Skip to main content

Table 3 Characteristics of included optimal treatment strategy studies

From: Statistical design and analysis in trials of proportionate interventions: a systematic review

First author

Intervention

Tailoring variable and decision rules (response unless otherwise stated)

Primary outcome

Statistical analysis

Analysis of stages

Weiss [32]

Two-stage intervention. Stage 1: buprenorphine–naloxone induction, 2 weeks of stabilisation, a 2-week taper and 8 weeks of follow-up. Stage 2: 12 weeks of buprenorphine–naloxone stabilisation, a 4-week taper and 8 weeks of follow-up. In each phase, patients were randomised to (1) standard medical management or (2) standard medical management plus individual drug counselling

Stage 1: self-reported opioid use on ≤4 days in a month, absence of two consecutive opioid-positive urine test results, no additional substance use disorder treatment and ≤1 missing urine sample. Stage 2: abstaining from opioids during week 12 and during ≥2 of the previous 3 weeks

Composite measures indicating minimal or no opioid use based on urine test-confirmed self-reports

Compare two treatment conditions using the stage 2 endpoint; generalised estimating equations to account for clustering of patients by site

Yes

Shortreed [33]

Two-stage intervention. Initially randomised to newer atypical antipsychotics or to perphenazine. Patients randomised at stage 1 to perphenazine who discontinue were randomised to a newer atypical antipsychotic. Patients randomised at stage 1 to a newer atypical antipsychotic who discontinue were given the choice of two randomisation arms: (1) with ziprasidone, olanzapine, risperidone or quetiapine, excluding their previous treatment or (2) with clozapine, olanzapine, risperidone or quetiapine, again excluding their previous treatment. Dissatisfied patients could opt to switch treatment again; at this stage treatment was neither randomised nor blinded

Non-response if patient discontinues treatment and then eligible for randomisation to next stage

12-month PANSS score and 12-month QoL score

Marginal structural modelling using a weighted analysis to compare treatment regimes: the always atypical antipsychotic regime or the perphenazine and atypical regime

No

Wang [34]

Three-stage intervention. Stage 1: randomised to one of four combination chemotherapies. Stage 2: (2a) responders receive second course of same chemo, (2b) non-responders randomised to second-line treatment. Stage 3: After (2a): (3a) responders receive second course of same treatment, (3b) if treatment not finished. After (2b): (3c) if overall success, finish treatment, (3b) if not randomised to second treatment, process repeated once more. After (3a): finish treatment

Response defined as: prostate-specific antigen (PSA) decline of at least 40% from baseline, objective regression (of any magnitude) of any measurable disease, improvement in any cancer-related symptom and no new lesions or new cancer-related symptoms. Success defined as PSA decline of at least 80% from baseline, resolution of all cancer-related symptoms, an objective tumour regression of at least 50% from baseline for all measurable lesions and no new lesions or cancer-related symptoms

Long-term survival using log survival time. Efficiency in diminishing disease burden over 32 weeks using three specific scoring functions defined as functions of toxicity and efficacy taking values in the interval [0,1]

Inverse probability weighting methods to estimate the mean of counterfactual outcome for dynamic treatment regimens and sequentially randomised trials

No

Rose [30]

Two-stage intervention. Stage 1: all received nicotine patch treatment 2 weeks before quit date. Responders continue nicotine patch treatment. Non-responders randomised to (1) control (nicotine patch), (2) nicotine patch and bupropion or (3) varenicline alone. Stage 2: for pre-cessation nicotine patch responders, nonlapsers continue nicotine patch and for those who lapsed in the first week after quit date randomised to (1) control (nicotine patch), (2) nicotine patch and bupropion or (3) varenicline alone

Ad lib smoking (expired carbon monoxide levels) decreased by >50% after 1 week

Continued smoking abstinence at end of treatment

Logistic regression compared each rescue treatment against the control

Yes

Kasari [31]

Two-stage intervention. Stage 1: sessions of (a) JASP+EMT or (b) JASP+EMT+SGD. Stage 2: early responders continue stage 1 treatment. Slow responders from (1a) randomised to receive intensified stage 1 treatment or augmented stage 1. Slow responders from (1b) receive intensified stage 1 treatment

After stage 1, if child demonstrated 25% or greater change on at least half of the variables (7 out of 14), then the participant was considered an early responder

Total spontaneous, communicative utterances coded from a standardised Natural Language Sample

Mixed-effects models compared outcome between stage 1 treatments. Secondary aim analysis used a weighted regression to compare mean outcomes between the three embedded adaptive interventions, including an indicator for stage 1 and 2 treatments

Yes

Kilbourne [36]

SMART design for adaptive implementation strategy. Run-in phase: sites offered REP to implement life goals for patients with mood disorders. Sites not initially responding to REP are randomised to receive additional support from an EF or both EF/IF. Additionally, sites randomised to EF and still not responsive will be randomised to continue with EF alone or to receive EF/IF

<50% patients receiving ≥3 evidence-based practice sessions

SF-12 mental-health-related QoL and PHQ-9 scores

Linear mixed-effects models. Compare interventions in non-responding sites beginning with REP plus EF/IF versus interventions beginning with REP plus EF on longitudinal patient-level change in number of life-goal sessions received. Compare whether continuing REP plus EF versus augmenting with REP plus EF/IF leads to changes in outcomes, among sites who are non-responsive to REP plus EF at month 12

Yes

Wu [35]

Two-stage intervention. Stage 1: patients randomised to bupropion, paroxetine or placebo. Stage 2: non-responders assigned second intervention. If receiving bupropion or paroxetine at stage 1, current doses increased. If placebo at stage 1, bupropion or paroxetine

≥50% improvement over initial SUMD and not meeting DSM-IV criteria for hypomania or mania

SUMD

Q-learning to estimate optimal regime

Yes

  1. DSM-IV Diagnostic and Statistical Manual of Mental Disorders; EF external facilitator; EMT Enhanced milieu teaching; IF internal facilitator; JASP Joint Attention Symbolic Play Engagement and Regulation; PANSS Positive and Negative Syndrome Scale; PHQ-9 Patient Health Questionnaire; PSA prostate-specific antigen; QoL quality of life; REP Replicating Effective Programmes; SF-12 12-Item Short Form Health Survey; SGD speech-generating device; SMART Sequential Multiple Assignment Randomised Trial; SUMD Scale to Assess Unawareness of Mental Disorder