This study forms part of a larger project, a prospective longitudinal non-comparative study that followed up people with FES who were treated with the lowest effective dose of flupenthixol decanoate medication over 24 months.²⁵ This study is a secondary analysis of data collected during the parent study.

The parent study was conducted in Cape Town, South Africa, and FES patients were recruited from inpatient and outpatient services at a psychiatric hospital and surrounding psychiatric clinics in the Western Cape between 2007 and 2011.

As part of the parent study, inclusion criteria for patients were male or female subjects of 16–45 years age, with first episode of psychosis meeting Diagnostic and Statistical Manual of Mental Diseases, Fourth Edition (DSM-IV) diagnostic criteria for schizophrenia,²⁶ schizophreniform disorder or schizoaffective disorder and a negative history of current SUD (as confirmed by a urine drug screen and history taking) or a serious medical condition. Exclusion criteria included a lifetime exposure to more than 4 weeks of antipsychotic medication or prior treatment with a long acting injectable (LAI) antipsychotic or intellectual disability (intelligence quotient [IQ] < 70). A total of 126 participants were entered into the parent study.

For the purposes of this study, we focused on dropout of participants. ‘Dropout’ refers to withdrawal from the study that was participant initiated (e.g. did not return for follow-up appointments, self-initiated choice of alternate treatment options). Any withdrawal from the study that was clinician initiated (e.g. non-response, persistent side effects) or based on a collaborative clinician and participant decision (e.g. collaborative choice of alternate treatment) or mortality (one participant died of cancer) was not regarded as a dropout. A total of 16 of the initial 126 participants were excluded from this study because they withdrew from the study in consultation with and on advice of the treating clinician (non-response or side effects) and a further four participants were excluded because one died and three moved to an area with a planned alternate follow-up. Therefore, our final sample composed of 106 participants.

Sociodemographic, anthropometric and clinical data were collected at baseline, obtained from the participant and verified in a separate consultation by the caregiver. The data included age, gender, ethnicity, diagnoses, duration of untreated psychosis and history of SUD. Patient diagnosis of schizophrenia, schizophreniform disorder or schizoaffective disorder and SUD was based on the assessment on intake with the Structured Clinical Interview for DSM-IV (SCID) – Patient Edition.²⁷ Although there are some changes to the various levels for SUDs between DSM-IV and DSM 5, the basic definition of SUD remains unchanged.²⁸ Symptom severity of schizophrenia was assessed using the Positive and Negative Syndrome Scale (PANSS)²⁹ and we used factor analysis-derived symptom domains for positive, negative, depression or anxiety, excitement or hostility and disorganised symptoms.³⁰ Depressive symptoms were assessed with the Calgary Depression Rating Scale for Schizophrenia (CDSS).³¹ We administered the Clinical Global Impressions (CGI) Scale as an overall clinician-determined summary measure.³² In addition, for overall level of functioning we used Social and Occupational Functioning Assessment Scale (SOFAS)³³ and the World Health Organization Quality of Life-BREF Scale (WHOQOL-BREF)³⁴ for quality of life. Duration of untreated psychosis (DUP) was estimated from the onset of continuous positive psychotic symptoms to initiation of adequate treatment. Adequate treatment was defined as the start of structured treatment with antipsychotic medication. Subjects underwent physical examination. Tolerability measures comprised adverse event (AE) reporting, Extrapyramidal Symptom Rating Scale (ESRS),³⁵ weight, height and waist circumference.

Investigators were physicians who were trained in the use of the key assessment instruments and inter-rater reliability testing was conducted periodically (intraclass correlation 0.7 or greater).

All patients were treated with a standard regimen of flupenthixol decanoate intramuscular injection (IMI) medication. There was a washout phase of up to 7 days during which all psychotropic medications were discontinued. Patients were treated with oral flupenthixol 0.5 mg – 4 mg per day for 1 week prior to the first flupenthixol decanoate dose to test for hypersensitivity. The starting dose of flupenthixol decanoate was 10 mg two weekly IMI, with six weekly increments of 10 mg two weekly IMI permitted, to a maximum of 30 mg two weekly IMI. The starting dose could be reduced to 5 mg two weekly IMI in patients younger than 18 years. Other permitted concomitant medications included orphenadrine, trihexyphenidyl or biperiden for parkinsonism or dystonia; propranolol for akathisia; and antidepressants and medication for medical conditions at the investigators’ discretion. Other antipsychotics, mood stabilisers and psychostimulants were not permitted.

