Fatigue level changes with time in long-term Hodgkin and non-Hodgkin lymphoma survivors: a joint EORTC-LYSA cross-sectional study

In 2009–2010 the EORTC Lymphoma Group and the LYSA have designed a cross-sectional study to collect information on socio-demographic characteristics, health situation and fatigue of HL survivors enrolled in the nine clinical trials that were conducted from 1964 to 2004. Two self-administered questionnaires were used in addition to clinical data prospectively collected and stored in a unique secured database at the EORTC Head Quarter in Brussels, Belgium. In 2015 the LYSA repeated the cross-sectional study in NHL survivors enrolled in the 12 clinical studies that were conducted from 1993 to 2010. The same two self-administered questionnaires were used in addition to clinical data prospectively collected and stored in a unique secured database at the LYSA Academic Research Organisation, Centre Hospitalier Lyon-Sud, Pierre-Bénite, France. Survivors were eligible if they had no active lymphoma, had follow-up of 5 years or more, and were free from any cancer treatments since 4 years. Detailed descriptions of the cross-sectional studies were previously published [23, 24].

Authorizations were obtained from the EORTC Scientific and Ethical Committees, the ethical committee and legal authorities in France, and local ethical committees at each participating hospital in other European countries. The study was performed in accordance with the Declaration of Helsinki.

Survivors voluntarily participated in the survey and signed informed consent.

The Life Situation Questionnaire (LSQ) addresses issues not available in other validated questionnaires including: socio-demographic data, cohabitation status and highest level of education [25]; parenthood data; education, work, and insurance; health situation including height, weight, and detailed information (checklist) on post-treatment health disorders and current treatments; and social situation [23]. Self-reported health disorders that had occurred after the end of the lymphoma treatment were grouped as follows: cardiovascular, pulmonary, and musculoskeletal disorders; severe infections; anxiety; depression; and history of second cancer. No attempt was a posteriori made to confirm these diseases using data available in either medical records or computerized clinical data.

The MFI questionnaire was used to address the topic of fatigue [20]. It consisted of 20 items, each item coded 1 to 5. From the 20 items, five scales were generated: general fatigue, physical fatigue, reduced activity, reduced motivation, and mental fatigue. Each scale was constructed by summation of its four items; the total obtained was transformed to a linear score ranging from 0 to 100. Zero indicated absence of fatigue and the higher the score, the higher the level of fatigue.

Baseline patient characteristics and treatments administered were retrieved from the clinical databases. Age at survey was obtained by subtracting the date of birth to the date the questionnaires were completed. Follow-up time was obtained by subtracting the date of randomization or the date of first treatment to the date the questionnaires were completed. The weight (kg)-to-height (m²) ratio was used to calculate the body mass index (BMI) at the time of survivorship assessment; obesity was defined as a BMI ≥ 30 kg/m².

Overall, 6665 and 8113 patients were enrolled in the HL and NHL clinical studies, respectively. Of these, 5374 HL (80.6%) and 5051 NHL (62.3%) patients were alive at the time the surveys started. A postal address was obtained for 4038 HL and 3317 NHL individuals. Of these half participated in the survey giving 2032 HL (50.3%) and 1671 NHL (50.4%) cases available for analysis (Fig. 1).

Clinical characteristics, treatment protocols, and clinical outcome of NHL survivors have been recently published [17]; those of HL survivors are under submission for publication in another journal. Demographic characteristics, follow-up time since treatment initiation and medical history as reported by survivors were described using numbers and proportions for HL and NHL separately. Because incidence and initial clinical characteristics and treatment protocols differ between HL and NHL patients, no comparisons were made between the two populations. Fatigue scores at the time of survivorship assessment were first expressed using crude mean and standard deviation for the five dimensions of the MFI assessment tool for HL and NHL separately. Adjusted mean fatigue scores were also estimated using linear regression models with gender and education level, age, cohabitation status and obesity at time of survivorship assessment as covariates. The impact of self-reported health disorders on fatigue level was analyzed within each lymphoma population using adjusted t-test. Statistical tests were two-sided with statistical significance defined as a P <  0.05.

