Reducing the time before consulting with symptoms of lung cancer: a randomised controlled trial in primary care

Participants

Participants were long-term smokers (at least 20 pack-years) aged ≥55 years, including ex-smokers if their cessation date was within 10 years. Smokers and ex-smokers were identified from practice computerised records and a sample selected (every nth name from an alphabetical list, where n = [number eligible]/[number required]); these individuals were sent a letter from their general practice, inviting them to take part in the study. Two waves of recruitment were conducted, the first to establish recruitment rates and the second to achieve the target sample size.

Intervention

A consultation with a trained nurse guided each participant through a self-help manual, which the participant took home, and developed ‘if–then’ action plans (full details are published separately).10 The manual provided information, used behaviour-change techniques and sought to: engage high-risk individuals, using logos, celebrity endorsement by Liz Dawn (from Coronation Street), and ‘special attention’ messages; increase the salience and personal relevance of symptoms; reinforce the benefits of early intervention in lung cancer and other chest diseases using patient stories and frequent positive messages; sanction early consultation using messages from doctors, other patients, and Liz Dawn; provide prompts to self-monitoring; provide simple symptoms checklists linked to ‘if–then’ action plans; and tackle barriers to consultation using ‘if–then’ coping plans.16,17 Participants allocated to the intervention group who did not attend the nurse consultation were sent the self-help manual by post. The intervention was appraised at, and refined after, focus groups and interviews with patients from the target group, GPs, and individuals with lung cancer. The control group received usual care at their general practice, which included patient-initiated consultations, opportunistic smoking-cessation advice and, if applicable, annual reviews for chronic obstructive pulmonary disease (COPD).

Outcomes

The primary outcome was the number of general practice consultations, within target times, for individuals with new chest symptoms (≤3 days for haemoptysis and ≤3 weeks for other symptoms) in the year after the start of intervention. However, data on symptom duration were missing for most consultations, so the primary outcome was analysed in two parts: first, number of general practice consultations with new chest symptoms and, secondly, the proportion of those who consulted (and for whom duration data were available) that were within the target time. The secondary outcome was intention to consult with symptoms by a given time. This was assessed using four items: intervention, self-efficacy, knowledge, and mood; each requiring forced-choice direct estimation of the time taken before making an appointment to see a doctor for a given chest symptom scenario. Each item had 11 options ranging from <1 day to >6 months. Internal consistency (Cronbach’s alpha) was 0.78. These were converted into ‘number of days’ to give the analysis more meaning (for example, 3 weeks to 21 days), with <1 day converted to 1 day and >6 months to 180 days. The a priori decision was for the main outcome to be a single measure of intention combining all four items, but with the four items also presented separately for clinical relevance.

Additional measures of process were:

• self-efficacy for consulting without delay; 10 items each with a 10-point scale (from ‘not at all confident’ to ‘totally confident’) summed to give scores ranging from 10 to 100, Cronbach’s alpha was 0.85;

• knowledge of symptoms of lung disease; 21 symptoms with ‘yes’, ‘no’, ‘not sure’ answers, expressed as a percentage of answers correct;

• risk perception; ‘how would you rate your chance of getting lung disease?’, with a five-point scale ranging from ‘very low’ to ‘very high’;

Adverse effects were assessed by:

• the Hospital Anxiety and Depression Scale (HADS), a validated scale measuring anxiety (7 items, range 0–21) and depression (7 items, range 0–21), with minimal confounding by somatic symptoms;18

• the Cancer Worry Scale, a six-item scale (range 6–24) with internal consistency (Cronbach’s alpha) of 0.88 in this study; 19,20 and

• general practice consultation rates with anxiety or depression.

Wider health service effects were measured by total general practice consultations, and chest radiograph and respiratory medicine referrals.

Respiratory fitness for treatment was assessed by spirometry, which was measured for participants attending nurse consultations and categorised using Global initiative for chronic Obstructive Lung Disease (GOLD) criteria for COPD.21

Self-report data were collected by questionnaires at recruitment (pre-intervention) and 1 and 6 months after the nurse consultation. For the control group, who did not have a nurse consultation, matching dates were allocated by pairing each control participant at randomisation to an intervention participant and using the latter’s date of nurse consultation. For intervention-group participants who did not attend the nurse consultation, the date of posting their manual was used. Data on consultations and referrals during the years before and after the nurse consultation were collected by review of general practice case notes. Data on duration of symptoms were collected from both records of consultations in case notes and short questionnaires that participants were asked to complete if they consulted.

Sample size

It was estimated that, without intervention, 35% of participants would consult for respiratory symptoms,22 of whom 20% (7% of the total) would consult within target times.11 It was judged that an increase to 25% would be clinically worthwhile; a sample of 200 (100 in each group) would provide 90% power at 5% two-sided significance to detect this difference.

Randomisation

Invitations to participate and randomisation in equal numbers to intervention and control groups were both conducted at the start of the trial. Responses from consenting participants were stratified by general practice and given a unique identification number as they were received. When blocks of 24 had accumulated, random numbers were generated by the project senior statistician and assigned to the previously ordered identification numbers.

Blinding

Data entry from questionnaires was blind to group allocation. Detailed protocols and rules were used for abstraction of data from case notes, but blinding was not possible because indicators of nurse consultations were present in the case notes.

Statistical methods

Analysis was by intention to treat, with participants analysed according to the trial arm to which they were randomised. For counts of consultations by participants, Poisson regression was used to produce an estimate of the ratio of consultation rates between the two treatment groups. The proportion of consultations within the target time was analysed using a generalised linear model with a binomial distribution and logit link function. For outcome measures from questionnaires, linear mixed effects models were used to estimate the treatment differences at 1 month and 6 months, taking into account the correlation between the repeated outcomes. Main analyses were of all observed data adjusted for baseline count or score, age (in years), sex, and practice. This assumes that missing data were at random, therefore sensitivity analyses were conducted with all participants using multiple methods of missing value imputation, including the median value from completers, median values within a scale, hot-deck of completers (random selected value from completers), and last value carried forward, to check for important differences.23