The transtheoretical model as applied to dietary behaviour and outcomes

Abstract

The transtheoretical model (TTM) is a behavioural theory that describes behaviour change as occurring in five stages, ranging from precontemplation to maintenance. The purpose of the present paper is to review and synthesise the literature published since 1999 on applications of the TTM to dietary behaviour so that the evidence for the use of assessment tools and interventions based on this model might be evaluated. Six databases were identified and searched using combinations of key words. Sixty-five original, peer-reviewed studies were identified and summarised in one of three tables using the following categories: population (n 21), intervention (n 25) and validation (n 19). Internal validity ratings were given to each intervention, and the body of intervention studies as a whole was rated. The evidence for using stage-based interventions is rated as suggestive in the areas of fruit and vegetable consumption and dietary fat reduction. Valid and reliable staging algorithms are available for fruit and vegetable consumption and dietary fat intake, and are being developed for other dietary behaviours. Few assessment tools have been developed for other TTM constructs. Given the popularity of TTM-based assessments and interventions, more research is warranted to identify valid and reliable assessment tools and effective interventions. While the evidence supports the validity of the TTM to describe populations and to form interventions, evidence of the effectiveness of TTM-based interventions is not conclusive.

Introduced in 1981¹, the transtheoretical model (TTM) has become one of the most popular and enduring theories in the field of health promotion and health education. The central concept within the theory is that behaviour change is most likely to happen when individuals engage in the right activities, or processes of change, at the right time, or stage. According to the theory, individuals are most likely to experience success in changing behaviour when they engage in strategies that are appropriate to their stage of readiness to make the change. It has been applied to diverse health behaviours, including the cessation of addictions, participation in cancer screening, and the adoption of positive lifestyle behaviours².

Overview of the transtheoretical model

Many published papers provide a detailed description of the TTM, therefore only a brief overview of the model is offered here. The TTM is comprised of the following constructs: stages of change, processes of change, decisional balance, self-efficacy, and temptation. Stage of change refers to the degree of readiness an individual exhibits toward adopting healthful dietary behaviours. A generalised version of the five stages has been identified and used to classify behaviour³. The stages are: (1) precontemplation, in which an individual may or may not be aware that a behaviour change is warranted and has no intention of changing within the next 6 months; (2) contemplation, in which an individual is aware that a change is warranted and is intending to change within the next 6 months; (3) preparation, in which an individual is planning to change within the next 4 weeks; (4) action, which begins the day an individual makes the behaviour change and lasts until they have maintained the change for 6 months; (5) maintenance, which begins after an individual has successfully maintained a behaviour change for 6 months. For many behaviours to which the TTM has been applied, staging algorithms specific to the behaviour have been developed. Nutrition and dietary behaviours are no exception, and several versions of the staging algorithm have been identified in a recent literature review⁴.

A second construct of the TTM is the processes of change⁵. The processes of change are ten strategies that individuals use to facilitate forward movement through the stages. The idea is that one must use the appropriate processes for each stage to be most effective in achieving maintenance of the new behaviour. According to the TTM, the five cognitive processes (dramatic relief, consciousness raising, self-re-evaluation, environmental re-evaluation, and social liberation) are most effective for those in precontemplation and contemplation. The five behavioural processes (helping relationships, self-liberation, counter-conditioning, stimulus control, and reinforcement management) are most effective for those in preparation, action or maintenance.

Decisional balance describes the balance of the pros and cons of changing a behaviour⁵. Based on the work of Janis & Mann⁶, decisional balance assumes that an individual will identify both reasons for changing a behaviour and reasons for not changing. Behaviour change is influenced by the relative weight of the perceived pros v. cons of change. For most behaviours, the pros of changing outweigh the cons of changing as an individual moves from contemplation into preparation. For those in precontemplation, the cons of changing usually outweigh the pros.

Developed by Bandura⁷, self-efficacy has been adopted as a TTM construct⁸. Self-efficacy is defined as the confidence one has in performing a behaviour. In dietary behaviour research, it could be assessed by a series of questions designed to determine how confident an individual is that he or she could successfully perform a diet-related behaviour if he or she wanted to.

Temptation is a TTM construct⁹ that has been used in some, but not all, health behaviours to which the TTM has been applied. It is a measure of the degree to which an individual feels tempted to lapse from the new healthy behaviour. It has been applied to exercise and smoking behaviour^{10, 11} and may be applicable to dietary behaviour, although published studies have not attempted to measure this.

