.1 (sd=10.8) and 48.0 (sd=11.0), respectively. The overall proportion of survivors reporting above-average pain was 27.8% and 32.3%, respectively. Mean pain scores worsened over time ( t =3.99, p =0.0001). The proportion of women reporting above-average pain was higher at 10 years (χ 2 (1)= 5.30, p=0.02) . From 40 months to 10 years post-diagnosis, almost 10% of survivors maintained above-average pain, nearly 25% reported improved pain and another one-third reported worsened pain ( Figure 2 ). There were no associations between pain change groups and demographic/clinical variables except that women who maintained above-average pain reported higher comorbidity burden (p<0.05). In individual models, women who were overweight or obese reported significantly higher pain (lower pain scores) than normal weight survivors (p<0.05, Table 3 ). High television time was also associated with higher pain (p<0.05). In the combined model, the effects of overweight and obese BMI remained significant, while the effect of television time was attenuated. In individual models, obese women were more likely to report above-average pain than normal weight survivors (p<0.05, Table 3 ). Survivors meeting physical activity guidelines were approximately half as likely as inactive women to report above-average pain (p<0.01). In the combined model, only the effect of meeting physical activity guidelines remained significant. Always active survivors reported lower pain than always inactive survivors in the individual and combined models (p5% BMI were more likely, and survivors who were always active were less likely, to report above-average pain (all p<0.05, Table 4 ). Both associations were similar in the combined model. Additional adjustment for comorbidities attenuated some associations: the cross-sectional associations with pain scores at 10 years for overweight BMI (β=−1.66, p=0.13) and television time (β=−1.03, p= 0.26) in individual models and for overweight BMI in the combined model (β=−1.73, p=0.12), the cross-sectional association between obese BMI and above-average pain at 10 years (OR=1.56, 95% CI=0.96–2.56, p=0.08), and the association between being ‘always active’ and pain scores in the combined model (β=2.04, p=0.14). Survivors who decreased BMI were less likely than those who increased BMI to be in the pain worsened group (25.6% versus 39.8%; OR, 95% CI: 0.45, 0.23–0.88, p=0.02). No other associations were observed between changes in BMI, physical activity, or sedentary behavior with pain change groups (data not shown, all p>0.05).

Pain is common among post-treatment breast cancer survivors, even many years post-treatment.[ 7 , 12 , 17 , 19 , 25 , 36 , 57 ] Prevalence estimates vary widely: 9–72% of breast cancer survivors report pain[ 14 ] and 12–29% of those ≥5 years post-diagnosis report pain attributed to cancer.[ 12 , 55 ] Pain is associated with numerous negative sequelae in cancer survivors, [ 16 , 19 , 25 , 31 ] Increased knowledge about pain and strategies for prevention or mitigation among survivors, given their risk for multiple comorbid conditions, physical and psychosocial complaints, and functional limitations[ 59 , 63 ], can contribute to optimal functioning after breast cancer. The course of pain over time and factors affecting pain in specific cancer populations are not well understood.[ 41 ] Longitudinal studies are few in number and focus on average pain severity without defining the vulnerable subgroup of survivors who experience pain. Research suggests that pain resolves for only certain survivors[ 36 , 57 ].[ 36 ] More research is needed that examines the trajectories of pain over time for individuals. Further, a case definition for clinically meaningful pain would facilitate research synthesis.[ 2 ] Documentation of modifiable factors associated with pain in post-treatment survivors may identify survivors at risk for poor outcomes and interventions to mitigate pain. Body mass index (BMI)[ 8 , 30 , 46 , 53 , 58 , 62 , 69 ] and physical activity[ 29 , 46 , 56 ] have been associated with pain in the general population. Furthermore, physical activity[ 1 , 4 , 6 , 20 , 32 , 42 , 52 , 60 ] and, to a lesser extent, BMI[ 47 , 52 ] have been associated with pain in survivors specifically. Research examining associations between sedentary behavior and pain have been mixed,[ 10 ] but few studies define sedentary behavior as the amount of waking time spent sitting/reclining rather than the absence of physical activity. A better understanding of the unique effects of BMI, physical activity, and sedentary behavior on pain in long-term survivors is needed. This study examined the course and predictors of pain in long-term breast cancer survivors using a multicenter, multiethnic, prospective cohort. Specifically, we aimed to (1) describe pain at 40 months and 10 years post-diagnosis by examining mean pain scores, identifying survivors with above-average pain (used as a proxy for a case definition of pain), and comparing observed pain scores to population norms; (2) identify patterns of change in pain over time; and (3) test associations between BMI, physical activity, sedentary behavior with pain in three different ways: (a) testing associations cross-sectionally at 10-years post-diagnosis, (b) testing associations between changes in BMI, physical activity, and sedentary behavior (5 to 10 years post-diagnosis) and pain at 10 years post-diagnosis, and (c) testing associations between changes in BMI, physical activity, and sedentary behavior with changes in pain (40 months to 10 years post-diagnosis). We hypothesized that BMI and sedentary behavior would be positively associated with pain, and that physical activity would be inversely associated with pain, cross-sectionally and longitudinally.

