|Home | About | Journals | Submit | Contact Us | Français|
Little is known about the trajectory of post-traumatic stress disorder (PTSD) symptoms in cancer survivors, despite the fact that such knowledge can guide treatment. Therefore, this study examined changes in PTSD symptoms among long-term survivors of non-Hodgkin's lymphoma (NHL) and identified demographic, clinical, and psychosocial predictors and correlates of PTSD symptomatology.
Surveys were mailed to 682 NHL survivors who participated in an earlier survey and now were at least 7 years postdiagnosis. Information was obtained regarding PTSD symptoms, positive and negative perceptions of the cancer experience (ie, impact of cancer), and other potential correlates of PTSD.
A total of 566 individuals participated (83% response rate) with a median of 12.9 years since diagnosis; respondents were 52% female and 87% white. Although half (51%) of the respondents reported no PTSD symptoms and 12% reported a resolution of symptoms, more than one-third (37%) reported persistence or worsening of symptoms over 5 years. Survivors who reported a low income, stage ≥ 2 at diagnosis, aggressive lymphoma, having received chemotherapy, and greater impact of cancer (both positive and negative) at the initial survey had more PTSD symptoms at follow-up. In multivariable analysis, income and negative impacts of cancer were independent predictors of PTSD symptoms.
More than one-third of long-term NHL survivors experience persisting or worsening PTSD symptoms. Providers should be aware of enduring risk; early identification of those at prolonged risk with standardized measures and treatments that target perceptions of the cancer experience might improve long-term outcomes.
Recent advances in treating non-Hodgkin's lymphoma (NHL), one of the most common forms of cancer, have led to a more than doubling of the 5-year relative survival rate from the 1960s, from 31% to 69%.1 Thus, NHL is often perceived by survivors as a life-long chronic illness, with alternating symptom-free and symptom-exacerbation phases that may require treatment. As with other chronic illnesses, comorbidities are common in the context of NHL, including symptoms of post-traumatic stress disorder (PTSD), which were identified among 39% of NHL survivors2 and have been shown to be associated with poor quality of life3–5 and depression6 in other cancer samples.
PTSD symptoms of avoidance, arousal, and re-experiencing have been identified in individuals diagnosed with an adult cancer in several cross-sectional studies.2,7,8 A few longitudinal studies have examined cancer-related PTSD during the first year after a diagnosis.9–11 However, none have examined the trajectory of cancer-related PTSD symptoms in long-term survivors, so nothing is known regarding symptom stability in this population. Furthermore, little is known about the predictors of PTSD in this understudied group of NHL survivors, the knowledge of which may be helpful in identifying those at future need.
Our initial study defined the NHL survivor experience in terms of PTSD,2 quality of life,12 and the impact of cancer (IOC).13,14 That study was cross sectional with a median of 8 years postdiagnosis. This follow-up assessment resurveyed study participants 5 years later and extends our examination to a median of 13 years postdiagnosis, providing a unique window into the longitudinal NHL experience. The purpose of this follow-up study is to examine change in PTSD symptoms over a 5-year period among NHL survivors and identify demographic, clinical, and psychosocial factors associated with changes in PTSD symptoms. We hypothesized that PTSD symptoms would persist or worsen for a small proportion of the sample, particularly given the potential for disease relapse, and that non-white race, younger age, less education and social support, comorbidities, and negative perceptions of the cancer experience would be associated with follow-up PTSD symptoms after controlling for initial PTSD symptoms. These findings have important implications for clinical care and future research on the late effects of NHL diagnosis and treatment.
We conducted a longitudinal follow-up assessment of NHL survivors who consented to an earlier study and were treated at Duke University or the University of North Carolina Cancer Centers. The institutional review boards at both institutions approved this study. Initial study eligibility criteria required that individuals had been diagnosed with NHL ≥ 2 years previously and were ≥ 19 years of age; details of initial study procedures are published elsewhere.2 Hence, the study cohort for this study was ≥ 7 years postdiagnosis and ≥ 24 years of age. The consent form used in the initial study included a statement of willingness to be recontacted within the next 5 years. Follow-up data were collected in 2010; surveys from 2005 and 2010 were linked at the individual patient level to facilitate within-person analyses of change over time.
