Transtheoretical Model - Criticisms of Criticisms

Criticisms of Criticisms

• In a systematic review of 23 randomized controlled trials published in 2003, the authors reported that "stage based interventions are no more effective than non-stage based interventions or no intervention in changing smoking behaviour."

• A 2005 systematic review of 37 randomized controlled trials claimed that "there was limited evidence for the effectiveness of stage-based interventions as a basis for behavior change."

• A randomized controlled trial published in 2009 found "no evidence" that a smoking cessation intervention based on the Transtheoretical Model was more effective than a control intervention that was not tailored for stage of change.

• The designs of many studies supporting the model have been cross-sectional, but longitudinal study data would allow for stronger causal inferences.

Response:

• A number of longitudinal randomized controlled trials demonstrate that tailored TTM-based interventions do change behaviors. In fact, the Pro-Change LifeStyle Suite had sufficient longitudinal evidence to be awarded the URAC 2009 Gold Award for Best Practices in Health Management.

• Many studies that show the model to be ineffective have tailored interventions only to stage of change; if the studies had tailored interventions based on all core constructs of the model, they might have shown positive findings. In particular, the "processes of change" have been characterized as "under-researched." A 2007 meta-analysis of tailored print health behavior change interventions found that the "number and type of theoretical concepts tailored on," including stage of change and processes of change, were associated with behavior change (Noar et al., 2007). Hutchison et al. (2008) published a systematic review of 34 articles examining 24 physical activity interventions based on the Transtheoretical Model; only 7 of the 24 interventions addressed all four dimensions "stages of change," "processes of change," "decisional balance," and "self-efficacy."

• Studies that find the model ineffective are poorly designed; for example, they have small sample sizes, poor recruitment rates, or high loss to follow-up.

• Velicer et al. (2007) examined predictors of smoking cessation at 12 and 24 months among nearly 3000 smokers from 5 randomized effectiveness trials. They reported that stage was of the strongest predictors of smoking status at 12 and 24 months, refuting the claim that stage of change is descriptive rather than predictive.

Response: The conversion of continuous data into discrete categories is necessary for the model, similar to how decisions are made about the treatment of high cholesterol levels depending on the discrete category the cholesterol level is placed into.

Response: Velicer at al. (1999) conducted a study to examine the validity of 40 predictions based on the Transtheoretical Model regarding movement from one of three initial stages (Precontemplation, Contemplation, or Preparation) to stage membership 12 months later. Thirty-six predictions were confirmed in these longitudinal analyses.

Response: The TTM does not suggest that movement through the stages is always linear. Latent transition analyses on data from effectiveness trials of tailored interventions (e.g., Martin, Velicer, & Fava, 1996) reveal that movement through the stages is not always linear, that the probability of forward stage movement is greater than the probability of backward stage movement, and that the probability of adjacent stage movement is greater than the probability of two-stage progression.

Response:

• The interventions included in the review are treated as comparable even though they differ dramatically on which TTM variables are used for tailoring, length of follow-up, sample size, percentage of eligible smokers recruited, and intervention modalities used. Based on our analysis, approximately 60% of the studies in Spencer et al. (2002) and 70% in Riemsma et al. (2003) used only the stage variable from the TTM. Tailoring only on stage is the most common application of the TTM. Five studies in Spencer et al., and three in Riemsma et al., tested interventions tailored on a partial set of TTM variables, namely stage, decisional balance, and/or self-efficacy. Five studies in Spencer et al. and four in Riemsma et al. tested interventions tailored on the full set of TTM variables, including processes of change.

• To assume that tailoring simply on stage would be TTM-based is analogous to assuming that tailoring simply on self-efficacy is based on social cognitive theory. In both situations, important theoretical constructs are not being used, and an important percentage of variance is not being accounted for or controlled. From a practical perspective, it could mean that the only tailored information specific to an individual is based on a single variable. All other information must be general information that has to be assumed to be valid for all people in a particular stage. However, theory and data both contradict this assumption, as individuals in a particular stage, such as Precontemplation, are theoretically expected and have been empirically demonstrated to differ on key TTM variables like the pros and cons of changing and experiential processes of change.

• If effective tailoring requires feedback that is accurate for individuals, then tailoring on stage alone should be less effective than tailoring on a larger set of TTM variables. Of 13 studies in Spencer et al. (2002) and 16 in Riemsma et al. (2003) that used the single variable of stage, only 10 had positive results (about 35%). Of the eight that applied partial TTM tailoring, four (50%) had significant effects. Finally, of the seven studies that applied full tailoring, five (about 70%) had significant effects. The two fully tailored studies that were negative involved teenagers. The number of fully tailored TTM studies was relatively small, but the number of smokers studied was large (>10,000).