Previous literature review of the transtheoretical model as applied to dietary behaviour

In 1999, Horwath⁴ conducted a literature review of the TTM as applied to dietary behaviour that included thirty-four studies. Twenty-five studies tested the validity of the TTM as applied to dietary behaviour, five used the TTM to describe the dietary behaviour of populations, three used TTM constructs to assess a dietary intervention and one assessed a TTM-based dietary intervention (i.e. stage-matched intervention). Horwath⁴ made six significant observations at the conclusion of her review: (1) there were insufficient TTM-based interventions or prospective studies to determine whether these types of interventions are warranted; (2) most studies focused solely on the stage-of-change construct and excluded the equally important construct of processes of change; (3) nutrient-based dietary intake was frequently misclassified by the staging algorithms used; (4) many studies focused on nutrient-based dietary intake rather than more easily measured food-based eating behaviours; (5) there was no agreement as to how to measure stage for dietary behaviour; (6) the constructs decisional balance, processes of change, and self-efficacy appeared to be reliable in describing each stage of dietary behaviour.

Purpose of the present literature review

Since the publication of Horwath's literature review⁴, the body of nutrition-based TTM literature has more than doubled. A review of the more recent studies is warranted to determine if Horwath's original six conclusions still apply and if new evidence exists to support the application of the TTM to dietary behaviour. The present paper is fifth in a series of literature reviews of the TTM as applied to select health behaviours (including substance abuse, exercise, tobacco and cancer screening) using a systematic review process^10–13. The purpose of the present paper is to systematically review the studies of the TTM as applied to dietary behaviour that have been published since the 1999 Horwath review⁴ to answer the following questions:

1. (1) How has the TTM been used to describe the dietary behaviours of populations and their intentions with respect to following dietary recommendations? Have researchers been able to accurately classify participants according to stage for various dietary behaviour changes?

2. (2) What is the evidence for using stage-based interventions with populations to achieve positive nutritional goals and behaviours?

3. (3) Do valid and reliable measures exist to describe dietary behaviours using the TTM constructs other than stage of change?

Identification of studies to be included

Using a system that has been developed and applied to published literature reviews of the TTM as applied to tobacco use¹⁰, cancer screening behaviour¹² and exercise¹¹, the present body of literature on dietary behaviour was organised and evaluated. A database search was conducted using the following databases: CINAHL, Medline, Academic Search Premier, EMBASE, PsychInfo and ERIC. Combinations of the following keywords were used: transtheoretical model, stages of change, nutrition, diet, dietary behaviour, weight management, obesity and diabetes. Only original research studies published in peer-reviewed journals were included. All of the studies identified were in English, although this language restriction was not specified in the search. (We recognise that there is a growing body of non-English studies of the TTM, and that it is likely that some of these address dietary behaviour.) A total of sixty-five original research studies published between 1999 (the year that the previous literature review ended) and August 2006 were identified.

Organisation of the literature

Studies were classified as population, intervention or validation based on their study design and purpose. Population studies (n 21) were either cross-sectional or longitudinal and sought to describe the dietary behaviour of populations using TTM constructs. Intervention studies (n 25) featured a dietary behaviour-change programme that was evaluated using TTM constructs and, in some cases, based on the TTM (i.e. a ‘stage-matched’ intervention). Validation studies (n 19) attempted to test the validity of applying one or more TTM constructs to dietary behaviour. Each category of studies is summarised and presented in a table.

Analysis of the literature

While the literature review is a qualitative process, we sought to make it as objective as possible to provide a useful and accurate response to the research questions guiding the present review. All studies are rated according to study design using five categories based on established criteria¹⁴. Experimental studies were clinical, controlled trials with random assignment of participants to control and intervention conditions. Quasi-experimental studies were clinical, controlled trials without random assignment. Non-experimental studies analysed the effectiveness of interventions without the use of a control group, and included cohort and case–control designs. Multiple time series studies included evaluation and comparison of participants in the absence of an intervention, such as correlation and regression analysis. Descriptive studies used cross-sectional survey designs.

The intervention studies are each assigned an internal validity rating, using previously established criteria¹⁴. Threats to internal validity are defined as a lack of any of the following: initial assembly of comparable groups, maintenance of groups throughout study, small sample size, loss to follow-up analysis or intention-to-treat analysis, evidence of the reliability and validity of measures, a clearly defined intervention, considering all important outcomes, or adjustment for confounding variables. Studies are given a rating of ‘good’ if they contain no significant threats to internal validity, ‘fair’ if they contain one threat and ‘poor’ if they contain two or more threats.

