Pain in Long-Term Adult Survivors of Childhood Cancers and Their Siblings: A Report from the Childhood Cancer Survivor Study

Abstract

Little is known about pain among long-term adult survivors of childhood cancers. The study investigated pain prevalence in this population compared with sibling controls and examined pain-related risk factors. Three self-reported pain outcomes including pain conditions, prescription analgesics used, and pain attributed to cancer and treatment were assessed among 10,397 cancer survivors and 3,034 sibling controls from the Childhood Cancer Survivor Study (CCSS). Pain conditions (pain/abnormal sensation, migraines, and other headaches) were reported by 12.3%, 15.5%, and 20.5% of survivors respectively; 16.7% of survivors reported use of prescription analgesics, and 21% attributed pain to cancer and treatment. Risks of reporting pain conditions and using prescription analgesics were higher among survivors than siblings adjusting for sociodemographic factors. Younger age at diagnosis and a history of non-Hodgkin lymphoma, Wilms tumor, or neuroblastoma (compared to leukemia) were associated with greater risk of reporting pain conditions. A history of bone cancer or soft tissue sarcoma (compared to leukemia) was associated with greater risks of using prescription analgesics and cancer-related pain attribution. Non-brain directed scatter irradiation was associated with elevated risk for migraines and cancer-related pain attribution. Female gender and lower educational attainment were associated with increased reports of all three pain outcomes; minority status, unemployment, and being single were associated with greater risks for reporting pain conditions. These findings contribute to the understanding of pain and associated risk factors among adult survivors of childhood cancer and suggest areas of focus for pain intervention.

Introduction

Approximately 80% of children with cancer survive more than five years from diagnosis of their disease [31]. The significant improvements in survival rates for childhood cancers [31] have led to an increased focus on identification of late effects of treatment among long-term survivors. A wide array of medical and psychosocial late effects has been documented among adult survivors of childhood cancer such as poor physical health behaviors [25], risk for chronic illness [8], fatigue[22], cognitive and psychological impairment [23, 43], and negative social outcomes [17]. However, little is known about the prevalence of pain and associated risk factors in this population.

Several studies have estimated the prevalence of pain to range from 9–26% for outpatients and 39–54% for inpatients among children during treatment or follow-up for a variety of cancer diagnoses [12, 14, 24]. These estimates suggest a frequent symptom occurrence for these patients; however, the prevalence of pain among long-term childhood cancer survivors is unclear. Studies with adult survivors of childhood cancer at 10 years [35] and 14 years post diagnosis [22] revealed that 20–30% of survivors reported pain. However, it is difficult to place the prevalence in perspective because of the limited sample size (Ns < 200). In addition, because point prevalence of pain in the general adult population varies widely [6, 7, 15, 30, 36], appropriate comparison groups are needed in order to understand whether having cancer as a child places individuals at greater risk of experiencing pain during adulthood. Furthermore, risk factors associated with pain are largely unexamined for this population.

Among adult cancer survivors, the estimated prevalence of pain is 50–75% during cancer treatment [5] and 21–41% at one to two years post-treatment [18]. Although pain prevalence data exist in adult cancer survivors who have recently completed treatment, only a few large-scale studies investigated pain in long-term cancer survivors and these studies mostly focused on breast cancer survivors [26].

The Childhood Cancer Survivor Study (CCSS) provides a unique opportunity to contribute to our understanding of pain outcomes among long-term survivors of childhood cancer because of its large sample size, comprehensive characterization of the cohort, and the availability of a sibling comparison group. By utilizing existing data obtained from the CCSS, we aimed 1) to investigate the prevalence of pain among survivors compared with that among sibling controls, and 2) to understand how sociodemographic and treatment related risk factors were associated with pain outcomes. We expected that survivors would have higher rates of pain compared with sibling controls. Studies among adult populations suggest that risk factors for chronic pain include gender [13, 40], lower socioeconomic status [9], minority status [2], and unemployment [4]. Accordingly, we hypothesized that female gender, ethnic minority, lower SES, younger age at diagnosis, and treatments (surgery, radiotherapy, chemotherapy) would be associated with higher risk of reporting pain.

Methods

Participants and procedures

The CCSS was established in 1993 through funding from the National Cancer Institute as a large cohort of survivors of childhood cancer for use as a research resource. The CCSS represents the largest and most comprehensively characterized cohort of childhood cancer survivors ever assembled in North America. A detailed description of the CCSS study design, methods and cohort characteristics is provided elsewhere [32, 33].

All survivors included in CCSS fulfilled the following eligibility criteria: (a) diagnosis of leukemia, CNS malignancies (all histologies), Hodgkin lymphoma, non-Hodgkin lymphoma, Wilms tumor, neuroblastoma, soft tissue sarcoma, or malignant bone tumor; (b) diagnosis and initial treatment at one of the 26 collaborating CCSS institutions; (c) diagnosis date between January 1, 1970 and December 31, 1986; (d) age less than 21 years at the time of diagnosis; and (e) survival of at least five years from the time of diagnosis. The CCSS was approved by the Institutional Review Boards of all participating institutions. All participants provided informed consent for their participation and medical record abstraction.