In accordance with the Dillman method for administering surveys,15 a brief prenotice letter was mailed to patients who were assumed living and within 5 years of their initial study enrollment. The follow-up survey was mailed 2 weeks later and included a postage-paid return envelope, a $2 incentive, and a form to indicate interest in participating in future studies and/or receiving an educational CD and summary of research findings. Thank-you and/or reminder postcards were mailed 2 weeks later, and nonresponders were sent replacement surveys and later telephoned to confirm receipt of the survey.
Data collected in this second survey included demographic and clinical status (eg, change in marital status, recurrence of NHL). The Self-Administered Comorbidity Questionnaire (SCQ), a 12-item self-report version of the Charlson Index, was used to assess non–cancer-related problems.16 In scoring the SCQ, an individual can receive up to 3 points for each of 15 medical conditions (1 point each for presence of the problem, current treatment, and functional limitation; range, 0 to 45 points).
Psychosocial well-being was assessed with the Medical Outcomes Study Social Support Survey (MOS-SSS)17 and the Impact of Cancer Version 2 (IOCv2) surveys.13 Scores for the 20-item MOS-SSS range from 20 to 100, with higher scores representing better social support. The IOCv2 assessed the positive life changes and negative impacts attributed to the cancer experience. It includes 37 items to measure four positive (Altruism/Empathy, Health Awareness, Meaning of Cancer, Positive Self-Evaluation) and four negative (Appearance Concerns, Body Change Concerns, Life Interference, Worry) subscales, which total to Positive and Negative Impact Summary scores (range, 1 to 5). Higher Positive Impact Summary scores indicate greater positive perceptions, and higher Negative Impact Summary scores indicate more negative perceptions.
Post-traumatic stress symptoms were measured with the PTSD Checklist-Civilian Version (PCL-C),18 a self-report symptom checklist that closely mirrors the diagnosis criteria in Diagnostic and Statistical Manual of Mental Disorders, Revision IV (DSM-IV).19 The instructions were modified so that symptoms were keyed to the particular traumatic stressor of interest; specifically, participants were asked to rate each PTSD symptom in the past 4 weeks with respect to their NHL diagnosis and treatment. Each of 17 symptoms is rated with respect to intensity on a scale of 1 (not at all) to 5 (extremely bothersome). Two approaches were used to construct an aggregate score in assessing symptoms: the continuous score (range, 17 to 85) and the symptom cluster method, which follows the DSM-IV PTSD symptom structure. For example, individuals would be considered as having a PTSD symptom if they reported having been at least moderately bothered by (score ≥ 3) at least one of five re-experiencing symptoms (eg, nightmares), at least three of seven avoidance symptoms (eg, evading follow-ups), or at least two of five arousal symptoms (eg, easily startled).
To compare follow-up study participants and nonparticipants with respect to initial demographic, clinical, and psychosocial characteristics and PTSD, we tested for differences between participants and both decedents and nonresponders by using t tests for continuous measures and χ2 tests for categorical measures. Because initial PTSD symptomatology is one of the strongest predictors of PTSD symptoms at follow-up, all analyses that examined characteristics associated with follow-up PCL-C scores were adjusted for the initial score of this measure. To assess the association between each of the independent variables (ie, data from the initial survey as predictors and from the follow-up survey as correlates) and the follow-up PCL-C, we used a series of linear regression models with follow-up PCL-C as the dependent variable. We first tested each independent variable separately (ie, only the candidate variable and initial PCL-C score in the model). Then, variables that were at least marginally significantly associated with follow-up PCL-C in these models (P < .10) were included in a multiple linear regression to estimate the independent associations of initial survey predictors with follow-up PCL-C. An additional model was estimated that included both the initial survey predictors and follow-up correlates. For the psychosocial measures, change scores (follow-up score minus initial score) were used rather than follow-up scores, so that the effect of changes in these measures independent of initial status could be evaluated. Data management and statistical analyses were conducted by using SAS Version 9.2 (SAS Institute, Cary, NC).