The body of intervention studies, as a whole, is then rated based on how strong the evidence is to support using stage-based dietary interventions, using criteria established by Anderson & O'Donnell¹⁵. Literature is rated as conclusive if it includes many well-designed clinical controlled trials that demonstrate a cause-and-effect relationship between the intervention and the outcomes. Literature rated as acceptable includes some well-designed clinical controlled trials demonstrating a cause-and-effect relationship between the intervention and outcomes, but more studies supporting this relationship are desired. Literature considered indicative includes many non-experimental studies that suggest a cause-and-effect relationship between the intervention and outcomes; however, there are threats to internal validity with many of the studies. Literature rated as suggestive includes some non-experimental studies that suggest a cause-and-effect relationship between the intervention and outcomes, but the studies are not well designed and may suffer from threats to internal validity. Literature identified as weak does not offer valid evidence for a cause-and-effect relationship between the intervention and outcomes, although experts may believe that a relationship is plausible.

Summary of the literature

All of the studies included in the present review are summarised in one of three tables. Table 1 includes the twenty-one population studies^16–36, Table 2 includes the twenty-five intervention studies^37–61 and Table 3 includes the nineteen validation studies^62–80. A summary of each table is presented in this section.

Table 1 Summary of population studies of the transtheoretical model (TTM) as applied to dietary behaviour

Table 2 Summary of intervention studies of the transtheoretical model (TTM) as applied to dietary behaviour

Fair rating, one threat to internal validity; good rating, no significant threats to internal validity; PACE+, Patient-Centered Assessment and Counseling on Exercise plus Nutrition; poor rating, two or more threats to internal validity; PACE, Patient-Centered Assessment and Counseling on Exercise.

Table 3 Summary of validation studies of the transtheoretical model (TTM) as applied to dietary behaviour

MANOVA, multivariate ANOVA.

Population studies

A summary of the twenty-one population studies is presented in Table 1. Populations represented included: a variety of low-income, ethnic populations; populations of patients with diabetes or obesity; a representative sample of US adults; one community-based sample of adults; international populations, including Spanish and Australian samples; older populations; cardiac populations. Compared with the study populations included in the review by Horwath⁴, the present group of studies provides more information on low-income populations, which received less representation in the body of literature before 1999. Importantly, the current group of studies includes a variety of specific populations, whereas the Horwath⁴ review included many studies using general adult samples from communities and worksites.

When considering the aggregate of data from these studies, the TTM, particularly the stage-of-change construct, was consistent in describing either measured dietary intake or food-related habits and perceptions. All twenty-one studies included measures of the stage-of-change construct as applied to particular dietary behaviours, weight-loss behaviours or nutrient intakes. A minority of studies employed other TTM constructs, such as decisional balance (n 5; study numbers P8, P9, P10, P12 and P16), self-efficacy (n 6; study numbers P8, P10, P11, P16, P18 and P20) or processes of change (n 6; study numbers P4, P7, P8, P10, P15 and P20), limiting the ability to draw conclusions about the utility of these constructs.

The greatest number of studies (n 9; study numbers P2, P3, P4, P5, P8, P13, P17, P18 and P21) included assessments of fruit and vegetable intake. Only two studies focused on the intake of dairy or dairy products (study numbers P9 and P10). Four studies focused on broader eating patterns (study numbers P1, P12, P14 and P15), and two included portion size (study numbers P14 and P21). In relation to specific nutrients, ten studies considered dietary fat intake (study numbers P2, P5, P6, P7, P13, P16, P17, P18, P19 and P21). Finally, seven studies focused on other health-related behaviours (i.e. exercise or physical activity, smoking, alcohol use) and perceptions related to weight loss (study numbers P1, P2, P7, P11, P13, P16 and P21). Studies showed a fairly consistent pattern of increases in fruit, vegetable, or Ca intake, as well as decreases in fat intake or in the percentage of energy from fat across the stages of change, in a variety of populations. Researchers also demonstrated hypothesised relationships between later stages, such as preparation, action, and maintenance, and greater focus on health and health-related behaviours.