The participating sample in the current study was derived from a group of 20,720 individuals meeting the eligibility criteria above. Of these individuals identified by the collaborating institutions, 3,017 (14.6%) could not be located and were considered lost for follow-up. Among the 17,703 subjects located, 14,372 (81.1%) consented to participate and completed a self-report questionnaire. A randomly selected subset of 5,000 survivors was asked to nominate their closest age sibling [33]. The random selection of such a large number of siblings should adequately represent the larger population of siblings and create a control group of non-cancer survivors with similar demographic and socioeconomic characteristics. Survivors nominated 4,782 siblings who were invited for participation, of which 3,846 (80%) completed the survey. To minimize recall biases, the current analyses focuses on those survivors (n=10,397) and siblings (n= 3,034) who were 18 years or older at the time of the survey.

Measures

Demographic and medical characteristics

Medical record abstraction, according to a structured protocol [33], was conducted at each CCSS center and included detailed information about cancer type, treatments received, and clinical characteristics of the survivor. Data on cancer therapy was abstracted from medical records and did not require patient recall of the experience. A 24-page questionnaire, completed by the survivors and siblings, provided information on demographics, personal and family medical history, functional limitations, psychological status, work history, and living circumstances. Study questionnaires can be viewed at http://www.ccss.stjude.org/. The demographic variables analyzed in the study were gender, minority status, annual household income, educational attainment, marital status, and work status. Treatment and cancer related variables examined were treatment type (surgery, chemotherapy, and radiotherapy), diagnostic category, and age at diagnosis.

Specific pain conditions

On the questionnaire, participants were asked whether they had “ever been told by a doctor or other health care professional” that they “have or have had” any of the following pain conditions: “prolonged pain or abnormal sensation in arms, legs, or back,” “migraine,” or “other frequent headaches.” Participants were given response options of “yes,” “no,” or “not sure.” To minimize potential recall bias and to be conservative in pain estimations, “not sure” responses on these items were considered as a lack of pain and were collapsed together with the “no” responses. Participants who responded “yes” to having been diagnosed with one of these three pain conditions were then asked to indicate the age at which the pain condition first occurred and these ages were utilized in time-to-event analyses for pain outcomes.

Use of prescription medications for pain

All participants were asked whether they used specific types of prescription pain medications either consistently for more than one month or for at least 30 days in one year during the two-year period preceding the administration of the questionnaire. Examples of types of prescription pain medications were provided. Response options included “yes,” “no,” and “not sure.” To minimize potential recall biases and to be conservative in pain estimations, responses of “not sure” were combined with “no” responses.

Current Pain attributed to cancer or its treatment

Survivors were also asked whether they “currently have pain as a result of [their] cancer, leukemia, tumor or similar illness, or its treatment?” Participants were given the option to rate current pain on a continuum of “no pain,” “small amount of pain,” “medium amount of pain,” “a lot of pain,” or “very bad excruciating pain”.

Data Analysis

Survivor/sibling comparisons in the overall risk of being diagnosed with one of the three pain conditions were evaluated using Cox proportional hazards models, where age was used as the time scale, adjusting for gender, race/ethnicity, marital status, highest level of education, and annual household income. Hazard Ratios (HR), 95% confidence intervals (CI) and p-values are reported. These analyses accounted for within-family correlations using generalized estimating equations with sandwich standard error estimates [39]. Two types of within-survivor models were employed, one adjusted for primary diagnosis and another for treatment with cranial radiation, both of which adjusted for the same demographic covariates listed for the Cox model above as well as age at diagnosis. For the three pain conditions, pain/abnormal sensation, migraines, and other frequent headaches, multiple imputation procedures were employed to impute missing values for age of onset for those participants who reported the presence of a pain condition but did not report the age at which the condition was first diagnosed. The method of multiple imputation was based on that described by Taylor et al. [38] and used an expectation-maximization algorithm to generate ten imputed missing values for each missing age at first occurrence value. The final reported HR, 95% CI and p-values were estimated based on the model for each pain condition and the standard formula of the multiple-imputation inference [34]. Of the 10,397 survivors who completed a survey when they were age 18 years or older, a total of 8,905 had complete treatment data and were included in the Cox regression analyses. Among these survivors, age at first occurrence was imputed for 7.6% (43 of 563), 7.4% (53 of 721), and 3.7% (29 of 780) for pain/abnormal sensation, migraines and other frequent headaches, respectively. As for the siblings age 18 years or older, the imputation models were created from the whole cohort of 3,034. Among the siblings, age at first occurrence was imputed for 9.5% (18 of 190), 9.3% (43 of 461), and 16.5% (76 of 461) for pain/abnormal sensation, migraines and other frequent headaches, respectively.

