Pharmacologic interventions for fatigue in cancer and transplantation : a meta-analysis

Background Our objective was to determine whether, compared with control interventions, pharmacologic interventions reduce the severity of fatigue in patients with cancer or recipients of hematopoietic stem-cell transplantation (hsct). Methods For a systematic review, we searched medline, embase, the Cochrane Central Register of Controlled Trials, cinahl, and Psychinfo for randomized trials of systemic pharmacologic interventions for the management of fatigue in patients with cancer or recipients of hsct. Two authors independently identified studies and abstracted data. Methodologic quality was assessed using the Cochrane Risk of Bias tool. The primary outcome was fatigue severity measured using various fatigue scales. Data were synthesized using random-effects models. Results In the 117 included trials (19,819 patients), the pharmacologic agents used were erythropoietins (n = 31), stimulants (n = 19), l-carnitine (n = 6), corticosteroids (n = 5), antidepressants (n = 5), appetite stimulants (n = 3), and other agents (n = 48). Fatigue was significantly reduced with erythropoietin [standardized mean difference (smd): –0.52; 95% confidence interval (ci): –0.89 to –0.14] and with methylphenidate (smd: –0.36; 95% ci: –0.56 to –0.15); modafinil (or armodafinil) and corticosteroids were not effective. Conclusions Erythropoietin and methylphenidate significantly reduced fatigue severity in patients with cancer and in recipients of hsct. Concerns about the safety of those agents might limit their usefulness. Future research should identify effective interventions for fatigue that have minimal adverse effects.


INTRODUCTION
Cancer-related fatigue is increasingly being recognized as one of the most important symptoms in patients with cancer 1,2 .It has been described as an unexpected tiredness that is more intense and severe than the fatigue experienced in healthy people 3 .Cancer-related fatigue can affect up to 80%-90% of cancer patients, and it can occur before diagnosis, during cancer treatment, and after completion of cancer therapies 1,[4][5][6][7][8][9] .The origin of cancer-related fatigue is multifactorial: it can be a result of the cancer itself, of cancer treatments, and of comorbid medical and psychological conditions 10,11 .Recipients of hematopoietic stem-cell transplantation (hsct) also experience fatigue, likely related to similar underlying mechanisms 12,13 .
Interventions including physical activity and psychological and pharmacologic approaches have been investigated for the management of fatigue in cancer patients, and several systematic reviews have been published [14][15][16][17][18][19][20][21][22] .The evaluation of pharmacologic interventions is particularly important, because medications can be associated with adverse effects and high costs.Thus, a good understanding of the benefits and risks are necessary to guide decision-making.However, the systematic reviews of pharmacologic interventions published to date had restrictive inclusion and exclusion criteria, limiting the number of

METHODS
We followed the prisma (Preferred Reporting Items for Systematic Reviews and Meta-analyses) statement for the systematic review 23 .A search for eligible randomized trials indexed from 1980 to 11 May 2017 was conducted in the medline, medline in-process, embase, Cochrane Central Register of Controlled Trials, cinahl, and Psychinfo electronic databases.The search strategy included mesh terms and text words that identified patients with cancer or recipients of hsct who received an intervention to reduce fatigue.Table i shows the full search strategy.

