Referrals
Initially we accepted referrals from GPs and psychiatrists only. To be more inclusive, and to improve recruitment into the trial, on 15 May 2009 we started to accept referrals from other healthcare professionals including primary care liaison workers, social workers and mental health nurses.
To improve recruitment in the Swansea centre, we amended our protocol to accept self referrals from people already taking antidepressants. From 26 June 2010 we recruited participants through advertisements in the community, including posters, local newspapers, pharmacies, clinics and general practices. To help with the expected increase in referrals we gained permission for registered mental health nurses to screen people who refer themselves to the study.
Our original protocol stated that we would reimburse general practices £50 to cover administration costs for every patient recruited. In November 2010 we gained permission to extend reimbursement to secondary care teams.
Recruitment
We initially planned to complete all randomisations by August 2008. However unforeseen delays in research governance delayed the start of recruitment by 6 months for reasons including: change of sponsor; delays in appointing researchers; delays in obtaining regulatory approval; and delays in obtaining honorary contracts. We also encountered obstacles to recruitment including: shortage of psychiatrists to screen patients; fewer referrals from secondary care than expected; fewer referrals of newly diagnosed patients than expected; limited space in general practices for screening; and difficulty in contacting referred patients. The NIHR HTA programme therefore granted us a 24-month costed extension to extend recruitment by 18 months.
On 25 June 2009 we gained permission to extend the Wrexham centre to include Conwy and Denbighshire.
Exclusion criteria
In July 2007 we removed substance misuse from the list of exclusion criteria for three reasons:
to make the trial more pragmatic and inclusive
to increase recruitment, and
because we could detect folate deficiency from alcohol use through blood tests at screening.
We added malignancy and related conditions like intestinal polyposis as exclusion criteria, following advice from the TSC in January 2007 based on evidence from rat studies that high folate intake may increase growth of existing tumours.
In response to our original application for a CTA the MHRA asked us to add taking lithium as an exclusion criterion, which we did. However we found no evidence of adverse interactions between lithium and folate. Furthermore we identified that this exclusion criterion was impeding recruitment into the trial, particularly from secondary care. In April 2008, we therefore amended the exclusion criterion for taking lithium to exclude patients with bipolar disorder whether taking lithium or not. This ensured that only patients with unipolar depression entered the trial.
Data collection
It had been our intention for the FolATED trial to be paperless. However IT constraints led us had to substitute a rigorous paper-based system to collect data and we amended the protocol accordingly.
We identified a need for researchers to make home visits to conduct the research assessments. Depression can often make it difficult for patients to attend hospital and many potential participants had requested that researchers go to their homes for appointments. In May 2009 we amended the protocol to allow researchers to visit homes when needed.
Assessment and follow-up
Following the initial application for approval in October 2006, we received further requests from the MREC, local R&D Departments and Trial Management Group to amend the protocol, including:
taking extra samples of blood for homocysteine analysis, to guard against losing samples in the post
collecting red cell folate at screening to ensure that we did not randomise patients with folate deficiency
collecting serum folate at weeks 12 and 25 to check for compliance
measuring plasma vitamin B12 concentrations at minus 2, 12 and 25 weeks, to guard against participants becoming B12 deficient during treatment with folic acid and resulting in a neuropathy, and
asking patients at each follow-up whether they are taking additional supplements.
We found that the initial screening appointment with patients took longer than expected owing to the number of assessments. In April 2008 we gained permission to remove the majority of the assessments including the MADRS, UKU, SF12, EQ5D and CGI from this appointment. The full battery of assessments at randomisation, which provided the true baseline, continued as planned. This increased the number of patients be screened by psychiatrists and reduced the burden on participants.
After the MHRA recommended following up any pregnancy that occurs during a trial, we gained permission in April 2008 to alter the information sheet to allow us to track any such pregnancy.
On 17 September 2008, the TSC identified the need to ask participants explicitly about attempts at self-harm. Though we define self-harm as a SAE, this relies on participants reporting it without prompts. We therefore added the question: ‘Since we last saw you, have you tried to harm yourself? For example, have you taken any extra tablets or cut yourself or done anything else to injure yourself?’
Additional analyses
In December 2009 we gained permission to investigate the interaction between folic acid and MMA, a metabolic marker of the enzymatic function of vitamin B12 and a more sensitive marker of B12 status. High-dose folate is generally considered safe, provided that B12 deficiency is excluded.162 However, there is evidence that high folate levels combined with low B12 levels are associated with significant cognitive impairment in the elderly.163 Furthermore the US National Health and Nutrition Examination Survey observed a relationship between folic acid and MMA (57). The FolATED trial offers a rare opportunity to observe the effects of folic acid supplementation on MMA concentration (not reported in this monograph).
To ensure the safety of participants in the trial fieldworkers informed their Principal Investigator (PI) of any participant who attempted suicide or was at high risk. For this purpose they completed the MINI suicidality scale at the end of the MADRS. If there was a marked increase in suicidal risk, they reported this as a SAE. Though we had approval to monitor suicidality through our safety reporting system, we did not name the MINI suicidality explicitly in our protocol. The MREC therefore asked us not to analyse these data. However our TSC were concerned that not to report fully on the safety of our intervention would be unscientific. So they recommended that we seek permission to publish this information in the public domain. We therefore gained permission in May 2011 to publish analyses comparing differences between the two treatment groups in MINI suicidality scores at follow-up.
Clarification of protocol
Following meetings of TSC and Data Monitoring Committee early in 2007 we replaced the lay summary with a technical abstract at the beginning of the protocol.81 We elaborated on screening, informed consent, randomisation, withdrawal and safety reporting. We improved clarity, not least by integrating the flow diagram and outcome measures table into the main text of the protocol.81,82
Review of power calculation (approved by DMEC, TSC and NCCHTA)
We originally powered FolATED to detect a difference between the two treatment groups of three points on the Beck Depression Inventory (BDI-II) at 25 weeks, judging that a clinically important difference. As we estimated the SD of BDI-II scores in the trial population at 10.7, our protocol proposed a completed sample size of 400 at 25 weeks to yield 80% power to detect this difference using a significance level of 5%. As interim analysis of baseline BDI-II scores showed that their SD was about 10, we revised the target completed sample size to 358 at 25 weeks. The original protocol allowed 10% loss at each of the three follow-up assessments, thus requiring a randomised sample of 549 to achieve 400 completers at 25 weeks. As interim analysis also showed that retention at 25 weeks was 79% rather than the 73% expected, the new target of 358 completers needed a randomised sample of 453 participants.