Within-survivor analyses of the binary variable “use of prescribed pain medications” were conducted using logistic regression analyses. Within-survivor analyses of pain attributed to cancer or its treatment were conducted using a partial proportional odds model where each possible dichotomization of the outcome was considered so that odds ratios could be simply interpreted as odds of being lower or higher across the entire range of the outcome. Specific comparisons were made between 1) those reporting a small amount of pain and above vs. no pain, 2) a medium amount of pain and above vs. a small amount of pain and no pain, and 3) a lot of pain or very bad pain vs. a medium amount of pain and below. Both of these within-survivor analyses controlled for the same demographic and cancer related covariates used in the previous Cox proportional hazards models. Two-sided p-values are reported and statistical significance for all models was set at p < 0.05. All analyses were completed using SAS version 9.1.3 (SAS Institute, Cary, NC).

Results

Characteristics of the Sample

Survivors were largely Caucasian (84.0%), middle-income (52.0% reported annual household incomes between $20,000 and $59,999), and well-educated (65.8% reported having attended at least some college). On average, survivors were 16.5 years post diagnosis (SD=4.9, and range = 5 – 31). Important demographic differences between the survivors and siblings were noted (Table 1). Relative to the sibling controls, a significantly higher proportion of survivors were male and younger. Additionally, survivors were less likely than siblings to describe their ethnic identity as Caucasian, to report having an annual total household income in the highest income ranges, to be married, to have graduated from college, to have ever had a job, or to have worked within the past year.

Table 1.

Demographic Characteristics of Participating Childhood Cancer Survivors and Siblings ≥ 18 years of age

Variable	Value	% Survivors (n)	% Siblings (n)	p-value
Sex	Male	5593(53.8%)	1429(47.1%)	<0.001
	Female	4804(46.2%)	1605(52.9%)
Ethnicity	Caucasian	8708(84.0%)	2692(91.9%)	<0.001
	African American	452(4.4%)	68(2.3%)
	Hispanic	207(2.0%)	38(1.3%)
	American Indian/Alaskan Native	64(0.6%)	11(0.4%)
	Asian or Pacific Islander	99(1.0%)	28(1.0%)
	Other	832(8.0%)	93(3.2%)
Age at study	18–29	7165(68.9%)	1634(53.9%)	<0.001
	30–39	2901(27.9%)	1093(36.0%)
	>39	331(3.2%)	307(10.1%)
Annual household income	≤ $19,999	2063(22.9%)	353(12.8 %)	<0.001
	$20,000–$59,999	4679(52.0%)	1368(49.7%)
	≥ $60,000	2255(25.1%)	1029(37.4%)
Marital status	Never married	5286(53.0%)	1001(35.0%)	<0.001
	Divorced/Separated/Widowed	814(8.2%)	262(9.2%)
	Married/Living as married	3874(38.8%)	1596(55.8%)
Highest level of schooling	Did not complete high school	1099(11.2%)	189(6.5%)	<0.001
	high school grad	2257(23.0%)	535(18.4%)
	high school + some college	3682(37.4%)	1059(36.4%)
	College grad and post grad	2795(28.4%)	1123(38.6%)
Work status	Never worked	543(5.2%)	38(1.3%)	<0.001
	Worked, but not in last 12 months	1540(14.9%)	221(7.3%)
	Worked last 12 months	8279(79.9%)	2766(91.4%)
Age at primary diagnosis (yrs)	0 – 3	1899(18.3%)
	4 – 10	3607(34.7%)
	11 – 14	2398(23.1%)
	15 – 20	2493(24.0%)
Diagnosis	Bone cancer	1133(10.9%)
	CNS tumors	1322(12.7%)
	Hodgkin Lymphoma	1876(18.0%)
	Wilms’ Tumor	670(6.4%)
	Leukemia	3061(29.4%)
	Non-Hodgkin lymphoma	928(8.9%)
	Neuroblastoma	416(4.0%)
	Soft tissue sarcoma	991(9.5%)

Pain Conditions

The cumulative incidence of pain/abnormal sensation, migraine headaches, and other frequent headaches were 12.3%, 15.5%, and 20.5% among survivors, and 6.3%, 15.2%, and 15.3% among siblings, respectively. Occurrence of ever having pain was compared between survivors and siblings. After adjusting for gender, ethnicity, household income, marital status, and education, survivors reported significantly higher rates of all three pain conditions: pain/abnormal sensation, (HR = 3.9, 95% CI = 3.3 – 4.7), migraines (HR = 1.9, 95% CI = 1.7 –2.1), and other frequent headaches, (HR = 1.9, 95% CI = 1.7 – 2.1),with all p-values <0.001.

Table 2 summarizes risk factors for each of the pain conditions among long-term survivors of childhood cancer. As expected, demographic factors were significantly associated with each of the three pain conditions. Females were more likely than males to report migraines and other frequent headaches. Compared with Caucasian survivors, Hispanic survivors had a significantly elevated risk of reporting abnormal sensation and other frequent headaches, and African American survivors were at a significantly increased risk of reporting all three pain conditions. Being single was significantly associated with increased risk of reporting all three pain conditions compared with not being single. Survivors who did not complete high school reported higher rates of all three pain conditions compared to those who completed high school. Survivors who had never worked reported higher rates of abnormal sensation and headache in comparison to those who worked in the past year.

