events and deterioration reported by participants in the PACE trial of therapies for chronic fatigue syndrome

Objective: Adverse events (AEs) are health related events, reported by participants in clinical trials. We describe AEs in the PACE trial of treatments for chronic fatigue syndrome (CFS) and baseline characteristics associated with them. Methods: AEs were recorded on three occasions over one year in 641 participants. We compared the numbers and nature of AEs between treatment arms of specialist medical care (SMC) alone, or SMC supplemented by adaptive pacing therapy (APT), cognitive behaviour therapy (CBT) or graded exercise therapy (GET). We examined associations with baseline measures by binary logistic regression analyses, and compared the proportions of participants who deteriorated by clinically important amounts. Results: Serious adverse events and reactions were infrequent. Non-serious adverse events were common; the median (quartiles) number was 4 (2, 8) per participant, with no significant differences between treatments (p = 0.47). A greater number of NSAEs was associated with recruitment centre, and baseline physical symptom count, body mass index, and depressive disorder. Physical function deteriorated in 39 (25%) of participants after APT, 15 (9%) after CBT, 18 (11%) after GET, and 28 (18%) after SMC (p < 0.001), with no significant differences in worsening fatigue. Conclusions: The numbers of adverse events did not differ significantly between trial treatments, but physical deterioration occurred most often after APT. The reporting of non-serious adverse events may reflect the nature of the illness rather than the effect of treatments. Differences between centres suggest that both standardisation of ascertainment methods and training are important when collecting adverse event data.


Introduction
Clinical trials frequently attribute health problems that arise during a trial to the intervention. But, when health problems typically remit and relapse, the attribution of all new health problems to the intervention may be misleading. This study aims to explore this issue in patients with chronic fatigue syndrome (CFS) who participated in a treatment trial.
Adverse events reported by participants in clinical trials of treatments may be considered to be clinically serious or not, and to be reactions to trial treatments or not. Few studies have examined the associations and predictions of adverse events in trials. Several trials have suggested a relationship between the reporting of adverse events and negative affect; anxiety [1] depression [2] and neuroticism [3].
Female and introverted participants of phase 1 medical trials are more likely to report adverse events than males and extroverts [4]. Physical symptoms at baseline predicted having a treatment related adverse reaction in an antidepressant controlled trial [5]. As well as this small literature regarding adverse events in trials, there are well established associations between reporting physical symptoms, outside of trials, and both mood disorders [6][7][8][9][10] and symptom burden [11].
Chronic fatigue syndrome (CFS) is characterised by long-standing disabling fatigue and other symptoms that have no alternative medical or psychiatric explanation [12]. Its nosological status and aetiology are uncertain [13]. CFS is associated with functional somatic syndromes such as irritable bowel syndrome and fibromyalgia [14]. Treatments recommended by the National Institute of Healthcare and Clinical Excellence (NICE) include cognitive behaviour therapy (CBT) and graded exercise therapy (GET) [15], but patient organisations have expressed concern about their efficacy and safety [16].
The PACE trial was a four arm randomised trial, which was designed to compare three therapies each added to specialist medical care (SMC) against SMC A C C E P T E D M A N U S C R I P T ACCEPTED MANUSCRIPT 4 alone to determine both efficacy and safety [17]. The trial found that two therapies, CBT and GET, were more effective than adaptive pacing therapy (APT), when any of these therapies were added to SMC, and were more effective than SMC alone [17].
While CBT and GET were designed to be rehabilitative, the goal of APT was to optimise adaptation to the illness by planning and pacing activities to avoid or reduce fatigue [17]. The trial measures of safety included systematic assessments of adverse events (AE), which occur uncommonly in trials of behavioural interventions [18]. We have already reported that there were few serious adverse events (SAEs) and even fewer serious adverse reactions (SARs), the numbers of which did not differ significantly across treatment arms [17]. We have also reported various measures of deterioration, but not whether there are any differences across treatment arms in the proportions of participants who deteriorated in the two primary outcomes by a clinically important amount [17]. This paper reports the more commonly reported non-serious adverse events (NSAEs), compares their frequency between treatment arms, and also identifies baseline factors associated with reporting larger numbers of NSAEs [17,19]. On the basis of the previous literature, we hypothesised that NSAEs would be associated with female sex, a larger number of physical symptoms at baseline, and both depressive and anxiety disorders present at baseline. To our knowledge there has been no previous study examining associations of NSAEs in a trial of treatments for CFS or functional somatic syndromes.