The prior literature review⁴ noted inconsistency in application of the stages of change, perhaps due to a focus on nutrients rather than foods. In the body of literature reviewed in the present review, the focus was on foods, which might explain the improved consistency in staging algorithms. In fact, of the twenty-one studies reviewed, only two (study numbers P15 and P16) elicited data that either did not support the TTM as a useful descriptive tool or were inconclusive. It is interesting to note that both studies included stages-of-change assessment of nutrient intake (i.e. fat) alone, or along with, stages-of-change assessment for fruit and vegetable intake. Also, stages of change might not, according to one study, be effective in identifying differences between sexes; however, the results of this study were not reflective of results of previous studies in which sex differences have been successfully measured.

Even though the majority of studies successfully utilised the TTM and its constructs in describing various populations, studies employed different techniques for assessment of dietary intake, nutrition-related perceptions, or stages-of-change determination. A total of eighteen different algorithms for a variety of health behaviours were used among the population studies in Table 1. Seven of the reviewed studies included novel staging mechanisms. The staging algorithms used most often included one developed by Laforge et al. ⁸¹ for fruit and vegetable intake and one developed by Greene & Rossi⁸² for low-fat diets. Of those studies that included dietary intake or habits assessments, eight utilised some form of a FFQ (study numbers P2, P3, P9, P10, P15, P17, P19 and P20), two used a 3 d food record (study numbers P4 and P5), two used a 24 h recall (study numbers P1 and P8) and four included assessments of dietary habits or behaviours (study numbers P6, P7, P12 and P18).

Intervention studies

A summary of the twenty-five intervention studies is presented in Table 2. The earliest study was published in 2000 and the most recent in 2006. Of the twenty-five studies, fifteen used experimental designs, four were quasi-experimental and six used non-experimental designs. All but one of the nineteen experimental or quasi-experimental studies had an internal validity rating of good or fair. Populations included low-income adults (n 4), primary-care patients (n 8), African-Americans (n 2), native Hawaiians (n 1), Canadian adults (n 1), British adults (n 1), German adults (n 1), Dutch adults (n 3), Hispanic mothers (largely migrant farm workers) (n 1), denture-wearers (n 1), patients on cholesterol-lowering drugs (n 1), type 2 diabetics (n 4), overweight adults (n 3), middle-school students (n 1), parents (n 2) and undergraduate students (n 2). In assessing TTM constructs, twenty-four studies focused primarily on stage of change and one focused primarily on processes of change.

Studies varied according to the dietary habits they attempted to change and measure. It is important to note that dietary behaviour is not one behaviour, but a collection of many specific behaviours. All but one study focused on one or more specific behaviours, often using multiple staging algorithms, each designed to measure a specific dietary behaviour. Fifteen studies measured reductions in dietary fat, seven studies assessed fruit and vegetable intake, two studies assessed portion control, two studies assessed fibre intake, and one study assessed both Na intake and sugar intake. None had weight loss as a primary study goal or programme focus.

Sixteen studies used staging algorithms that had been previously validated for use with a specific nutrition behaviour. In two studies, authors adapted the tobacco staging algorithm. In three other studies, the staging algorithm used was either unspecified or described in terms too general to determine the specific algorithm used. Finally, one study used a non-validated algorithm developed by the authors of that study.

Of the twenty-five studies, nineteen supported the use of stage-based dietary interventions. These included twelve studies with experimental designs (study numbers I2, I3, I5, I10, I11, I14–16, I21, I22, I24 and I25), three studies with quasi-experimental designs (study numbers I7, I8 and I13) and four programme evaluations using non-experimental designs (study numbers I6, I9, I18 and I23). These studies included populations of primary-care patients, patients at risk for heart disease, patients on cholesterol-lowering drugs, type 2 diabetics, denture-wearers, parents, Canadian adults, English adults, Dutch adults, native Hawaiians, two of the four low-income populations, and students.

Two studies clearly did not support the use of stage-based dietary interventions. One of these was a well-designed, quasi-experimental study of African-American church members, many of whom were low income (study number I17). In this study, participants in precontemplation performed as well as those in preparation in terms of changing and maintaining healthful dietary habits. The second study was a non-experimental programme evaluation of a dietary intervention designed for and delivered to mothers in migrant farm working families (study number I23). The authors identified several significant barriers to reaching this population and concluded that the method of evaluation of stage of change for this population needs to be culturally sensitive and specific to them. In another study (number I1), it was unclear whether the intervention was stage-matched or not, as each participant designed his or her own intervention, and the interventions were not reviewed by the researchers. Among native Hawaiians, the stage-matched intervention was effective in promoting stage progression in earlier stages, but not in enabling those in later stages to maintain their changes (study number I13). While Keller et al. ⁴⁸ (study number I12) did not find support for the use of TTM-based counselling among physicians with their patients, it is important to note that the study suffered from several flaws, including low recruitment and retention rates. The final two studies (study numbers I4 and I18) did not provide follow-up measures of TTM constructs; however, a follow-up study of the Riebe et al. ⁵⁵ (study number I19) sample demonstrated support for the TTM.