Table 2.

Relative Risk Ratios and 95% Confidence Intervals for Predictors of Pain in Childhood Cancer Survivors by Pain Condition

	a
		Pain or abnormal sensation			Migraine			Other frequent headaches
		HR	95 % CI	p-value	HR	95 % CI	p-value	HR	95 % CI	p-value
Sex
	Male	1.0			1.0			1.0
	Female	0.94	(0.84– 1.04)	0.23	1.48	(1.33– 1.64)	<0.001	1.40	(1.26– 1.54)	<0.001
Ethnicity
	Caucasian	1.0			1.0			1.0
	Hispanic	1.74	(1.27– 2.39)	0.001	1.35	(0.97– 1.86)	0.08	1.44	(1.06– 1.96)	0.02
	African American	1.91	(1.58– 2.30)	<0.001	1.85	(1.54– 2.22)	<0.001	1.68	(1.40– 2.02)	<0.001
	American Indian/Alaskan Native	1.44	(0.85– 2.46)	0.18	1.18	(0.66– 2.09)	0.58	1.3	(0.75– 2.26)	0.35
	Asian or Pacific Islander	0.41	(0.16– 1.11)	0.08	0.97	(0.52– 1.81)	0.92	0.88	(0.45– 1.69)	0.69
	Other	0.97	(0.79– 1.17)	0.73	0.99	(0.83– 1.19)	0.94	1.05	(0.88– 1.25)	0.61
Annual Household Income($)
	<=19,999	1.17	(0.99– 1.38)	0.08	1.04	(0.89– 1.22)	0.59	1.08	(0.92– 1.26)	0.34
	20,000–59,999	1.07	(0.93– 1.23)	0.34	0.99	(0.87– 1.13)	0.91	1.0	(0.88– 1.14)	0.95
	>=60,000	1.0			1.0			1.0
Marital Status
	Single(never married)	1.0			1.0			1.0
	Married/Living as married	0.65	(0 57– 0.73)	<0.001	0.74	(0.66– 0.83)	<0.001	0.72	(0.64– 0.80)	<0.001
	Divorced/Separated/Widowed	0.66	(0.54– 0.80)	<0.001	0.74	(0.61– 0.88)	0.001	0.74	(0.61– 0.88)	0.001
Highest Level of Schooling
	Did not complete high school	1.0			1.0			1.0
	high school grad	0.73	(0.61– 0.87)	<0.001	0.75	(0.64– 0.89)	0.001	0.76	(0.65– 0.90)	0.001
	high school + some college	0.68	(0.57– 0.80)	<0.001	0.7	(0.59– 0.82)	<0.001	0.68	(0.58– 0.80)	<0.001
	College grad and post grad	0.32	(0.26– 0.39)	<0.001	0.38	(0.31– 0.45)	<0.001	0.33	(0.27– 0.40)	<0.001
Work Status
	Never worked	1.0			1.0			1.0
	worked last 12 months	0.7	(0.55– 0.89)	0.004	0.82	(0.65– 1.05)	0.12	0.76	(0.60– 0.96)	0.02
	Worked, but not in last 12 months	0.95	(0.73– 1.24)	0.72	0.99	(0.76– 1.29)	0.94	0.86	(0.67– 1.11)	0.25
Age at Diagnosis(yrs)
	0–3	1.0			1.0			1.0
	4–10	0.66	(0.56– 0.77)	<0.001	0.7	(0.60– 0.82)	<0.001	0.70	(0.60– 0.81)	<0.001
	11–14	0.33	(0.27– 0.40)	<0.001	0.34	(0.28– 0.41)	<0.001	0.35	(0.29– 0.42)	<0.001
	15–20	0.19	(0.15– 0.24)	<0.001	0.19	(0.15– 0.24)	<0.001	0.20	(0.16– 0.25)	<0.001
Primary Diagnosis
	Leukemia	1.0			1.0			1.0
	CNS tumors	0.70	(0.57– 0.85)	<0.001	0.74	(0.61– 0.90)	0.003	0.88	(0.73– 1.06)	0.18
	Hodgkin Lymphoma	0.79	(0.66– 0.94)	0.01	1.06	(0.90– 1.25)	0.48	0.88	(0.75– 1.04)	0.14
	Non-Hodgkin lymphoma	1.08	(0.89– 1.32)	0.44	1.31	(1.09– 1.59)	0.01	1.15	(0.95– 1.39)	0.14
	Wilms tumor	1.22	(0.98– 1.50)	0.07	1.34	(1.10– 1.62)	0.003	1.22	(1.01– 1.48)	0.04
	Neuroblastoma	1.29	(0.98– 1.68)	0.07	1.32	(1.03– 1.69)	0.03	1.07	(0.83– 1.37)	0.61
	Soft tissue sarcoma	0.87	(0.72– 1.06)	0.18	0.94	(0.78– 1.14)	0.54	0.82	(0.68– 0.99)	0.04
	Bone cancer	1.03	(0.84– 1.25)	0.79	1.01	(0.83– 1.23)	0.89	0.94	(0.77– 1.14)	0.51