Study Selection and Data Abstraction
Inclusion and exclusion criteria were defined a priori.Studies were included if participants were adults or children with cancer or recipients of hsct and if the study was a fully published primary randomized or quasi-randomized trial with a parallel-group design that evaluated a pharmacologic intervention for the management of fatigue.
Studies were excluded if fewer than 75% of the participants had cancer or were undergoing hsct, if fatigue was not an endpoint or was reported as an adverse effect, if the intervention was direct cancer treatment, and if fewer than 5 participants were randomized to any study arm.Inclusion was not restricted by language.For the purpose of the analysis, studies were limited to those using a systemically administered pharmacologic agent.Studies using non-systemically administered pharmacologic agents were excluded, as were studies in which only education or advice was provided.
Two reviewers (PDR and SO or LS) independently evaluated the titles and abstracts of publications identified by the search.Any publication considered potentially relevant by at least one reviewer was retrieved in full and assessed for eligibility.Inclusion of studies in this meta-analysis was determined by agreement of two reviewers (PDR and SO or LS).Discrepancies between the two reviewers were resolved by consensus and adjudication by a third reviewer if required (LLD or LS).The kappa statistic was used to evaluate agreement for study inclusion between the two reviewers.Strength of agreement was defined as slight (0.00-0.20), fair (0.21-0.40), moderate (0.41-0.60), substantial (0.61-0.80), or almost perfect (0.81-1.00) 24 .
Data were abstracted in duplicate by two reviewers (DT and PDR) and any discrepancies were resolved by consensus.We contacted authors of manuscripts when publications were missing data for the primary fatigue outcome.

Outcomes
The primary outcome was severity of self-reported fatigue using various fatigue scales.Change scores and end-ofintervention scores were both evaluated.For studies that used more than one fatigue scale, we a priori defined a hierarchy, based on prevalence, for the inclusion of scales in the analysis.Table ii shows the prevalence of the scales reported in our systematic review.The secondary outcome was the severity of selfreported fatigue using the most common fatigue scale (determined after all scales had been categorized).

Intervention and Control Groups
The intervention was any systemically administered pharmaceutical agent.In studies with more than two arms, the least "active" agent (for example, placebo, usual care, or lowest dose) was used as the control group.Where multiple pharmacologic agents were evaluated, the "intervention group" was the highest dose or the most commonly evaluated intervention (determined after all interventions had been abstracted and categorized).
We categorized the control group type as placebo, usual care, or other pharmacologic intervention.

Study Covariates
Study-level variables included age of the participants (adult or child), cancer diagnosis (breast, lung, other single cancer type, or more than one cancer type), inclusion of hsct patients, timing of the intervention (during cancer treatment, after completion of treatment, or both during and after treatment), exclusive enrolment of palliative care patients (as defined by each study), presence of fatigue as an eligibility criterion for enrolment (as defined by each study), and duration of intervention [<8 weeks, ≥8 weeks, or variable (based on median duration reported by each study)].We also evaluated the methodologic aspects of the studies.

Risk-of-Bias Assessment
We used the Cochrane Collaboration tool for assessing the risk of bias in randomized trials 25 .We evaluated sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessors, and attrition bias.Because of their potential effect on bias, adequate sequence generation and adequate allocation concealment were prioritized a priori for the stratified analyses 26 .

Data Analysis
For this meta-analysis, we combined data at the study level and not at the individual patient level.All synthesized outcomes were continuous.For fatigue scores with missing summary measures, we made these assumptions to facilitate data synthesis: the mean can be approximated by the median; the range contains 6 standard deviations; the 95% confidence interval (ci) contains 4 standard errors; and the interquartile range contains 1.35 standard deviations.Where required, instruments were rescaled such that higher scores reflected more fatigue.We synthesized outcomes when data from at least three studies within a stratum were available.
For the primary outcome of severity of fatigue for all fatigue scales, data were synthesized using the standardized mean difference (smd).For the secondary outcome of the most commonly used fatigue scale, data were synthesized using the weighted mean difference (wmd).A smd or wmd less than 0 indicates that the mean fatigue scores were lower (better) in the intervention group than in the control group.Effect sizes were weighted using the inverse variance method.Given an anticipation of heterogeneity between the studies, a random-effects model was used for all analyses.Statistical heterogeneity between the trials was assessed using the I 2 value, which describes the percentage total variation for all studies attributable to heterogeneity rather than to chance.
For the primary analysis, individual pharmacologic intervention groups were compared with all control groups using all fatigue severity scales.Change scores and end-of-intervention scores were both evaluated.Where possible, interventions were also evaluated against placebo.A secondary analysis evaluating the most commonly used fatigue severity scale was similarly conducted.
Potential publication bias was explored by a visual inspection of funnel plots when at least 10 studies were available for synthesis 25 .In the event of potential publication bias, the ''trim and fill'' technique was used to determine the effect of such bias 27 .In that technique, outlying studies are deleted, and hypothetical negative studies with equal weight are created.
Meta-analyses were conducted using Review Manager (version 5.2: Cochrane Collaboration, Nordic Cochrane Centre, Copenhagen, Denmark).All tests of significance were two-sided, and statistical significance was defined as p < 0.05.