Compared with the three intervention studies included in the Horwath⁴ literature review, the twenty-five intervention studies in the present paper add considerable evidence to support the effectiveness of stage-based interventions for dietary behaviour. Of the three interventions included in the Horwath review, only one was stage-matched, meaning that the intervention was tailored to the stage of each participant. In the present study, nineteen were stage-matched, including all of the studies using experimental and quasi-experimental designs.

Validation studies

Of the nineteen validation studies summarised in Table 3, thirteen focused on the development and/or validation of a stage-of-change algorithm. These included one for each of the following behaviours or situations: fruit and vegetable intake (study number V8), dietary fat reduction (study numbers V5 and V16), readiness for recovery from anorexia nervosa (study number V6) and weight-loss behaviours (study numbers V16 and V18). In two studies, an algorithm was created for Ca intake (study numbers V17 and V19). Six studies included the development and testing of other TTM measures with specific populations. In four of these studies, decisional-balance measures for fruit and vegetable intake were created for Chinese adults (study number V9), senior citizens (study number V14), low-income African-American adolescents (study number V4) and middle-class adolescents (study number V15). One study featured the development of a decisional-balance scale to assess fruit and vegetable intake for culturally diverse male college students (study number V1). A second study featured the development of a decisional-balance scale to assess dietary fat reduction and other weight-loss behaviours among Pacific islanders (study number V16). Padula et al. ⁷⁵ (study number V14) also tested the application of self-efficacy to fruit and vegetable intake with a senior citizen population. Di Noia et al. ⁶⁵ (study number V4) developed both a processes-of-change scale and a self-efficacy assessment tool for low-income African-American adolescents to assess fruit and vegetable intake. Rossi et al. ⁷⁶ (study number V15) also tested an instrument to measure situational temptations with the fruit and vegetable intake of adolescents.

Of the nineteen studies, only three used a prospective design and were designated as ‘multiple time series’ in Table 3. Eight studies randomly selected participants and eleven used convenience sampling methods. The populations of these studies varied widely, and included Dutch adults, British adults, Pacific islanders, Finnish diabetic patients, young adults, senior citizens, overweight females, African-American females, low-income African-American adolescents, middle-class adolescents, females with anorexia nervosa, female college students and culturally diverse male college students.

Of the three categories of studies in the present review (validation, population and intervention), the category studied in greatest depth in the Horwath⁴ review was the group of validation studies. In her analysis of twenty-five studies attempting to validate one or more constructs of the TTM as related to dietary behaviour, Horwath⁴ drew several important conclusions. The first was that it is much more complicated to stage individuals for dietary behaviour than for smoking behaviour. Many of the staging algorithms in her review tended to place individuals in action or maintenance, when in reality they were not yet in action for a specific nutrition behaviour. Another staging concern among staging nutrition behaviour is that some behaviours, such as consuming 30 % or less energy from fat, are harder for consumers to measure than others, such as eating five servings of fruit or vegetables per d. Horwath⁴ suggested that future studies needed to offer simpler and more specific goals that consumers can easily assess within a short time frame. In the present review, two studies successfully addressed the issue of inaccurately identifying pre-action participants as being in action or maintenance through the development of new staging algorithms for fruit and vegetable intake (study number V8) and for multiple weight-loss behaviours (study number V18). No studies specifically attempted to identify dietary behaviours that were easy for participants to measure, and the most common behaviours measured were fruit and vegetable consumption and the reduction of dietary fat.

The second important finding of Horwath⁴ was that the processes of change for dietary behaviour were minimally studied and validated, except for a few studies related to dietary fat reduction. In the present review, two studies (numbers V4 and V14) included the development of a tool to assess the use of the processes of change for fruit and vegetable consumption among low-income African-American middle-school students and diverse, male college students.