Of the 886 initial study participants, 176 had died, four were ineligible (because of change in diagnosis or dementia), and 24 of the mailed surveys were returned as undeliverable at follow-up. Of the 682 eligible individuals who were assumed to have received a follow-up survey, 566 (83%) completed and returned their survey, 107 (16%) did not respond, and nine (1%) refused to participate (Fig 1).
Initial demographic, clinical, and psychosocial characteristics and PTSD symptoms of the 566 individuals who participated in the follow-up study and the 144 nonresponders are provided in Table 1. The mean and median time since diagnosis for all participants was 10.4 and 8.0 years, respectively; the median time since diagnosis for the subset that reported a recurrence was 8.8 years. Compared with decedents (not shown in Table 1), follow-up study participants were more likely to report the following characteristics on the initial survey: income ≥ $30,000, college degree, married or living with a partner, younger age, aggressive lymphoma, no current treatment, no active disease, lower comorbidity, lower IOC Negative Impact Summary score, lower PCL-C score, and fewer PTSD symptoms (all P < .05). In addition, follow-up study participants compared with nonresponders were more likely to report white race, income ≥ $30,000, married or living with a partner, older age, received biologic therapy, no active disease, and lower PCL-C scores (all P < .05). At follow-up, 41 (7.3%) of 557 had a PCL-C score ≥ 44, which is indicative of PTSD,20 whereas 28 (5.0%) of 557 scored ≥ 44 at the time of the initial survey. The most common PCL-C problems endorsed by participants as moderately to extremely bothersome were sleep and concentration difficulties and loss of interest in activities (Table 2).
The follow-up survey was completed an average of 4.8 years after initial survey completion (range, 4.3 to 5.4 years). Total and subscale scores for the comparison of the initial survey and the follow-up PCL-C are listed in Table 3. Paired t tests found no significant change in mean scores reported by the 557 participants who completed the PCL-C at initial and follow-up surveys. However, the avoidance/numbing subscale had the largest mean change or increase in symptoms of 0.2 (standard deviation, 3.6; P = .15).
Table 4 provides a cross tabulation of PTSD symptoms at the time of the initial survey and follow-up. Among the 557 participants who completed the PCL-C at both times, 281 (50.4%) did not report symptoms at either time point, 68 (12.2%) reported a resolution of symptoms, 23 (4.1%) reported improvement but persistent symptoms, 79 (14.2%) reported stable and persistent symptoms, and 106 (19.0%) reported a worsening of symptoms. In addition, 73 (13.1%) reported a 5- to less than 10-point increase in PCL-C scores, which represents reliable change (ie, worsening not due to chance). In addition, 39 (7.0%) reported at least a 10-point increase, considered to be a clinically significant worsening of symptoms. Conversely, 88 (15.8%) reported at least a 5-point decrease in PCL-C scores (ie, improvement of symptoms).21
The results of linear regression models are provided in Table 5, the first set of models adjusting only for initial PCL-C score, the second model adjusting for initial PCL-C score and significant predictors (initial survey responses), and the third model adjusting for initial PCL-C score, significant predictors, and significant correlates (follow-up survey responses). For the models with adjustment for initial PCL-C only (Column 1), the initial survey predictors of having PTSD symptoms at follow-up were income less than $30,000 (P < .001), having an aggressive lymphoma (P = .025), stage ≥ II at diagnosis (P = .007), having had chemotherapy (P = .027), and IOC Negative Impact Summary (P < .001) and IOC Positive Impact Summary (P = .006) scores. A recurrence in the last 5 years (P = .011), higher follow-up comorbidity score (P = .003), and increase in IOC Negative Impact Summary score (P < .001) were associated with higher follow-up PCL-C. For the multiple linear regression model with adjustment for initial PCL-C and other initial survey measures (Column 2), the significant independent predictors of having more PTSD symptoms at follow-up were income less than $30,000 (P < .001) and IOC Negative Impact Summary score (P < .001). Regarding the multiple linear regression model with adjustment for initial PCL-C, other initial measures, and changes in status between the initial survey and follow-up (Column 3), the significant initial survey predictors of greater PTSD symptoms at follow-up were income less than $30,000 (P < .001), not currently receiving treatment (P = .013), and IOC Negative Impact Summary (P < .001) and IOC Positive Impact Summary (P = .030) scores. Increase in IOC Negative Impact Summary score was also a significant correlate of follow-up PCL-C (P < .001). For the models in Columns 2 and 3, R2 values were 0.55 and 0.65, respectively, indicating that the demographic, clinical, and psychosocial variables accounted for a substantial amount of the variance in follow-up PCL-C.