RESULTS
Figure 1 presents the flow diagram of study identification and selection.The search strategy identified 11,793 citations, of which 617 were retrieved for full-text evaluation.Within those 617 citations, 117 studies met the eligibility criteria and were included in the systematic review.Figure 1 indicates the reasons for exclusions.Agreement for study inclusion was almost perfect between the two reviewers (kappa: 0.97; 95% ci: 0.95 to 0.99).
Tables iii and iv present the characteristics and details of the 117 included studies, which were conducted in more than 30 countries.Most of the studies (69.2%) were published during or after 2007.All were conducted exclusively in adults; no pediatric patients were included in any study.Breast cancer (15.4%) was the most common cancer   diagnosis studied.Twenty studies (17.1%) were conducted exclusively in the palliative care setting.
Table ii lists all the fatigue assessment scales used in the various studies.The scale most commonly used was the Functional Assessment of Cancer Therapy (fact) 13-item fatigue scale (FACIT.org,Elmhurst, IL, U.S.A.).Of all the studies included in our systematic review, only 35 (29.9%) could be included in any synthesis because of the requirements that an estimate of central tendency (mean or median) and a measure of variability be presented and that at least three studies with such data be included within a stratum.The pharmacologic agents for which synthesizable data were available were erythropoietins, stimulants, and corticosteroids.
Table v shows the effects of the evaluable pharmacologic agents by either change scores or end-of-intervention score.In evaluating erythropoietin, only change scores could be evaluated because too few studies reported end-of-intervention scores for any analysis.Compared with all controls and placebo, erythropoietin significantly improved fatigue.Compared with all controls, its smd was -0.52 (95% ci: -0.89 to -0.14).When the comparison was restricted to studies that reported fatigue using the fact, fatigue was significantly improved in patients receiving erythropoietin compared with all control patients (wmd: -2.98; 95% ci: -4.41 to -1.55).
Table v also shows the effect of stimulants compared with all control treatments and with placebo.As a group, stimulants were not effective for improving change or end-of-intervention fatigue scores.However, when stratified by specific agent, methylphenidate was associated with a significant improvement in fatigue (smd: -0.36; 95% ci: -0.56 to -0.15; and wmd: -2.87; 95% ci: -4.68 to -1.07); modafinil (or armodafinil) was not effective in any comparison.Corticosteroids were not associated with improvement in fatigue (Table v).
Given the small number of studies having data available for synthesis, stratified analyses could not be conducted for l-carnitine, antidepressants, and appetite stimulants.All other agents were examined in only one or two studies, and thus data synthesis was not possible (see Table iv).Figure 2 presents the funnel plot for erythropoietin compared with all controls; no evidence of publication bias was observed.