Horwath⁴ identified a stronger body of studies linking self-efficacy with dietary behaviour, most notably dietary fat reduction and increases in fruit and vegetable consumption. As expected, individuals in pre-action stages had lower self-efficacy scores than did those in action and maintenance. One study in the present review included the application of self-efficacy to senior citizens for fruit and vegetable consumption (study number V14), although the authors did not specify how it applied to the study findings. Di Noia et al. ⁶⁵ (study number V4) created a process-of-change scale for low-income African-American middle-school students and found that those in action or maintenance were more likely than their peers in pre-action stages to report using a greater number of the processes, although they did not report using the behavioural processes more frequently than the cognitive ones. (The TTM postulates that individuals are more likely to use the cognitive processes in pre-action stages and the behavioural processes in action and maintenance.)

Decisional balance also appeared to relate predictably with stage of change for dietary fat reduction, fruit and vegetable intake and increased consumption of milk products⁴, although two of the seven studies cited in Horwath's review were of unpublished results. Four studies in the present review included the development of a decisional balance instrument for a specific population. Ling & Horwath⁷⁰ (study number V9) found that a culturally specific decisional-balance scale to assess fruit and vegetable consumption among Chinese adults followed the same pattern as it does for other populations, with pros increasing in contemplation and cons decreasing in preparation, action and maintenance. A focus group method was used to identify pros and cons of fruit and vegetable consumption among senior citizens (study number V14). Di Noia et al. ⁶⁵ (study number V4) found that decisional balance applied in predicted ways to a low-income African-American middle-school sample, with pros rated higher than the cons by the action stage. Simmons & Mesui⁷⁷ (study number V16) found that Pacific islanders identified pros and cons that were culturally specific, yet also found support for the use of the decisional-balance construct with this population. Finally, a decisional-balance questionnaire was successfully developed to assess fruit and vegetable consumption among ninth-graders (study number V15).

Conclusions

In this section of the review, we answer the three research questions on which the present review is based. We also evaluate the evidence for using stage-matched dietary behaviour interventions, using the criteria presented earlier in the paper. We begin by answering the three research questions in order.

How has the transtheoretical model been used to describe the dietary behaviours and nutrition goals of populations? Have researchers been able to accurately classify participants according to stage for various dietary behaviours and nutritional goals?

In the body of literature reviewed, the TTM has been applied mainly to describe fruit and vegetable intake or dietary fat intake, with very few studies focusing on other food groups or nutrients. Researchers have been able to classify participants accurately, and this success has been most consistently related to use of FFQ. Success in staging based on other food groups was demonstrated, but studies are too few to be able to identify scientific consensus. Inconsistency in staging occurred mainly when the focus was on nutrient intake rather than food intake, similar to what has been noted by Horwath⁴.

What is the evidence for using stage-based interventions with populations to achieve positive nutritional goals and behaviours?

Using the criteria established by Anderson & O'Donnell¹⁵, we rated the evidence as indicative for dietary fat reduction and suggestive for other dietary behaviours. While there are a growing number of studies supporting the use of stage-matched interventions for fruit and vegetable consumption, there are few, if any, studies which provide evidence for stage-matched interventions for other dietary behaviours. While nineteen studies supported the use of stage-matched interventions, two did not. These two studies both utilised non-white, low-income populations, suggesting that the interventions themselves and the algorithms used to assess stage of change may not apply equally to all socio-economic and cultural groups. Of the nineteen interventions using experimental or quasi-experimental designs, all were well designed with one or no threats to internal validity, adding to the evidence for the use of stage-matched interventions for dietary fat reduction and fruit and vegetable consumption.

In a recent review of thirty-seven TTM-based interventions employing a randomised control trial design⁸³, the authors raised several important concerns with the quality of these studies and the inferences which can and should be drawn from them. This review, which included trials of interventions for dietary behaviour, smoking cessation, exercise, screening mammography, mental health treatment, tobacco or alcohol prevention, and multiple lifestyle changes, demonstrated a lack of evidence for TTM-based interventions. One of the critical issues raised was the potential for poorly designed TTM-based interventions due to limited application of the model in assessing participants and creating interventions for them. Bridle et al. ⁸³ found that, in many cases, interventions were vaguely described and staging mechanisms lacked evidence for validity. They attributed this to a limitation of the TTM itself in clearly discerning discrete stages and specifically identifying the processes of change that should be used in each stage. They also pointed out that, in many of these studies, the intensity of the TTM-based intervention was potentially greater than that of the comparison intervention, which could explain its apparent effectiveness in these studies. One potential limitation of drawing inferences from the Bridle et al. ⁸³ review specific to dietary behaviour is that only five diet-related interventions met the review criteria and were included. These included three published articles in peer-reviewed journals, one doctoral dissertation and one conference proceedings abstract.