b Cranial Radiation substituted for Diagnosis in above model
		Pain or abnormal sensation			Migraine			Other frequent headaches
		RR	95 % CI	p-value	RR	95 % CI	p-value	RR	95 % CI	p-value
Cranial Radiation(Gy)b
	No Radiation	1.0			1.0			1.0
	1.5–50	0.78	(0.59– 1.04)	0.09	0.70	(0.55– 0.89)	0.004	0.79	(0.64– 0.99)	0.04
	50–110	0.76	(0.47– 1.22)	0.26	0.81	(0.52– 1.27)	0.35	0.65	(0.41– 1.03)	0.07
	Scatter c	1.15	(0.91– 1.45)	0.24	1.25	(1.03– 1.52)	0.02	0.86	(0.71– 1.05)	0.14

Note:

^bAdjusted for same covariates as Diagnosis Model in Table 2A with the exception of Primary Diagnosis

^cScatter i.e. indirect exposure to brain due to scatter from non-cranial radiation treatment

Age at diagnosis ≤3 years was associated with a higher risk of reporting all three pain conditions as compared with ages at diagnosis from 4–20 in years. Differences in risk for the three pain conditions across cancer diagnoses were examined using the leukemia survivors as a comparison group. Survivors with a history of Non-Hodgkin lymphoma, and those who survived neuroblastoma had a higher risk of reporting migraines. Survivors with a history of Wilms tumor were at a higher risk of reporting migraines and other frequent headaches. Survivors with a history of CNS tumor had a lower risk of reporting abnormal sensation and migraines. Survivors with a history of Hodgkin lymphoma were at lower risk of reporting migraines. Survivors with a history of soft tissues sarcoma were at lower risk of reporting other frequent headaches.

No significant associations between pain conditions and treatment with chemotherapy or surgery were found (data not shown). As shown in Table 2b from a model without diagnosis included, survivors who received scatter radiation (i.e. indirect exposure to brain due to scatter from non-cranial radiation treatment) had elevated risk of migraines compared with those who received no radiation. Survivors who received direct cranial radiation doses in the range of 1.5–50 Gy had a lower risk of reporting migraines and other frequent headaches compared with those who received no irradiation.

Use of Prescribed Pain Medication

Among survivors, 17.9% reported using prescribed analgesics compared to only 12.6% of the siblings. The survivors were 1.4 times more likely to have used prescribed analgesics relative to the siblings (95% CI=1.2–1.6), adjusted for gender, race, household income, age at study, marital status, education, and work status.

As shown in Table 3a, female survivors were 1.8 times more likely to use prescribed analgesics than were male survivors. Survivors who had < $20,000 total household income were more likely to use prescribed analgesics relative to those with income ≥ $60,000. Survivors who did not complete high school were more likely to use prescribed analgesics than were high school graduates. Those who had been employed before, but not in the last 12 months, were more likely to have used prescribed analgesics compared to those who had never been employed. Survivors who were separated, divorced, or widowed were more likely to use prescribed pain medication relative to those who were never married/partnered. No effect of ethnicity on use of pain medication was observed among survivors.

Table 3.

Odds Ratios for Using Pain Medication among Childhood Cancer Survivors

a
	Covariates	Odds Ratio	95 % CI	p-value
Sex
	Male	1.0
	Female	1.77	(1.56– 2.00)	<0.001
Race/Ethnicity
	Caucasian	1.0
	Hispanic	1.12	(0.75– 1.67)	0.59
	African American	0.84	(0.61– 1.14)	0.26
	American Indian/Alaskan Native	1.21	(0.58– 2.52)	0.61
	Asian/Pacific Islander	0.65	(0.30– 1.37)	0.25
	Other	1.02	(0.82– 1.27)	0.85
Age at baseline(yrs)
	18–29	1.0
	30–39	0.90	(0.76– 1.07)	0.22
	>=39	1.0	(0.70– 1.43)	0.99
Annual household income($)
	<=19,999	1.38	(1.15– 1.66)	<0.001
	20,000–59,999	1.16	(0.99– 1.35)	0.06
	>=60,000	1.0
Marital Status
	Single (never married)	1.0
	Divorced/Separated/Widowed	1.32	(1.05– 1.65)	0.02
	Living as Married	1.25	(1.00– 1.57)	0.05
	Married	1.09	(0.93– 1.27)	0.28
Education
	Did not complete high school	1.0
	high school grad	0.80	(0.65– 0.99)	0.04
	high school + some college	0.79	(0.64– 0.96)	0.02
	College grad and post grad	0.57	(0.46– 0.72)	<0.001
Work Status
	Never worked	1.0
	Worked, but not in past 12months	1.55	(1.13– 2.13)	0.01
	Worked in past 12months	0.94	(0.70– 1.27)	0.68
Age at Diagnosis(yrs)
	0–3	1.0
	4–10	1.04	(0.86– 1.26)	0.67
	11–14	1.22	(0.97– 1.53)	0.08
	15–20	1.31	(1.02– 1.68)	0.03
Primary Diagnosis
	Leukemia	1.0
	Bone Cancer	1.29	(1.04– 1.60)	0.02
	CNS tumor	0.89	(0.71– 1.10)	0.29
	Hodgkin lymphoma	0.91	(0.74– 1.11)	0.35
	Wilms tumor	1.03	(0.79– 1.34)	0.82
	Non-Hodgkin lymphoma	0.91	(0.72– 1.17)	0.47
	Neuroblastoma	1.33	(0.97– 1.82)	0.08
	Soft Tissue sarcoma	1.32	(1.07– 1.64)	0.01