Bivariate associations between categorical exposures and outcome (adherence) were assessed using Pearson’s chi-square tests or Fisher’s exact tests as appropriate. In the case of continuous predictor variables, t-tests were used if they are normally distributed and Mann–Whitney tests were used for non-parametric data. We then used unadjusted and adjusted logistic regression to assess for an association between adherence variables that were identified on bivariate analyses. A backward stepwise model was constructed with entry and exit probabilities set at 0.1 and 0.05, respectively. Odds ratios (ORs) and 95% confidence intervals (CIs) were presented in the final model. All analyses were performed using STATA version 15 and statistically significant differences were established at p < 0.05.

Of 106 patients, 40 (38%) dropped out and 66 (62%) completed 24 months of treatment. Reasons for dropout were as follows: two relocated without planning transition to a new service, seven withdrew consent to participate in the study and the remainder were uncontactable or did not give reasons for dropping out. Males made up the majority in both dropout (n = 49, 74%) and completion groups (n = 30, 75%) (Table 1). Those who dropped out were significantly more likely to have a SUD (p = 0.01). There were no other significant differences in demographic and clinical characteristics between those who dropped out and those who completed treatment. Although comorbid SUD was strongly associated with dropout, SUD was also highly prevalent (41%) amongst those participants who completed 24 months of treatment. Therefore, for further analysis, we compared sociodemographic and clinical factors in those participants with dual diagnosis (FES and SUD) who dropped out of treatment versus those who completed treatment.

In our cohort, patients with schizophrenia who also used illicit substances (dual diagnosis) were at greater risk of dropping out. On univariate analyses, those with a dual diagnosis who dropped out differed from those with a dual diagnosis who completed treatment (Table 2). Non-completers with SUD had a later onset of illness (p = 0.02), were older at presentation (p = 0.04), completed more years of schooling (p = 0.001) and had more positive symptoms at baseline (p = 0.05) than participant with SUD who completed the study. Participant with SUD who dropped out had poorer insight on the Birchwood Insight Scale (BIS) subscale 2 (awareness of illness) compared with those who completed treatment, but the difference was not significant (p = 0.06). Participants with SUD who dropped out where more ill at time of discontinuation compared with those who completed the study as measured by PANSS total score at study exit (p < 0.01).

We entered all factors that were associated with dropout in those with SUD at the significance level of p < 0.1 into the regression analyses. We found that non-completer participants with SUD had completed more years of schooling (OR = 3.87, 95% CI: 1.34–11.11, p = 0.01). No other factor that was significant on univariate analysis remained significant on regression analyses (Table 3).

The strength of our study is that we prospectively followed up a comparatively large group of carefully characterised FES patients using validated instruments over a 24-month period of treatment. In addition, all participants were assessed by qualified psychiatrists at a research unit and did not have to endure long clinic queues. There were several limitations. Our study was part of a large parent study and dropping out may be reflective of the fact that participants did not want to participate in the study rather than they did not want to take treatment. Studies have reported that coercion in people with schizophrenia may be associated with greater adherence to treatment and participating in a research study potentially adds to the concern regarding coercion. We therefore ensured that the informed consent document was in plain language and was revisited regularly. We also explained to stable patients that should they wish to stop medication before the 24-month period, we would assist them to do so as safely as possible and without prejudice. We also advised that we would assist with the referral process should they wish to obtain treatment elsewhere. The study procedure made it difficult to adjust for the effects of income and system limitations on adherence, two factors that are relevant in South Africa. As part of the ethical requirements of the study, we reimbursed participants for transport costs and medication was provided at no cost. These costs have been cited as factors associated with non-adherence.^44,45