Hobbs 2005.

Study characteristics
Methods	Multi‐centre cluster‐randomised controlled trial involving 50 (computerised) primary care centres across the West Midlands, UK, over a 12‐month period. Randomisation was stratified by levels of deprivation (Townsend quartiles) and practice size. A subsidiary trial embedded in the intervention arm compared 2 different screening strategies. The overall time period was from October 2001 to February 2003
Participants	Male and female patients over 65 years of age attending general practices in the UK Age range was 65 to 98 years; average age was 73.5 years A random sample of 10,000 participants from the intervention group were allocated randomly to systematic or opportunistic screening. Randomisation was stratified according to whether or not atrial fibrillation (AF) had been previously diagnosed to have an equal prevalence of known AF on both arms A random sample of 5000 was selected from the control population. After sampling, lists were returned to practices to remove those who had died, moved or were terminally ill. These patients were replaced from a back‐up list, which had been randomised at the same time as the initial list Final number of participants in control arm = 4963 from 25 general practices Final number of participants in intervention arms = 4933 for opportunistic screening and 4933 for systematic screening from 25 general practices Baseline AF prevalence in the control population was higher than in the intervention populations (7.9% vs 6.9%)
Interventions	Training: Staff at primary care centres in the intervention arms were given training on the importance of AF detection and available treatment options, and were encouraged to consider opportunistic screening of patients. Staff at control centres were given no training. Practice nurses received ECG training before starting ECG clinics Systematic screening: All participants in the systematic screening arm were sent an invitation to attend a screening clinic along with an information sheet. Non‐responders were sent a reminder Opportunistic screening: Participants in the opportunistic screening arm had their records flagged to encourage staff to undertake pulse recordings during routine consultation. Those who had an irregular pulse were given an information sheet and were invited to attend a screening clinic Screening clinics: Screening clinics were run by practice nurses, who took patient histories, checked radial pulse rate and whether it was regular or irregular and recorded a 12‐lead electrocardiogram (ECG). Participants were then asked to complete a questionnaire on the acceptability of the intervention. All 12‐lead ECGs were sent to 2 cardiologists for reporting. If disagreement over the diagnosis arose, a third cardiologist decided. Patients were informed of the results within 2 weeks
Outcomes	Primary outcomes: New cases of atrial fibrillation detected within the 12‐month study period Incremental cost per case detected Secondary outcomes: Cost‐effectiveness of screening in the UK Community prevalence and incidence of AF Acceptability of AF screening and patient uptake
Funding	This research was funded by the NHS research and development health technology assessment programme (No. 96/22/11)
Notes	Intention‐to‐treat analysis was performed, and patients who already had a diagnosis of AF were excluded from the calculation of newly detected cases
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Probably done for control and intervention groups: "After stratification for practice size and deprivation (based on Townsend score), we used MINITAB to select randomly two equal size groups from those practices within a particular stratum. We used a simulated value from a Bernoulli distribution, comprising two values equally likely to occur, to determine which group became the intervention arm (the other being the control arm)" Also probably done for embedded trial within the intervention arm: "We used SPSS to allocate patients randomly from this list to either systematic or opportunistic screening to create two equal size groups of patients within each stratum so that each strategy (systematic or opportunistic screening) had an equal chance of detecting known, unknown, and suspected atrial fibrillation (n=4933). Which group then became the systematic arm (the other being opportunistic) was again decided by using a simulated value from a Bernoulli distribution, comprising two values equally likely to occur"
Allocation concealment (selection bias)	Unclear risk	Study authors state "there was no deliberate concealment of allocation to the trial arms...the trial statistician determined allocation, which was implemented by the trial coordinator". However, clusters (general practitioner (GP) practices) were identified and recruited before randomisation was conducted, so allocation was concealed from the people providing permission for the cluster to be included in the trial. Similarly, patients in the intervention arm were identified and randomly allocated to 2 groups before it was known to anyone involved in the trial which group would be allocated to which treatment (opportunistic or systematic), because this was decided at the end of the randomisation process on the basis of a simulated value from a Bernoulli distribution, comprising 2 values equally likely to occur. However, as no deliberate attempt was made to conceal allocation, it is unclear to what extent a risk of selection bias might have arisen from awareness of practices in the intervention arm that they were in the intervention arm and not the control arm before participants were recruited
Blinding of participants and personnel (performance bias) All outcomes	High risk	It was not possible to blind participants, who were notified by letter that they were being offered the opportunity to participate in an atrial fibrillation (AF) screening clinic or were encouraged to have their pulse recorded during routine consultation. Neither were primary care physicians and healthcare staff blinded because the intervention arm received training during which they were informed of the importance of detecting AF and receiving treatment. Practice nurses at screening clinics who took patients' medical history, pulse and electrocardiogram (ECG) probably were not blinded to whether the patient came from the systematic or opportunistic arm. Blinding is not feasible in a situation where well‐informed patients who need to decide whether they want to avail themselves of screening are a key component of the systematic screening intervention. However, because a study that cannot be blinded is not equal to a blinded study, it is classified as high risk. Screening clinics were used to test patients from each group according to the same protocol and with the aid of a 12‐lead ECG machine (Biolog)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinding was performed when possible; cardiologists who interpreted the 12‐lead ECG reading to make a diagnosis of AF were blinded as to the allocation of the participant from whom the ECG was taken
Incomplete outcome data (attrition bias) All outcomes	Low risk	After random sampling to identify participants from the cluster‐randomised primary care centres, general practices were contacted to exclude people who had died, had moved away or were terminally ill. These exclusions were randomly filled from a reserve list of 10% of practice patients, which was randomised at the same time as the original list. Immediately before sending screening invitations or flagging notes, investigators again contacted general practices to exclude people who had since died, moved or were terminally ill, and these exclusions were not replaced; the numbers in each arm were reported. The primary outcome was calculated by taking the original figure and using an intention‐to‐treat (ITT) approach. Patients within each group who had already received a diagnosis of AF were excluded from the calculation of the primary outcome (new cases detected). This necessitated a review of patient records to identify those with a pre‐existing diagnosis. Records for some people in each group were missing and were reported for each group individually. Both participants with AF and those with missing notes were excluded from the calculation of the rate of new cases detected
Selective reporting (reporting bias)	Low risk	All outcomes specified in the trial protocol were reported
Other bias	Unclear risk	The potential for recruitment bias and contamination is noted in the study. Recruitment bias could have been introduced at the stage where general practitioners were asked to exclude unsuitable patients from opportunistic and systematic screening arms within the intervention group. Advice was given to exclude those from both groups who had died, had moved away or were terminally ill. People excluded at this stage were replaced from a back‐up list of patients that had been randomised at the same time as the groups. No data are provided about how many from each group were replaced at this stage, nor is the breakdown of reasons for their exclusion given. Immediately before the intervention was provided, GPs again were asked to exclude any patients who had died, had moved away or were terminally ill from both groups. Data concerning exclusions at this stage are reported, and a considerable difference in numbers excluded was noted between the 2 arms; 500 were excluded from the systematic screening arm (10% of total) and 195 from the opportunistic arm (4% of total). However, individual reasons for exclusion from the systematic screening arm were reported, and only a small minority of these (9 people, 0.2% of total) were deemed unsuitable, as opposed to having died or moved away (491 people, 9.9% of total). An ITT analysis included participants removed from the intervention group at this stage in the calculation of the primary outcome