Analysis of fatigue level changes in long-term survivors of HL and NHL was performed on cases with fatigue assessment (at least one dimension score) available. Multivariate linear regression models were used to assess the influence of age and time since diagnosis and treatment as covariates on changes of the five fatigue level scores, i.e. general fatigue, physical fatigue, mental fatigue, reduced activity, and reduced motivation. Variables included in the models were age at the time of survivorship assessment, follow-up time, gender, education level, cohabitation status, obesity and presence of health disorders at fatigue assessment. Primary treatments (including autologous stem-cell transplantation administered upfront in NHL patients) were not included in the models because they did not influence long-term fatigue levels as previously shown [2, 3, 14, 17]. Salvage treatments delivered for a relapse were not considered as well for the same reason. In the results, the intercept (reference score) is the estimation of the mean fatigue score for a male aged 20 years, living with partner, with follow-up time equals zero. For age and follow-up time since treatment initiation the regression coefficient β estimates the change in score associated with a 10-year increase. Cases aged ≥70 years (10% of the population) were grouped because in a previous study focusing on NHL it was shown that fatigue level remained unchanged until 69 years of age and significantly increased beyond [17]. A score can be estimated by simply adding the following terms: intercept + (β_age < 70 x (age-20)/10) + β_{age ≥ 70} + (β_follow-up x follow-up time/10) + β_iV_i (where V_i represents any covariate included in the model and β_i its regression coefficient). Predicted fatigue scores were plotted according to age assuming that cases had been treated at age 45 years. No data being available in lymphoma survivors on minimal (clinical) important difference of fatigue changes based on the MFI, the regression coefficient β estimates were used to test for slopes different from zero.

An attempt was made to compare fatigue level changes with time to general population data in which fatigue level was assessed by use of the MFI instrument [26]. Data consisted of a sample of 1082 individuals (50.3% women; age range, 20 to 79) with equal size aged strata for whom socio-demographic determinants were available such as: education level, cohabitation status, and presence of self-reported health disorders (i.e. somatic or psychological disorders including cancer). Estimations of fatigue levels by age were made with adjustment on gender, education level and cohabitation status. Fatigue levels by age were plotted for individuals without and with health disorders separately. No statistical comparisons were made.

Data were analyzed at the Centre de Traitement des Données du Cancéropôle Nord-Ouest, Plateforme de Recherche Clinique Ligue Contre le Cancer, Centre François Baclesse (Caen, France). All analyses were performed with STATA software (version 14.2; STATA Corp, College Station, Texas 77,845 USA).

As expected, NHL cases were 15 years older than HL cases in average (Table 1). The age also explained the excess of low educated (elementary school) cases and the higher proportion of cases living without partner in NHL population. The number of health disorders reported by the participants at the time of survivorship assessment were similarly distributed in HL and NHL cases, i.e. 61.5 and 64.4%, respectively; those reporting three or more health disorders were 26.0 and 26.9%. However, NHL cases reported twice as many history of second cancer than HL cases. At the time of survivorship assessment HL and NHL cases expressed similar crude levels of fatigue than NHL cases in all dimensions. Levels of fatigue adjusted on gender, age, education level, cohabitation status and obesity were influenced by the presence of health disorders at the time the survivorship assessment was made. HL and NHL survivors reporting health disorders (any types) had significantly higher levels of fatigue than those who did not report health disorders (P <  0.001) (Table 2).

The effects of age and time since treatment (adjusted on gender, age, education level, cohabitation status, obesity and the presence of health disorders) on the five dimensions of fatigue are shown in Table 3. In HL, mean fatigue levels significantly increased from age 20 to 69 for all dimensions except mental fatigue. In individuals aged 70 or older, age increased the mean fatigue levels for physical fatigue, reduced activity and reduced motivation only. Similarly, mean fatigue levels increased with follow-up time: a marked influence was noticed for physical fatigue; it was less important for general fatigue, reduced activity and reduced motivation. In NHL, an increase in mean fatigue levels with increased age from 20 to 69 years was observed for reduced activity. In contrast, physical fatigue level decreased with increasing age until 69 years. In older cases, the effect of age was of the same magnitude in all dimensions except mental fatigue. However, in contrast to HL, follow-up time did not influence fatigue. The combined influence of age and follow-up time on mean fatigue scores are illustrated in Fig. 2. The figures show the predicted mean fatigue scores 5 years and beyond the start of primary treatment for non-obese highly educated male survivors treated at 45 years of age, and living with partner. Main differences between HL and NHL are seen before 70 years of age with fatigue increasing in HL (Fig. 2a) and being stable or decreasing in NHL (Fig. 2b). For example, the negative effect of increasing age (β = − 1.2, Table 3) on general fatigue was more pronounced by follow-up time (β = − 1.3) ending at a slight decrease of fatigue score with increasing age in NHL. Beyond 70 years of age, the curves paralleled whatever the fatigue dimension. The same analyses were repeated on the subgroup of survivors who never relapsed of their disease, and who had a HL, a diffuse large B-cell lymphoma, or a follicular lymphoma. Of the 3642 cases, 514 (14.1%) were excluded either because they experienced a relapse (205 HL and 290 NHL), or were of T-cells histological type (n = 19). Overall, results remained unchanged.

Predicted mean fatigue scores by age were higher for both HL and NHL survivors compared with general population data [26]. Illustrations are given for low educated males living without partner for whom those who survived HL had higher fatigue levels in all dimensions than individuals who survived NHL (Fig. 3a to e). For all scale scores, HL survivors (61.5% with health disorders) displayed changes with age higher than those of the general population with health disorders; in contrast plots for NHL survivors (64.4% with health disorders) were in between those of the general population with and without health disorders except for mental fatigue (Fig. 3e).