b Cranial Radiation substituted for Diagnosis in above model
	Covariates	Odds Ratio	95 % CI	p-value
Cranial Radiation(Gy)
	No radiation	1.0
	1.5–<50	1.0	(0.84– 1.19)	0.99
	50–110	0.93	(0.70– 1.23)	0.60
	Scatter	0.93	(0.80– 1.10)	0.40

Survivors who were diagnosed with cancer between the ages of 15–20 years were more likely to use prescribed analgesics compared with those diagnosed at <3 years of age. Survivors with a history of bone cancer or soft tissue sarcoma were more likely to use prescribed analgesics compared with leukemia survivors. None of the other types of cancer diagnoses was significantly associated with use of pain medication compared to leukemia. Cranial radiation was not significantly associated with use of pain medication.

Current Pain Attributed to Cancer or its Treatment

Among survivors, 21.0% reported having pain in the past week that they attributed to their history of cancer or its treatment. Specifically, 10.8 % reported having only a “small amount” of cancer-related pain; 6.7 % reported having a “medium amount” of cancer-related pain; and 3.4% reported having “a lot, or very bad excruciating” cancer-related pain.

As shown in Table 4a, survivors with annual household incomes of < $20,000, or $20,000–59,000 were more likely to report pain attributed to cancer or its treatment than those with incomes ≥ $60,000. Survivors who did not complete high school were more likely to report pain attributed to cancer or its treatment compared with high school graduates. Survivors who had ever been married/partnered were more likely to report pain attributed to cancer or its treatment than those who were never married/partnered. Survivors with a history of bone cancer and soft tissue sarcoma were more likely to report pain attributed to cancer or its treatment compared to those with a history of leukemia, while Hodgkin lymphoma survivors were less likely to report pain attributed to cancer or its treatment compared to leukemia survivors.

Table 4.

Odds Ratios for Predictors of Recent Pain Attributed to Cancer or its Treatment in Childhood Cancer Survivors

a
Factors meeting Proportional Odds Assumption		Odds Ratio	95 % CI	p-value
Race/Ethnicitya
	Non-Caucasian	1.0
	Caucasian	1.01	(0.86– 1.19)	0.87
Age at study (yrs)
	18–29	1.0
	30–39	1.11	(0.95– 1.29)	0.19
	≥ 39	1.22	(0.86– 1.72)	0.26
Annual household income($)
	≤$19,999	1.64	(1.37– 1.95)	<0.001
	$20,000–$59,999	1.17	(1.01– 1.35)	0.03
	≥$60,000	1.0
Marital Status
	Single (never married)	1.0
	Divorced/Separated/Widowed	1.40	(1.12– 1.74)	0.003
	Living as Married	1.34	(1.08– 1.67)	0.01
	Married	1.11	(0.96– 1.29)	0.17
Education
	Did not complete high school	1.0
	high school grad	0.67	(0.54– 0.84)	<0.001
	high school + some college	0.77	(0.63– 0.94)	0.01
	College grad and post grad	0.67	(0.54– 0.83)	<0.001
Primary Diagnosis
	Leukemia	1.0
	Bone Cancer	3.29	(2.73– 3.98)	<0.001
	CNS tumor	0.96	(0.77– 1.19)	0.69
	Hodgkin lymphoma	0.59	(0.48– 0.73)	<0.001
	Wilms tumor	1.13	(0.86– 1.47)	0.38
	Non Hodgkin lymphoma	0.89	(0.70– 1.13)	0.34
	Neuroblastoma	1.18	(0.84– 1.66)	0.35
	Soft Tissue Sarcoma	1.93	(1.58– 2.36)	<0.001