To shed further light on the findings related to negative and positive impacts, Table 6 provides the IOC Subscale predictors and correlates of follow-up PCL-C, with three models again controlling for initial PCL-C and other factors. For the linear regression model with adjustment for initial PCL-C only (Column 1), all of the IOC Negative and Positive Impact Subscale scores were significant predictors of more PTSD symptoms at follow-up (all P < .05), except “Health Awareness.” In addition, increases in IOC Negative Impact Subscale scores were associated with greater PTSD symptoms at follow-up (all P < .001), but there were no associations for change in Positive Impact Subscale scores. For the model with adjustment for initial PCL-C and other initial measures (Column 2), a higher “Worry” score was the only IOC Subscale associated with more PTSD symptoms at follow-up (P = .019). Regarding the multiple linear regression model adjusted for initial PCL-C and other initial and follow-up measures (Column 3), IOC Negative Impact “Appearance Concerns,” “Life Interferences,” and “Worry” scores were significant initial survey predictors of more follow-up PTSD symptoms (all P < .05). Further, increases in IOC Negative Impact “Life Interferences” and “Worry” scores were independently associated with higher PCL-C scores (both P < .001). In addition, a decrease in the IOC Positive Impact “Positive Self-Evaluation” score was an independent correlate of follow-up PCL-C (P = .005).
In this follow-up study, the largest longitudinal study of PTSD symptomatology among adult cancer survivors reported in the literature to date, we found that although 51% of survivors did not have PTSD symptoms at either time point, and 12% had PTSD symptoms that resolved over 5 years, more than one third of the sample reported persistent (18%) or worsening (19%) PTSD symptoms over a 5-year period. These findings of persistent PTSD symptomatology among NHL survivors are consistent with those found among disaster victims22,23 and victims of violence.24–26 Importantly, several characteristics were identified that could help screen and target treatments for those survivors who are at risk for prolonged PTSD symptoms and inform opportunities to reduce the impact of potentially PTSD-inducing cancer care scenarios.
Although our hypotheses were partially supported (eg, finding of persisting and worsening of symptoms), we found that the explanation for follow-up PTSD symptoms was not limited to recurrence status. Specifically, individuals with an initial status of lower income, aggressive lymphoma, stage ≥ II at diagnosis, having had chemotherapy, and higher IOC Negative and Positive Impact Summary scores were more likely to have more PTSD symptoms 5 years later. Importantly, income and Negative Impacts were most influential in adjusted analyses (ie, more strongly predictive of PTSD symptoms) than the clinical aspects of the disease and treatment.
How might these findings be used? Two strong messages emerge: (1) the upfront cancer treatment experience directly influences downstream patient experience and risk of PTSD, and (2) specific individuals are at increased risk. Therefore, whole-person interventions targeted at improving the experience of patients with NHL, including mitigating PTSD, could seek to do either or both of the following: (1) improve those elements of the treatment experience that lead to negative impressions (eg, reducing life interference by social work intervention, improving a sense of body image by physical therapy, relieving worry by cognitive behavioral therapy), and (2) recognize that NHL survivors from poor socioeconomic circumstances are at the highest risk and triage efficiently to social support services. In addition, clinical care should include a formal assessment of symptoms that includes domains similar to those on the IOC. However, additional testing of the IOC in other samples is needed to support its use earlier in the cancer trajectory.