Outline of the PACE trial
This report uses data from the PACE trial, relevant aspects of which are described; more comprehensive accounts are available in the protocol [19], and the primary paper [17]. The PACE trial recruited 641 patients from secondary care clinics with a diagnosis of CFS, using the Oxford criteria, which require six or more months of A C C E P T E D M A N U S C R I P T ACCEPTED MANUSCRIPT 5 disabling fatigue, with fatigue being the principal symptom, and no alternative, explanatory diagnosis [20]. Participants were randomly allocated to one of four treatment arms consisting of specialist medical care (SMC) alone or SMC with one of APT, CBT or GET. Randomisation to the four treatment arms was stratified by centre, co-morbid depressive disorder, and different CFS and myalgic encephalomyelitis (ME) criteria [12,21] Inclusion criteria were meeting Oxford research diagnostic criteria for CFS [20], a Chalder Fatigue Questionnaire binary score of 6 or more [22], a SF-36 physical function sub-scale score of 65 or less [23] and age at least 18 years old.
Exclusion criteria were a significant risk of self-harm, being considered by the RA to be unable to participate in the trial, participation in the PACE trial being inappropriate for clinical needs, and patients who had previously attended a PACE centre specialist fatigue clinic and received a course of PACE trial consistent treatment [19].
The Structured Clinical Interview for DSM-IV was administered by the RA, after appropriate training, and used to assess psychiatric comorbidity and psychiatric exclusions [24]. Further baseline information collected included demographic details, current membership of a local or national ME self-help group, and body mass index (BMI). Additional self-report questionnaires included the Chronic Disease Self- Efficacy measure [25], physical symptoms (Patient Health Questionnaire; PHQ-15) [26], Cognitive Behavioural Responses Questionnaire (CBRQ) [27], Jenkins sleep scale of subjective sleep problems [28], and the Hospital Anxiety and Depression Scale (HADS) [29]. Further assessments consisted of the International (CDC) criteria for CFS [12], the London criteria for myalgic encephalomyelitis [21] and presence or absence of fibromyalgia [30].

Assessment of adverse events
Follow-up assessment interviews were conducted by the RA at each centre on three occasions: 12, 24 and 52 weeks after randomisation. At each of these time points the RA asked participants whether any new events or illnesses had taken place since the last assessment including any events for which the participant visited the GP or hospital department, or took medication. [19] AEs were also recorded by treating specialist doctors and therapists if spontaneously reported to them during the trial. An AE was defined as 'any clinical change, disease or disorder experienced by the participant during their participation in the trial, whether or not considered related to the use of treatments being studied in the trial' [19]. We did not examine inter-rater reliability between RAs since we did not foresee variability in these assessments. as serious, the scrutineers were unmasked to treatment allocation to establish whether or not the event was a serious adverse reaction (SAR). A serious adverse reaction was considered to be a reaction to one of the supplementary therapies or a drug prescribed as part of SMC [19]. All those judged as definitely, probably, or possibly related were considered to be SARs.
A non-serious adverse event (NSAE) was any health event, which was not categorised as an SAE or SAR. Each NSAE was ascribed to the appropriate body system (gastroenterological, neurological, etc.) independently by two senior medical clinicians (one a consultant infectious diseases physician, the other a consultant liaison psychiatrist; both experienced in CFS), who were different from the independent scrutineers. NSAEs attributed to CFS (i.e. considered to be a symptom of CFS) were put into a separate category since there is no consensually agreed body system for CFS, and because of specific interest in these symptoms.
Differences in clinicians' ratings were resolved by discussion until consensus was reached. To summarise, adverse events were any new health related event reported by the participant in any context. These were independently judged as serious adverse events, using an a priori guideline of seriousness, and as serious adverse reactions if independently judged to be a reaction to a trial intervention.

Statistical analysis
Firstly Thirdly, all associated univariate variables, significant at p≤0.1, were entered into a multivariate binary logistic regression model for all NSAEs, followed by a separate regression analysis for NSAEs attributed to CFS. Age, sex and treatment arm were also entered into all models. We also modelled those with one or more NSAEs versus those without any, using a logistic regression analysis to establish characteristic differences between these two groups of patients.
In order to provide further checks on the relatedness of adverse events in general to trial treatments, we compared both serious adverse events and reactions, and the numbers of participants in each treatment group who reported being "much worse" or "very much worse" in their overall health at 52 weeks after randomisation. [31] We also compared the numbers of participants in each treatment arm who had deteriorated by more than a clinically important difference (at least 2 points on the fatigue questionnaire and/or at least 8 points on physical function, which represented 0.5 of a standard deviation of baseline outcome measures) between randomisation and 52 weeks later [17]. We then examined the number of NSAEs in those who had deteriorated by either of the latter measures. We analysed the data using SPSS v18 and v22.
As a post hoc analysis, in order to better understand the differences in NSAE counts between centres, we stratified centres into three groups: low (3 centres), medium (2 centres) and high (1 centre) numbers of NSAEs per participant. Using these strata, we undertook a one way analysis of variance (ANOVA) of those continuous variables that showed statistically significant differences by NSAE count on univariate analyses. We examined linear trends across centre strata. Significant associations were also found between reporting more NSAEs and having any psychiatric disorder, particularly depressive disorder, dysthymia, and major depressive disorder (table 4). There were no significant associations with anxiety disorders.