Factors not meeting Proportional Odds Assumption
			≥'a lot or very bad excruciating'			≥'Medium amount			≥'Small amount'
		Odds Ratio	95 % CI	p-value	Odds Ratio	95 % CI	p-value	Odds Ratio	95 % CI	p-value
Sex
	Male	1.0			1.0			1.0
	Female	0.69	(0.52– 0.91)	0.01	0.98	(0.84– 1.16)	0.84	1.15	(1.02– 1.30)	0.03
Work Status
	Never worked	1.0			1.0			1.0
	Worked, but not in the previous 12month	2.57	(1.19– 5.54)	0.02	2.15	(1.36– 3.39)	0.001	2.00	(1.36– 2.94)	<0.001
	Worked in the previous 12months	0.93	(0.45– 1.93)	0.84	0.92	(0.60– 1.41)	0.69	1.28	(0.90– 1.84)	0.17
Age at Diagnosis(yrs)
	0–3	1.0			1.0			1.0
	4–10	0.92	(0.61– 1.39)	0.70	0.92	(0.71– 1.19)	0.52	1.17	(0.96– 1.42)	0.12
	11–14	0.89	(0.57– 1.38)	0.59	1.18	(0.89– 1.57)	0.25	1.65	(1.31– 2.08)	<0.001
	15–20	1.21	(0.79– 1.87)	0.39	1.31	(0.98– 1.75)	0.07	1.69	(1.32– 2.16)	<0.001

b: Cranial Radiation substituted for Diagnosis in above model
		Odds Ratio	95 % CI	p-value
Cranial Radiaiton(Gy)b
	No radiation	1.0
	1.5–<50	0.95	(0.80– 1.13)	0.57
	50–110	0.99	(0.75– 1.31)	0.97
	Scatter	1.20	(1.04– 1.40)	0.02

^aThis covariate was simplified to two levels due to sparse number

^bAdjusted for same covariates as Diagnosis Model in Table 3a with the exception of Primary Diagnosis

The lower section of Table 4a displays the factors for which odds ratios differed across different cut-points for the cancer pain attribution outcome. These results illustrate that although females were more likely to attribute equal to or greater than “small amount” of pain to their cancer and treatment than were males, they were less likely to attribute “a lot or very bad excruciating” pain to cancer and its treatment compared with males. Those who had worked before, but not in the last 12 months were more likely to attribute a higher level of pain to cancer or its treatment compared to survivors who never worked. Compared to survivors diagnosed at or before 3 years of age, survivors diagnosed between 11–20 years of age were more likely to attribute equal to or greater than a ‘small amount’ of pain to cancer and its treatment.

In a separate model with all factors except diagnosis, survivors who received scatter radiation were more likely to report pain attributed to cancer or its treatment compared to those who did not receive any radiation (Table 4b).

Discussion

Current findings indicate that adult survivors of childhood cancer experience increased risk of reporting a pain condition and for using prescription analgesics compared to a sibling control group. We observed elevated prevalence estimates in survivors for self-reported pain/abnormal sensation (12.3%), migraines (15.5%), and other headaches (20.5%). The pain prevalence estimates correspond with the proportion of survivors who reported recently using prescription analgesics (16.7%) and who reported current pain attributed to their cancer or its treatment (21%). Moreover, moderate and severe levels of current pain attributed to cancer or treatment were reported by 6.7% and 3.4% of survivors. Although participants were not asked to report current pain attributable to other sources, it is reasonable to assume that the overall prevalence of pain, disregarding attribution source, is greater than 21%.

Survivors reported significantly higher rates of all types of pain conditions assessed as well as higher use of prescription analgesics, even after controlling for observed demographic differences in income, employment, educational attainment, and marital status. These findings suggest a greater risk of being diagnosed with a pain condition among long-term childhood cancer survivors, consistent with data from studies with survivors of adult-onset cancer [29]. Despite previous reports that pain prevalence increases with age [4, 36], rates among our young cohort of childhood cancer survivors (with the majority less than 39 years) are similar to rates in the general population of older adults [7, 30, 36]. The findings are consistent with the known risk for health-related complications in childhood cancer survivors [8].

Hypotheses regarding demographic risk factors were largely confirmed. Female gender and lower educational attainment were associated with greater risks for all three pain outcomes. Minority status, unemployment, and being single were associated with greater risks for reporting all three pain conditions. Lower income was associated with great risk of using prescription pain medication and attributing pain to cancer. These risk factors correspond with known risk factors for pain in childhood cancer survivors and clinical and general adult populations. For instance, previous studies have found risk factors for pain, such as gender [13, 40], lower socioeconomic status [9], minority status [2, 27], and unemployment [4]. These factors may each relate to the availability of health insurance and access to health care [10, 16]. These risk factors are also similar to known risk factors for poor health status [19, 20] and poor psychological functioning [43] in childhood cancer survivors.

