The LION study protocol has already been published : it is a pragmatic single-blind multi-site cluster randomized controlled trial. The Medical Ethical Committee of the University Medical Center Groningen approved the study. Eligible patients received an information letter and signed informed consent before participating in the trial. The study was performed in accordance with the Declaration of Helsinki and registered in the Dutch Trial Registry (NTR3765, www.trialregister.nl, 21 December 2012). The trial adheres to the CONSORT guidelines  and the CONSORT-EHEALTH checklist (V.1.6.1) .
Participants, recruitment and randomization
SMI patients treated by 21 Flexible Assertive Community Treatment (F-ACT) teams  and eight sheltered facility teams of five mental health organizations in the northern Netherlands (catchment area 3.6 million inhabitants) were invited to participate in the study within twelve months after the inclusion of the teams (January 2014 to October 2015). F-ACT teams offer outreach mental health care to community-dwelling patients in their own living environment, ranging from low intensive support to high intensive treatment . When independent living is not possible, patients may reside in sheltered facilities in the community where they are supported in budgeting and other independent living skills. The F-ACT teams were clustered based on organization, caseload size, patient mean age, mean duration of patient admission, most prevalent diagnosis and location (urban or rural), and were randomized equally between intervention or control group. Randomization within eight blocks of two to three teams was performed using a random number generator by a member of the research team (FJ) not involved in the training of staff or the recruitment of patients. To avoid spill-over effects of the intervention, sheltered housing teams which relied on F-ACT teams for their patients’ mental health care, were assigned to the same condition as the F-ACT teams. In some teams, all of the nurses participated, while in others the team leader selected nurses.
MH nurses invited SMI patients to participate if their annual physical screening revealed at least one of the following metabolic risk factors: WC > 88/102 cm (females/males); fasting glucose > 5.6 mmol/L or HbA1c > 5.7% or > 39 mmol/mol; BMI > 25 kg/m2. Exclusion criteria were pregnancy, BMI < 19 kg/m2, or impairment in performing physical activity. Patients received an information letter and provided written informed consent. In total, with an alpha = 0.05 and power of 0.80, 275 patients were required to detect a clinically relevant reduction of 5.8 cm in the primary outcome for WC , taking into account 10% dropout.
The twelve-month multimodal, patient-centred lifestyle intervention was delivered by MH nurses, and included the use of the web tool ‘Traffic Light Method for Somatic Screening and Lifestyle’ (see below). MH nurses coached patients by using the web tool during regular care visits, ideally once every two weeks. Key features of the intervention were:  patient readiness for behaviour change was not a prerequisite for starting the intervention,  patients decided what lifestyle behaviour would be targeted and created their own lifestyle plan,  the intervention addressed diet and physical exercise and incorporated behavioural techniques,  active support from friends and family was incorporated into the lifestyle plan, and  nurses focused on coaching the patient and creating a healthier environment in the mental health care institution, if applicable. The patients’ level of readiness to change their diet and physical activity behaviours was assessed in the web tool, so MH nurses could better tailor the content of the intervention to the stages of change.
Nurses received one day of training on: (a) basic components of motivational interviewing  and the stage-of-change model , (b) side effects of psychotropic medication, (c) lifestyles of and risks for SMI patients, (d) working with the web tool Traffic Light Method, and (e) environmental factors that may influence effectively working with the Traffic Light Method (e.g. the availability of unhealthy products in the home environment). In addition, the study protocol was explained. After three months, nurses attended an evaluation session to discuss obstacles in the use of the tool, obstacles in motivating patients to participate and to recollect the study protocol. Due to the nature of the intervention, the trained LION nurses were not blind for study allocation.
Patients in the control condition received care-as-usual, which entails an annual Routine Outcome Monitoring (ROM) assessment, of which results are discussed with patients. They are referred to their GP when ROM results indicate this is necessary. Lifestyle guidance is more or less provided when patients express an interest.
The web tool ‘Traffic Light Method for Somatic Screening and Lifestyle’
The Traffic Light Method was developed as a practical tool for use by nurses and patients in a Dutch mental health care organization and further advanced by a small spin-off company (Charly Green, Bilthoven, the Netherlands). The Traffic Light Method was constructed based on national and international guidelines for a healthy diet and physical activity and on the literature on somatic screening and lifestyle coaching, and was adapted after two rounds of Delphi panel expert discussion . The tool provides knowledge and incorporates behavioural techniques to elicit behavioural change, such as creating awareness, goal setting, providing feedback and self-management. The tool addresses the lifestyle themes of diet, physical activity, medication use, personal hygiene, stressors, substance use and sexual behaviour.