Although there are no known psychosocial interventions developed specifically for NHL survivors, evidence-based offerings could be tailored to meet the unique needs of this population. The alternating symptom-free and symptom exacerbation characteristics of NHL coupled with the difficulty in distinguishing between signs of aging and long-term symptoms from treatment may exacerbate worry or fear of recurrence in survivors. The Managing Uncertainty Day-to-Day intervention27 is designed to help older breast cancer survivors manage fears of recurrence and improve coping skills by delivering cognitive strategies via audiotape. In addition, a supportive-expressive group therapy intervention has been shown to significantly reduce trauma symptoms and mood disturbance in women with advanced breast cancer.28 Thus, treatments exist that might benefit PTSD outcomes of NHL survivors, especially if they are targeted to those of greatest presumed risk.
The findings in this study are especially beneficial, given the large sample size, excellent response rate, use of standardized measures, and longitudinal design. Although our study included only two survey administrations, it is a starting point of depicting the experiences of patients with cancer. Study limitations include the representation of a predominantly married and white sample and potential nonresponse bias. However, our racial profile closely mirrors that of the national population of NHL survivors, thereby strengthening the generalizability of our findings. There is also evidence that individuals who chose not to participate at follow-up were not doing well at the initial survey; therefore, the level of PTSD symptomatology may be underestimated. Second, the 28-page survey lacked measures assessing other psychological problems and life stressors in an effort to minimize the burden on respondents. Third, potential biases inherent in the use of self-report measures were minimized by using standardized instruments. And last, there is a potential overlap of PTSD symptoms with those related to cancer and treatment. However, only a small proportion (9.4%) of the sample reported active NHL or receiving treatment at follow-up. Furthermore, follow-up disease and treatment status were not predictive of PTSD symptoms in the linear regression analysis.
In conclusion, this article provides new information about the trajectory of PTSD symptoms in long-term NHL survivors. The initial cancer experience (and the need to improve it) becomes paramount in improving downstream outcomes, such as PTSD symptomatology. In addition, the identification of several patient characteristics related to PTSD risk could inform the screening process early in the survivorship trajectory. Furthermore, the strong predictive role of the IOC is consistent with previous cross-sectional studies and suggests that negative perceptions related to the cancer experience could be targeted in interventions as a means to minimize PTSD symptomatology. Future work should focus on the identification of patients at risk by using predictive models applied at the point of care and development of low-cost interventions that are delivered to those exhibiting clinically significant PTSD symptomatology during treatment to improve the long-term patient experience.
We thank the many non-Hodgkin's lymphoma survivors who participated in our study for their continued support. We also thank Kimberly Ward and other members of the research staff of the Cecil G. Sheps Center for Health Services Research, University of North Carolina at Chapel Hill, for assisting with the fielding of this survey.
Supported by Grant No. R03CA101492 and Cancer Care Quality Fellowship R25CA116339 from the National Cancer Institute, Grant No. DSW-0321301-SW from the American Cancer Society, Grant No. 10KR71019 from the University of North Carolina (UNC) Translational and Clinical Sciences Institute, and an award from the UNC Research Council.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
Although all authors completed the disclosure declaration, the following author(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
Employment or Leadership Position: None Consultant or Advisory Role: None Stock Ownership: None Honoraria: Amy P. Abernethy, Helsinn Therapeutics (US), Proventys, Amgen Research Funding: Amy P. Abernethy, Pfizer, Eli Lilly, Bristol-Myers Squibb, Helsinn Therapeutics (US), Amgen, KangLaiTe, Alexion, BioVex Expert Testimony: None Other Remuneration: None
Conception and design: Sophia K. Smith, Sheryl Zimmerman
Collection and assembly of data: Sophia K. Smith
Data analysis and interpretation: All authors
Manuscript writing: All authors
Final approval of manuscript: All authors