DISCUSSION
In the present systematic review and meta-analysis, erythropoietin and methylphenidate were found to be associated with significant improvements in fatigue for patients with cancer and for recipients of hsct; modafinil (or armodafinil) and corticosteroids were not found to be effective.Also, despite a very large number of randomized trials, data synthesis was limited.Most interventions were studied only  once or twice; and even for agents that were studied more often, the data could not be synthesized because of limited data reporting from many of the studies.
Erythropoietin was found to be effective in reducing fatigue, but the size of the effect-a wmd of 2.49 compared with placebo according to the fact 13-item fatigue subscale-was small.The minimal clinically important difference for the fact 13-item fatigue subscale has been reported to be 3-3.5 145 , which suggests that, although statistically significant, the observed effect is not meaningful to patients.Combined with concerns about the tumour protection, venothrombotic events, and worse survival potentially associated with erythropoietin 146,147 , that minimal change in outcome suggests that this agent should not routinely be used in clinical practice for fatigue reduction.
The other pharmacologic agent that was found to be effective for fatigue was methylphenidate.However, the wmd of methylphenidate also did not meet the threshold for clinical importance.Further, a Cochrane review of methylphenidate for attention deficit hyperactivity disorder suggested that this agent is associated with an increased risk of non-serious adverse events-sleep problems and decreased appetite being most common 148 .Those issues suggest that methylphenidate should not routinely be used to manage fatigue in patients with cancer and in recipients of hsct, but could selectively be used in specific patients for whom the potential benefits outweigh the disadvantages.
None of the studies found during the systematic review of literature included children.That omission is important, because patients with childhood cancer experience severe fatigue 149,150 and are vulnerable to long-term side effects of treatments 151 .Pharmacologic interventions might not have been applied in children because dosing considerations and safety concerns add complexity.However, future studies should consider the pediatric population when formulating eligibility criteria.
An interesting observation was that, despite the large number of randomized trials, relatively few studies had data available for meta-analysis.Although the fact 13-item fatigue subscale was used in many of the trials, publications were inconsistent in whether they reported fact change scores or end-of-intervention scores.Additionally, many of the studies did not report a measure of central tendency and a measure of variability for either of the two fatigue outcomes (change or end-of-intervention score).The lack of well-reported fatigue data raises potential concerns about a form of publication bias in which negative endpoints are not reported or the data are not shown.Future randomized studies focused on fatigue reduction should be encouraged to explicitly report data that could be combined for analysis in systematic reviews.
The present systematic review complements two previously published meta-analyses evaluating the effects of pharmacologic agents on fatigue in cancer patients 18,152 .Our review adds important insights, given that the review by Mustian et al. 18  The strengths of the present review are its broad eligibility criteria, its inclusion of publications in all languages, and its focus on systemically administered pharmacologic agents.However, our meta-analysis was limited because of the data reporting in the primary studies.Furthermore, wide variations in dose and schedule were noted for the individual pharmacologic agents studied, and the limited number of studies available for synthesis meant that stratified analyses were not possible.

CONCLUSIONS
Erythropoietin and methylphenidate significantly reduce fatigue severity in patients with cancer and recipients of hsct; however, the magnitude of the benefit is of questionable clinical significance.Use of those agents is potentially further limited by concerns about safety.Pharmacologic interventions should not routinely be used to reduce fatigue b FACIT.org,Elmhurst, IL, U.S.A. c MD Anderson Cancer Center, Houston, TX, U.S.A. d MHS Assessments, Toronto, ON.EORTC = European Organisation for Research and Treatment of Cancer; QLQ-C30 = 30-question core Quality of Life Questionnaire; NA = not available.

FIGURE 1
FIGURE 1 Study identification and selection, and reasons for study exclusion.RCT = randomized controlled trial; AE = adverse event; SRs = systematic reviews.
severity.Future meta-analyses should obtain individual data from trials to better understand how pharmacologic interventions affect fatigue.Research is required to identify interventions for fatigue that are effective and have minimal adverse effects.

TABLE II
Self-report fatigue assessment scales used in the included trials a a Some studies used more than one fatigue scale.

TABLE III
Characteristics of 117 studies included in the systematic review e157 Current Oncology, Vol. 25, No. 2, April 2018 © 2018 Multimed Inc.

TABLE IV
Details of the 117 included studies reported many types of interventions, citing 14 studies of pharmacologic interventions that were analyzed as a single group.To inform practice, studies must evaluate pharmacologic agents separately.The review by Minton and Stone 152 , which analyzed specific pharmacologic interventions, is now outdated, being based on a literature search conducted in 2009.

TABLE V
Effect of erythropoietins, stimulants, and corticosteroids on fatigue using all fatigue scales and the FACT scale a