The tool consists of a screening phase and a follow-up phase. In the screening phase, patients and nurses appraise the patient’s lifestyle behaviours. The Traffic Light Method generates a visual risk profile for each lifestyle behaviour based on the level of risk, represented by green (healthy), orange (medium-healthy) or red (unhealthy) traffic lights. Based on the lifestyle screening, patients construct a lifestyle plan with SMART (specific, measurable, attainable, realistic and timely ) behavioural goals. During the follow-up phase, nurses and patients evaluate the patient’s progress in achieving the lifestyle goals using follow-up reports during fortnightly regular care visits for approximately 15 min. After six months, patients and nurses screen lifestyle behaviours again, revisit and adjust or create a new lifestyle plan and evaluate this plan for the next six months until the intervention ends. In our study, it was estimated that patients would complete 23 follow-up reports over 12 months, i.e. 26 fortnightly visits minus the visits for the three lifestyle screening sessions. For a detailed description of the Traffic Light Method, see [21, 22].
Measurements and outcomes
The primary outcome concerned waist circumference (WC; cm) after six and twelve months intervention. Secondary outcomes were measured by Body Mass Index (BMI; kg/m2) and metabolic syndrome Z-score (MS Z-score; SD), a standardized score for the cluster of five cardiometabolic risk factors. Information on age, sex, diagnosis and medication use was derived from patient record forms. As part of standard care, trained ROM nurses screen patients annually on physical and psychosocial outcomes according to protocol . The data were used for baseline and twelve-month measurements. For the additional physical exam and lab test after six months of intervention (six-month measurement), participants received a nominal remuneration (EUR 5/USD 5.45). ROM nurses were blinded to treatment allocation.
WC, weight, height, systolic and diastolic blood pressure (BP) were measured according to protocol . Fasting blood samples were collected in a hospital or other laboratory for levels of lipids (total cholesterol, LDL-cholesterol, HDL-cholesterol and triglycerides) and glucose metabolism (glucose, HbA1c). If not fasting, this was routinely indicated on the form.
Metabolic syndrome (MS) was defined as the presence of three or more of the following criteria : WC ≥ 88/102 cm (female/male); systolic BP ≥ 130 and/or diastolic BP ≥ 85 mmHg or receiving antihypertensive medication; HDL-C < 1.03/1.3 mmol/L (female/male) or receiving lipid-lowering medication; fasting triglycerides ≥1.7 mmol/L or receiving lipid-lowering medication; and fasting glucose ≥6.1 mmol/L  or receiving antihyperglycemic medication. When fasting glucose levels were not available, patients were considered to fulfil the glucose risk criterion if they reported to have diabetes or if HbA1c ≥ 42.0 mmol/mol . Since the dichotomization of the MS components reduces sensitivity to changes over time, the individual components were standardized into Z-scores (with HDL-cholesterol Z-score multiplied by − 1) [32, 33] and the sum, divided by five, was used as a continuous variable for the degree of metabolic syndrome (MS Z-score). BP was standardized using mean arterial pressure (MAP).
Antipsychotic medication (AP) was categorized into three groups according to the strength of its side effect on cardiometabolic health (none, mild or strong) based on the literature [34, 35] (see e Additional file 1 Table S1).
Patient readiness to change physical activity or dietary behaviour was assessed by a question representing the five phases of the stage-of-change model . Answers ranged from ‘not willing to change within six months’ (precontemplation), ‘willing to change within six months’ (contemplation), ‘willing to change within one month’ (preparation), ‘consider myself acting healthily for less than six months’ (action) to ‘consider myself acting healthily for more than six months’ (maintenance phase).
The number of follow-up reports in the Traffic Light Method web tool represents the level of adherence to the intervention.
Results were presented as means (95% confidence interval) or medians (25-75th percentile). Data were analysed using SPSS version 22 , with a p-value of 0.05 considered statistically significant. The intervention effect was tested using an intention-to-treat approach with a multi-level, subject-specific linear mixed model that had an unstructured variance structure, controlling for teams to adjust for clustering of patients within teams and adjusting for type of AP medication. Intervention effects were tested in stratified analyses of pre-specified subgroups based on sex, age and type of housing. In explorative per-protocol analyses, adhering participants (high users) were compared to the control group using the same linear mixed models as described above. Participants who completed at least one lifestyle behaviour screening, constructed lifestyle goals and completed ten or more follow-up reports were considered high users. We also compared the percentage of participants who improved or deteriorated by ≥5% in WC or BMI after six and twelve months between intervention and control groups using Chi-square tests.
We also used a Chi-square test to determine the intervention effects on patient readiness to change dietary or physical activity behaviour by comparing the percentage of intervention participants who had shifted towards more readiness to change between baseline, six and twelve months to the percentage of participants in the control group who had done so.