Approximately one-third of survivors reported above-average pain relative to their same aged-peers at both 40 months and 10 years post-diagnosis, which is consistent with expectations based on population norms.[ 68 ] Other research has suggested that pain is more common or severe among survivors than the general population,[ 3 , 55 ] but these studies measured pain continuously or as present versus absent. More research is needed to establish a reliable, valid, and clinically relevant case definition of pain to determine whether pain is different in survivors compared to the general population. Nonetheless, examination of survivors with above-average pain is a first step toward identifying the subgroup most in need of intervention. Nearly 10% of women maintained above average pain, approximately one-third reported increased pain, and the likelihood of above-average pain was greater further out from cancer diagnosis. These results suggest that pain is a common and perhaps an increasingly important complaint many years post-diagnosis. Consistent with observational studies in the general population[ 8 – 9 , 30 , 46 , 53 , 58 , 62 , 69 ] and in cancer survivors,[ 47 ] sxcess weight was uniquely associated with higher pain 10 years after breast cancer; moreover, weight gain was positively associated with above-average pain and worsened pain over time. Several studies failed to find cross-sectional associations between BMI and pain among cancer survivors[ 5 , 21 , 36 , 55 ], again perhaps due to a focus on the presence versus absence of pain[ 36 , 55 ] or on average pain severity[ 5 , 21 ] rather than a case definition of meaningful pain. Additionally, building on a previous HEAL study demonstrating associations between physical activity and pain up to 40 months post-diagnosis,[ 1 ] meeting physical activity guidelines at 10 years post-diagnosis and consistently over time was uniquely associated with a lower likelihood of above-average pain. Although this prospective observational study cannot fully determine whether changes in physical activity caused or were a consequence of reduced pain, intervention studies show that short-term physical activity programs reduce pain in cancer survivors.[ 6 , 32 , 42 ] Taken together, these results document that BMI and physical activity are independently associated with pain in breast cancer survivors, both acutely and prospectively, over the course of long-term survivorship. Consistent with a previous cross-sectional HEAL analysis on sedentary behavior and QOL 3.5 years post-diagnosis,[ 18 ] sedentary behavior was not uniquely associated cross-sectionally or longitudinally with pain in this sample. The attenuation of the cross-sectional association between television time and pain after adjustment for BMI and physical activity suggests sedentary behavior may not be uniquely associated with pain among breast cancer survivors. Alternatively, the self-report nature of our television watching measure or the definitions we chose for change in sedentary behavior may have resulted in exposure misclassification. Even if sedentary behavior is not a predictor of pain, sedentary behavior is common[ 39 ] and associated with obesity among breast cancer survivors[ 34 ] and with biomarkers of postmenopausal breast cancer risk[ 35 ] and mortality[ 13 , 15 , 26 , 40 , 50 , 61 , 67 ] among adults without cancer. Additional studies of sedentary behavior and QOL in cancer survivors with validated, comprehensive self-report measures of sedentary behavior, complemented by objective monitoring of sedentary behavior where feasible, would be helpful towards understanding the role of sedentary behavior and health of cancer survivors. Strengths of this study include a large group of breast cancer survivors recruited through population-based cancer registries, a prospective study design with follow-up through 10 years post-diagnosis, repeat measures of pain, examination of above-average pain and pain change patterns, and assessment of the unique role of BMI, physical activity, and sedentary behavior. However, several limitations should be considered. Data on BMI, physical activity, and sedentary behavior were self-reported using multiple modes of administration of unknown comparability. Participants may have difficultly accurately recalling activities over the past year. Data for total sedentary time were not available, thus time spent on computers or in sedentary employment was not captured. Our operational definitions for above-average pain, high television time, and changes in BMI, physical activity, and sedentary behavior may be mis-specified. Due to sample size constraints, we were unable to examine a more stringent case definition for pain. Data were not available at 10 years to control for all possible confounders (e.g., pain medication use) or fully assess the role of comorbidities as a confounder or a moderator. We cannot examine the course of pain prior to 40 months post-diagnosis, thus pain cannot be attributed to cancer, and the extent to which changes in BMI and physical activity were a cause or a consequence of pain is unclear. Small cell sizes likely limited power for analyses associating changes in BMI, physical activity, and sedentary behavior with changes in pain. Selection bias due to differential attrition may reduce the generalizability of our findings. Finally, given racial/ethnic differences in associations between obesity, physical activity, and pain in breast cancer survivors,[ 33 , 52 ] future research should examine pain in more diverse samples. This study suggests that pain remains or becomes a significant complaint among some long-term breast cancer survivors and that excess weight and lack of physical activity place survivors at continuing risk for pain long after cancer treatment ends. Assessment and treatment of pain are already considered standard for quality care during cancer treatment,[ 48 ] but they should continue to be a routine part of long-term survivorship care. Future research should determine the directionality and causal mechanisms for associations of BMI and physical activity with pain, as well as examine the unique role of BMI and physical activity in symptom clusters (e.g., pain, depression, fatigue, sleep disturbances)[ 27 ]. Associations between BMI and pain are strongest among older women[ 62 ]; thus, if weight control and physical activity interventions are supported by randomized controlled trials, these interventions may be particularly relevant for preventing or mitigating pain among older long-term breast cancer survivors.