Multivariable binary logistic regression of NSAEs
When modelling all NSAEs, using a median split of 4 or less versus 5 or more, centre effects dominated the models, so we remodelled both with and without centre. Adding centre replaced baseline depressive disorder with baseline major depressive episode (p = 0.03) and added physical symptom count (p = 0.09) as well as CFS symptom count (p = 0.006).

Post hoc exploration of centre effects on NSAE count
Seven of ten variables measured at baseline were significantly correlated in a linear trend with centres stratified by NSAE count in the ANOVA (appendix table B).
However, the variation between mean scores per centre varied little, with the most significant different variables of Chalder fatigue, SF36, and PHQ15 varying by 2, 6 and 2.8 points respectively. There were no significant correlations for age, sex, duration of illness, and embarrassment scores. Table 5 shows that there were no statistically significant differences between treatment arms in those who had an SAE, SAR, or in those who had deteriorated either by CGI score (of either "much" or "very much worse"). This was also the case for fatigue alone and fatigue and disability combined. Similarly, there were no significant differences across treatment arms between the median number of NSAEs in those who had deteriorated as measured by the global impression change score (p = 0.97) and those reporting deterioration in both primary outcomes of fatigue and physical disability (p = 0.16) between treatment groups. However, there was a significant difference in deterioration of physical function, across treatment arms, with a quarter of those who received APT deteriorating, compared to 18% after SMC, and 11% and 9% after GET or CBT respectively.

Discussion
There were no important differences between treatment arms in any of the adverse events, however they were measured or classified. Most importantly there was no evidence of more frequent adverse events after either CBT or GET. The factors associated with a higher number of NSAEs were the centre where the participant was seen, followed by the number of physical symptoms at baseline, having a depressive episode, and higher body mass index. Those variables associated with CFS related NSAEs were centre, CFS symptom count and a depressive episode.
The common baseline associations in both models were centre, depressive disorder and physical symptom count. Those who received APT were most likely to deteriorate by a clinically important amount in physical function, with those in receipt of CBT being least likely to deteriorate.
The substantial variation in the frequency of the reporting of NSAEs between centres is our most unexpected finding, although variation between centres is not uncommon in multi-centre trials [32], and this did not influence treatment response [17]. We found statistically significant linear associations between centres stratified by NSAE counts and a number of baseline variables, but the differences between centres were small, and nothing like the size of differences in NSAE frequency across centres. ThIs suggests that the large differences in NSAE numbers between centres is unlikely to be related to the small differences found between centres in baseline factors. Although the research assessments across centres were standardised and training was provided at the start of the trial, it might be that the differences were due to different methods of ascertainment. This apparent variation in recording NSAEs, despite a standard protocol for doing so, has important implications for recording adverse events in future trials.
Having more symptoms at baseline, particularly those associated with CFS, predicted subsequent NSAEs in general and also NSAEs attributed to CFS. This

ACCEPTED MANUSCRIPT
14 replicates previous work [5]. Higher symptom counts are associated with somatoform disorders in secondary care [33], and may reflect a general tendency to report symptoms, which is associated with, but also independent of, mood disorders [34].
The specificity of CFS symptoms at baseline being associated with NSAEs attributed to CFS suggests a specific tendency to report these symptoms, rather than a generic influence of reporting any symptom. It may also reflect the relapsing and remitting nature of CFS.
Our finding that a diagnosis of a depressive disorder at baseline predicted increased reporting of NSAEs is consistent with previous studies that found negative affect to be associated with NSAEs specifically [1,3], and somatic symptoms in general [6][7][8][9][10]35]. This association remained significant for both NSAEs as a whole and for CFS attributed NSAEs in one regression model. Unlike some previous studies, we did not find an association with anxiety, either with the HADS score or through the SCID interview. One other trial failed to find an association between anxiety and adverse events [5].
We found that a higher BMI was associated with NSAEs in general. This observation may have several explanations; obese people generally report both more physical and mental health related problems [36], and our sample included 123 (19%) participants who were morbidly obese. We were not able to replicate a previous research finding that female participants are more likely to report adverse events [4].
The strengths of this paper are that it used data from a large trial from multiple centres. The assessment of NSAEs on three occasions improved sensitivity.
The limitations include the difference in frequency of NSAE reporting between centres, implying variation in ascertainment, although controlling for centre did not significantly affect our main findings. We only measured deterioration using selfratings, rather than objective measures. We were unable to model the full distribution of NSAEs, which may have limited the power of our regression models.

Conclusions
We found that there were no important differences in any of the adverse events between treatment arms, and no excess associated with either CBT or GET.
Clinically important deterioration occurred least often after CBT and GET; APT may be associated with more frequent deterioration in physical functioning. We also noted that the reporting of non-serious adverse events in a clinical trial of treatment for CFS