Results of the search

The results of the search are summarised in Figure 1. A total of 9200 records were identified after removal of duplicates and first assessment. One additional paper was identified after contacting a study author (Johansson 2015). A total of 676 papers were reviewed for suitability for inclusion, of which four studies met the criteria for inclusion. Two cross‐over studies could not be included as they did not report separate phase one outcome data, which the authors were unable to provide when contacted (Speech 1993; Tenovuo 2009). One study did not include a placebo control group (Johansson 2015). One study investigating rivastigmine was identified to be ongoing, and is described in the Characteristics of ongoing studies table.

Figure 1

---

---

Study flow diagram.

Included studies

The four studies meeting inclusion criteria are described in the Characteristics of included studies tables. One study investigated modafinil (Jha 2008), a wake‐promoting agent with effects on multiple neurotransmitter systems, including histaminergic, serotonergic, and glutaminergic activity that may be secondary to stimulation of catecholamine systems (Minzenberg 2008). The second investigated (−)‐OSU6162 (Johansson 2012), a monoamine stabiliser agent with dopaminergic (Lahti 2007), and serotonergic effects (Burstein 2011; Carlsson 2011). The third investigated atomoxetine (Ripley 2014), a noradrenaline reuptake inhibitor (Zerbe 1985). Lastly, Silver 2006 investigated rivastigmine which is an acetylcholinesterase and butyrylcholinesterase inhibitor (Giacobini 2002). Three studies (Jha 2008, Ripley 2014 and Silver 2006) were conducted in the United States, whilst Johansson 2012 was conducted in Sweden.

Jha 2008 is a single centre randomised placebo‐controlled crossover trial, examining the efficacy of modafinil in the treatment of fatigue or excessive day‐time sleepiness, which included secondary measures assessing global and cognitive function. Fifty‐one men and women aged 16 to 65 were recruited from an inpatient rehabilitation unit. For inclusion, participants were required to suffer from fatigue or excessive daytime sleepiness that compromised functioning. Although the mean baseline ImPACT (Immediate Post Concussion Assessment Cognitive Testing) score was provided for the group this is difficult to interpret. The ImPACT is a neuropsychological test which compares the subject against aged‐matched normative data. It would therefore be important to know the baseline for each individual to ascertain the degree of cognitive impairment at baseline. Relevant exclusion criteria were: the presence of a neurological or neuropsychiatric diagnosis that would impair evaluation of the intervention; other diagnoses that may cause excessive daytime sleepiness; concurrent medication use; clinically significant systemic medical conditions; epilepsy; cardiovascular disease; severe hepatic or renal impairment; psychiatric or behavioural disturbance that would impair evaluation of the intervention; pregnancy. In the treatment arm, modafinil was titrated to 200 mg twice daily orally over a two‐week period and continued for a further eight weeks. Participants in the placebo arm were administered tablets at the same frequency as for modafinil. After a four‐week washout period participants were switched to the opposite arm. The primary outcome measures of the study related to fatigue and daytime sleepiness. Relevant cognitive or global outcome measures included the Medical Outcome Study 12‐Item Short Form Survey (SF‐12), ImPACT, and the Conners' Continuous Performance Test II (CPT‐II). The measures were administered at baseline, week four and week ten of treatment (repeated at week four and ten following cross‐over). Only results from the first treatment phase were included in this review. We contacted the study authors who provided unpublished phase one outcome data for cognitive outcomes and adverse effects.