• The impact of fully tailored TTM interventions for smoking has been repeatedly demonstrated in randomized, population-based studies with diverse populations since Riesma et al.(2003). These studies tended to produce the same magnitude of effects at long-term follow-up (22% to 26% point-prevalence abstinence), as was found in our first sample of convenience, a representative sample of 5130 smokers, and an HMO population of 4653 smokers. Similar abstinence rates (23.9%) have been found when treating a population of adolescents in primary care. Hall et al., 2006 reported comparable results in a population of smokers being treated for depression. With pregnant smokers in the UK, adding a TTM-tailored intervention to the traditional treatment of midwife counseling produced more than eight times the impact compared to the traditional treatment alone.

• Recent research demonstrated the same range of abstinence when treating populations with TTM fully tailored interventions for multiple behaviors. This was the case with a population of 2460 parents of teenagers who were treated for three behaviors. The significant abstinence rate was 22.9% with an even higher success for those progressing from high-fat to low-fat diets (38.2%) and for those progressing from high-risk to low-risk ultraviolet (UV) exposure (35.2%). Similar results were produced with a population of 5545 primary care patients (Prochaska et al. 2005). Long-term significant abstinence was 25.6% with even greater success for diet and sun exposure. Such studies are causing us to change traditional impact equations from (impact = participation rate x efficacy) to (impact = participation rate x efficacy x number of behaviors changed).

Response: The authors claimed to be studying the impact of TTM-based interventions on weight loss and reported that the selection criteria included randomized controlled trials using the "TTM SOC as a model, theoretical framework, or guideline in designing lifestyle modification strategies (mainly dietary and physical activity versus a comparison intervention of usual care). One of the outcome measures of the study was weight loss, and participants were overweight and obese adults." These criteria, however, were not systematically applied. Most glaringly, two of the five trials (Dinger et al., 2007 & Steptoe et al., 2001) did not include weight loss as an outcome. Furthermore, those two studies included participants who were not overweight or obese. Jones et al., (2003) included no physical activity intervention and measured weight only as a secondary outcome. That leaves two studies that potentially met the inclusion criteria. A careful reading of Logue et al. (2005), however, indicates that behavior change targets were not clearly specified in that intervention, which the authors defined as a "minimal intervention for obesity." Rather than using public health criteria for reaching Action for diet and physical activity, Logue et al. (2005) reported focusing on small, non-specific increases in exercise and eating.

Though the stated outcome of the review was to assess the potential for TTM-based interventions to measure sustained weight loss, sustainability of weight loss was not adequately assessed. Of the three studies that measured weight loss, two of the three (Jones et al., 2003 and Logue et al., 2005) measured weight loss only at the end of treatment. No follow-up beyond the end of treatment was included. Only one of five studies measured weight loss at one year post-intervention (Johnson et al., 2008). When examined carefully, the results of this study demonstrate that in the context of a truly effective, evidence-based TTM individualized intervention, weight loss in the treatment and control groups begins to diverge at 24 months (a full 12 months after treatment ended). In fact, Johnson et al. (2008) reported that among participants in the pre-action stages (i.e., those at risk for diet and/or physical activity), there was a significant and increasing difference over time in the proportion of participants losing at least 5% of their body weight. At the 24-month follow-up, 30% of those in a pre-action stage for both healthy eating and exercise at baseline had lost at least 5% of their body weight in the treatment group versus only 18.6% of the comparison group.

The bar for being defined as TTM-based intervention study was set far too low. The authors note that listing stage names "fulfills criteria for using TTM SOC." The only thing common to the included studies, however, is that SOC names appeared in the abstracts. As the authors acknowledge, the TTM was inconsistently applied in everything from one size fits all email reminders (improperly using primarily behavioral processes of change for a sample almost entirely in Contemplation at pre-test) in an under-powered 6-week long study with no follow-up in which weight wasn’t even measured (Dinger et al., 2007) to stage-matched messages in 2-3 interactions from a nurse with only brief training (Steptoe et al., 2001), to weight loss advisors who adhered to the intervention protocol less than 50% of the time (Logue et al., 2005). Investigators with adequate knowledge of the TTM recognize that it is a comprehensive model of behavior change in which stage of change is one of 14 variables that make up the model. To date, the best practices for TTM-based interventions employ statistical decision making to derive evidence-based decision rules about how to best match messages to a participant’s readiness to change and status on multiple behavior change variables. Conclusions regarding the efficacy and effectiveness of TTM-based interventions should be based on high quality research that applies the model appropriately, just as conclusions about the efficacy of medications are based on well controlled trials of pharmacologic agents manufactured under the strictest quality controlled procedures. Unfortunately, those standards were not applied here.

• The review gave no consideration to the quality of studies included beyond the reporting of potential biases that were often, as the authors admitted, inappropriate for consideration for the trials included. No mention, for example, was made about whether the studies reviewed had adequate statistical power.