The observed pattern of differences in pain outcomes across cancer treatments, diagnoses, and age at diagnosis adds to what has previously been reported in smaller studies [3, 41]. Compared with those who received no irradiation, survivors who received indirect (scatter) radiation to the brain had elevated risk for migraines. They were also more likely to attribute pain to cancer or its treatment. The finding provides indirect evidence for the argument that radiation induced pain might sustain decades after treatment completion [5]. On the other hand, unexpectedly, compared with those who received no radiation, survivors who received cranial radiation doses in the range of 1.5–50 Gy to the brain reported lower risk for migraines and headaches. Results for direct brain irradiation in the range of 50–110 Gy, although not significant, also followed similar patterns. One possible explanation is that cognitive dysfunction associated with whole brain doses of ≥18 Gy may have impacted the reporting of pain on the questionnaire [21]. A previous study found that brain tumor survivors who received the highest doses of cranial irradiation had the greatest severity of cognitive late effects, and were less likely to report pain/abnormal sensation and migraines [11]. It is plausible that the impact of brain irradiation on cognition may lead to both underreporting of pain and lack of attribution of current pain to cancer or its treatment. The differential finding of pain between survivors who received direct brain irradiation and those who only received indirect brain irradiation suggests the need for screening for cognitive dysfunction among the former group of survivors. Alternatively, perhaps those survivors exposed to direct cranial radiation differed in some other aspect of their treatment. We examined contributions in risk for pain from common systemic chemotherapy agents and found no clear effect. Further research may be warranted to confirm and explore this unexpected inverse pattern between cranial radiation and pain.

Our hypotheses regarding surgery and chemotherapy were not supported, as no significant associations with pain were detected (data not shown). Perhaps the impact of surgery and chemotherapy on pain decreases as survival time increases. Future studies should examine the impact of treatment within specific diagnostic groups as well as its trajectory longitudinally.

We observed increased risks for migraines and other headaches among survivors of non-Hodgkin lymphoma, Wilms tumors, and neuroblastoma in comparison to leukemia survivors, consistent with findings from a previous study [3] that survivors of Wilms tumor or advanced neuroblastoma exhibited high levels of pain during childhood. Tumors that potentially invade peripheral sensory afferents or paraspinal ganglia may produce peripheral sensitization followed by central sensitization [28]. This finding might relate to disruption of pain inhibiting mechanisms or development of pathological pain pathways [37]. Although survivors of bone cancer or soft tissue sarcoma did not report higher levels of the specific pain conditions assessed, they were at increased risk of attributing pain to their cancer or treatment and of using prescription analgesia. The higher risk in these two diagnostic groups may relate to current ongoing pain in areas associated with their original tumor rather than in the specific pain locations included on the questionnaire. CNS cancer survivors reported lower rates of pain conditions compared with leukemia survivors. These findings were possibly related to cognitive dysfunction differences in the two diagnostic groups or to other yet unexplored reasons.

Consistent with our hypothesis and a previous study [1], we found that survivors diagnosed at a very young age (≤ 3 years of age) had higher risk of reporting all three pain conditions compared with those diagnosed at an older age (3–20 years). In contrast, survivors diagnosed at an older age (15–20 years) were more likely to attribute pain to cancer and report prescription pain medication use compared to those diagnosed at a very young age (≤ 3 years of age). One possible explanation is that knowledge of pain assessment and management in infants and toddlers was likely to be lower during the dates when these young children were diagnosed, which may have resulted in less adequate pain management. Also, survivors who were diagnosed at an older age might take “cancer diagnosis” as a more salient cue and thus be more likely to attribute current pain to their original cancer or treatment and be more likely to use analgesics. Survivors diagnosed at a very young age might not have remembered their cancer treatment and thus not associate any current pain with their original cancer.

Several limitations of the current study are worth noting. First, the study was cross-sectional in design, and thus causal relationships could not be determined. Furthermore, neither a standardized pain measure was used nor were queries about pain in all parts of the body made (such as chest, back, abdomen, or leg). Participants were not asked to report on their use of over-the-counter analgesics, such as NSAIDs. Thus, the limited choice of locations of pain and exclusion of over-the-counter pain medication may have led to the under-reporting of pain and use of analgesics in this population. To be conservative, we treated “not sure” response to the pain condition and prescription medication questions as a “no” response, a categorizing that might also have led to underestimated pain rates. Finally, the sample was primarily comprised of Caucasians. The limited ethnic diversity of the sample may have underestimated the overall pain rates, weakened the ability to detect ethnic differences, and reduced the generalizability of the findings. Lower percentage of minorities among siblings compared with survivors may also have overestimated differences between survivors and siblings; although this likelihood was reduced by the adjustment for race/ethnicity in all analyses. There may have also been differences between survivors and siblings in medical visits. Siblings however, represent a better control group compared with population norms because of genetic and family of origin similarities [42]. It is possible, however, that siblings, due to the experience of having a sibling with cancer are at elevated risk of pain outcomes themselves. If this is the case, then the relative risk results shown here would underestimate the differences between survivors and a non-sibling control group. Despite these deficiencies in the questionnaire and possible underestimate of pain prevalence, the study did show differences in the pain outcomes assessed between survivors and siblings.

Importantly, risk factors for developing pain (i.e., female gender, minority status, lower SES, unemployment, and being single) identified in this study overlap considerably with known risk factors for other complications. Health-care providers should be mindful that adult survivors of childhood cancer who present with risk factors for developing poor health status are also likely to be at risk for pain. By identifying those at risk for poor outcomes, it may be possible to develop targeted, multifaceted treatments to prevent adverse outcomes such as pain and high levels of analgesic use. Evaluating the psychological, social, environmental, and medical factors associated with pain in this population is an important direction for future research.