Johansson 2012 is a single centre cross‐over randomised placebo controlled study, examining the efficacy of (−)‐OSU6162 on long‐term mental fatigue following TBI or stroke. Twelve men and women aged 30 to 65 were recruited through a local Swedish newspaper advertisement or from the Department of Neurology at Sahlgrenska University Hospital in Gothenburg. Participants were included if they were at least one year post‐TBI or stroke, but not more than 10 years, and had suffered pathological mental fatigue for at least one year. Participants were required to have otherwise recovered from neurological symptoms, have a Glasgow Outcome Scale (Extended) (GOS‐E) score indicating at least moderate disability or better (> 5), and have completed education to high school equivalent and speak Swedish. Participants were excluded if they were deemed to suffer significant cognitive impairment, as measured by scores more than two standard deviations worse on the cognitive tests used by the study, or in the case of significant neurological or psychiatric comorbidity, substance misuse, or pregnancy. Participants receiving (−)‐OSU6162 were administered 15 mg twice daily in week one, 30 mg twice daily in week two, and 45 mg twice daily in weeks three and four. Dose variation was possible if a participant had responded at a lower dose and then experienced decreased efficacy or adverse effects at a higher dose, therefore permitting a reduction to the previous dose. The process of administration in the placebo arm was not explicitly described. The primary outcome measure of the study assessed fatigue. The relevant secondary cognitive outcome measures were the Trail Making Tests A,B,C & D, WAIS‐III digit symbol coding, WAIS‐III digit span and the FAS verbal fluency test. These were administered at baseline and four weeks. A week 8 assessment was conducted following the crossover, the results of which are not relevant to this review. We contacted the study authors, who provided unpublished outcome data specific to TBI participants in phase one of the trial.

Ripley 2014 is a single centre randomised placebo‐controlled cross‐over trial, examining the efficacy of atomoxetine in the treatment of attention impairment. Sixty men and women aged 18 to 65 were recruited by posting flyers to people who had received inpatient rehabilitation following TBI, or through clinicians handing leaflets to people with TBI in the community. Flyers were also distributed to various brain injury organisations. For inclusion, participants were required to have a history of moderate to severe TBI, defined by a Glasgow Coma Scale score of 12 or less, post‐traumatic amnesia which lasted more than 24 hours or radiographic evidence of intracranial trauma. Telephone screening was conducted to administer the Adult ADHD Self‐Report Scale (ASRS‐v1.1) and the Cognitive Failures Questionnaire (CFQ). The Other CFQ was also administered if a primary carer or significant other was available. A score of four or higher on the ASRS, 0.35 or higher on the CFQ, or 0.42 or higher on the Other CFQ was required for inclusion. Exclusion criteria were: non‐English speaking participants; current seizure history; cardiovascular disease; history of psychiatric illness requiring hospital treatment; use of monoamine oxidase inhibitors; severe renal or hepatic impairment; pregnancy or lactating. After baseline assessment all participants received a placebo run‐in for two weeks. In the treatment arm, atomoxetine 40 mg was administered twice a day for two weeks. Participants in the placebo arm were administered tablets of identical appearance according to the same schedule. After a two‐week placebo washout period participants were switched to the opposite arm for a further two weeks. The primary outcome measure of the study was the Power of Attention factor of the Cognitive Drug Research (CDR) Computerized Cognitive Assessment System. The CDR is a computer‐controlled battery of tests that provides factor scores for Power of Attention (POA), Continuity of Attention (COA), Quality of Working Memory (QWM), Quality of Episodic Memory (QEM), and Speed of Memory (SOM). Secondary measures in the study included COA, QEM and SOM factor scores of the CDR. Further secondary measures were the Efficiency of Attention score calculated from the Power of Attention score, Stroop Colour and Word Test, the Adult ADHD Self‐Report Scale and Neurobehavioral Functioning Inventory scores. Only results from the first treatment phase were included in this review. The study authors provided unpublished phase one outcome data for cognitive outcomes and adverse effects on request.

Silver 2006 is a multicentre randomised placebo controlled trial examining the efficacy of rivastigmine on cognitive function. One‐hundred and fifty men and women aged 18 to 50 were recruited across 19 sites. Participants with cognitive impairment following non‐penetrating traumatic brain injury were included. At baseline, participants were required to have a difference of at least one standard deviation between current estimated intelligence measured by the Wechsler Adult Intelligence Scale, third edition (WAIS‐III) and current verbal attention or verbal memory functioning measured by the Cambridge Neuropsychological Test Automated Battery Rapid Visual Information Processing A subtest (CANTAB RVIP A) or the Hopkins Verbal Learning Test (HVLT) total trials one to three. Relevant exclusion criteria were: medical, psychiatric or substance use disorders that could impair evaluation of the intervention; acute or severe pulmonary or cardiovascular disease; primary neurodegenerative disorders; epilepsy; medication known to affect cognitive functioning in TBI; history of major brain surgery or penetrating brain injury. Participants receiving rivastigmine started at 1.5 mg twice daily for four weeks, after which the dose was increased to 3 mg twice daily. If not tolerated due to adverse side effects, the dose could either be reduced to 4.5 mg or 3 mg daily (also in divided doses). A matching placebo was administered according to the same schedule as for rivastigmine. Primary outcome measures were the CANTAB RVIP A and the HVLT. These were administered at baseline and weeks four, eight and twelve. Secondary outcome measures were only administered at baseline and week twelve, and included; Controlled Oral Word Association; WAIS‐III Digit span; WAIS‐III Letter Number Sequencing; and Trail Making Tests Parts A and B. We contacted the study authors for additional data, which was unavailable. Therefore, outcomes for the Neurobehavioral Functioning Inventory (NFI), Diener Satisfaction with Life Scale, and Clinical Global Impression of Change (CGI) could not be reported.

Excluded studies

We excluded two cross‐over studies as neither reported phase 1 outcome data, which the authors were unable to provide when contacted (Speech 1993; Tenovuo 2009). We excluded three other studies as the time since TBI was not clearly stated (León‐Carrión 2000; Plenger 1996; Schneider 1999). One study (Johansson 2015) was excluded as the control group did not receive a placebo.

Allocation

Jha 2008 and Ripley 2014 described adequate procedures to reduce selection bias and were rated as low risk. Johansson 2012 did not describe the randomisation process, other than reporting that an external system was employed; therefore we rated the study as unclear risk for random sequence generation, and low risk of allocation concealment due to the use of an external process. A limited description of the randomisation and concealment process was provided by Silver 2006; therefore we rated the risks of bias as unclear.

Blinding

All studies were assessed to be of low risk of bias.

Incomplete outcome data

We assessed risk of attrition bias as high in Jha 2008 as not all participants who were randomised were included in the analysis and reference was not made to intention‐to‐treat analysis. We assessed Johansson 2012, Ripley 2014 and Silver 2006 as low risk.

Selective reporting

Outcomes were reported as planned in the protocol by Jha 2008, Johansson 2012, and Ripley 2014 , and therefore we assessed reporting bias as low risk.

Silver 2006 varied from the protocol in terms of inclusion criteria and outcome measurement time‐scales. The published protocol registered with ClinicalTrials.gov planned to include participants aged 18 to 65, while the published study states the inclusion criteria as 18 to 50 years. Of more concern, a different cognitive battery was used in the study to that proposed in the protocol to determine inclusion into the trial. The protocol included subjects one standard deviation or more below the mean on the California Verbal Learning test, Tower of London test, Verbal Memory Learning test and the Test Battery for Attentional Performance. However, the completed study included subjects who performed below one standard deviation on the Verbal subtest on the Wechsler Adult Intelligence Scale (WAIS) and attention or verbal memory as assessed by the Cambridge Neuropsychological Test Automated Battery CANTAB), Rapid Visual Information processing (RVIP), or the Hopkins Verbal learning Test (HVLT). Also, the protocol planned for treatment for 20 weeks, as opposed to 12 weeks in the completed study. Both these discrepancies appear to have been balanced across both arms of the study: we therefore assessed the risk of bias as low.

Other potential sources of bias

Only Ripley 2014 and Silver 2006 investigated treatment of cognitive impairment as a primary outcome measure, with Jha 2008 and Johansson 2012 primarily investigating treatment of fatigue. Limited information on the severity of TBI was provided by the studies, and similarly limited data on baseline cognitive impairment at inclusion were provided. These factors may have introduced bias through variation in the recruitment or selection of participants, and consequently assessment of outcome.