In the present review, we found some improvement among the intervention studies in two of the three problem areas identified by Bridle et al. ⁸³. In our judgment, ten of the nineteen experimental or quasi-experimental intervention studies provided sufficient detail about the interventions to allow other practitioners to replicate them. Four studies utilised the ‘expert systems’ approach^{46, 47, 50, 51}. Two studies used the Patient-Centered Assessment and Counseling for Exercise plus Nutrition^{40, 52}. Auslander et al. ³⁸ clearly referenced the Eat Well, Live Well programme, Clark et al. ⁴¹ clearly cited the sources of the assessments used in patient counselling and Steptoe et al. ⁵⁸ provided references for the Changes of Heart Program. In each of these studies, the authors specified the name of the intervention and referred to another source in which the intervention was described in detail. In addition, Resnikow et al. ⁵³ both described within the article and referenced the Motivational Interviewing techniques used in this intervention. Of this same group of nineteen experimental or quasi-experimental intervention studies, only four included assessments of TTM measures other than stage of change. Prochaska et al. ^{50, 51} also assessed processes of change and decisional balance in two studies. Finckenor et al. ⁴³ used the processes of change to created stage-matched interventions for college students. Johnson et al. ⁴⁶ also assessed processes of change and decisional balance in their expert system intervention for patients using cholesterol-lowering medications. While self-efficacy was measured in several other interventions, the additional presence of this measure alone was not considered evidence of using multiple TTM constructs, given that many behaviour-change theories incorporate self-efficacy. Finally, we fully concur with Bridle et al. ⁸³ on the lack of intensity of interventions provided to participants in control or comparison groups. Almost all of the intervention studies in this review compared a stage-matched intervention to ‘usual care’ or to no intervention at all. It is no surprise that a TTM-based intervention would yield significantly better results than no intervention at all; this would be true for most interventions based on many of the existing health-behaviour theories. The more important question is whether a TTM-based intervention is more effective than other types of interventions, particularly if other interventions require fewer resources to implement. We recommend that future research compare multiple interventions to determine which are the most effective and efficient.

It would be useful to see future research compare stage-matched to stage-mismatched dietary interventions. These types of studies have been conducted for tobacco^84–88 and physical activity^{89, 90}. It is important to note that these seven interventions varied in methodology and overall quality of design, yet all but one of the studies raise doubts as to the superiority of the stage-matched intervention over the stage-mismatched intervention.

Do valid and reliable measures exist to describe the nutritional status and behaviours of populations using the transtheoretical model constructs? Do they still only exist for stage of change, or have measures been developed for processes of change, situational temptation, decisional balance and self-efficacy?

Progress has clearly been made in developing valid and reliable stage-of-change algorithms for dietary behaviour, although more validation research is needed. Since the Horwath review in 1999⁴, the bulk of published validation literature has focused on the development of stage-of-change algorithms for specific populations. This is an important step forward, since staging algorithms appear to be culturally and demographically specific. Few researchers have sought to refine and improve existing algorithms to correct for the problem of placing pre-action individuals into action and maintenance, although more have developed algorithms for dietary behaviours that are easier for the average consumer to measure. It is encouraging that four published studies since the Horwath review have attempted to develop and validate measures for decisional balance, processes of change, situational temptations, and self-efficacy, but more studies on these constructs are needed.

Based on this body of literature, it was apparent that there is little consensus on the best algorithm to use for each dietary behaviour, although the existence of multiple validated algorithms add to the validity of the stage-of-change construct. Twenty-four different staging algorithms were used in these studies (including the eighteen identified among the population studies) and another four studies employed staging algorithms that were not identified at all (noted as ‘unspecified’ in Tables 1, 2 and 3). We estimate that approximately fifteen of these algorithms have been tested for validity and reliability, although it is difficult to discern this in a few studies. Some studies used algorithms that are specific to one behaviour, such as fruit and vegetable consumption or dietary fat reduction, while others used a form of the algorithm that was initially developed for smoking and simply adapted it for a nutrition behaviour or status. The most frequently used algorithms were for fruit and vegetable consumption⁸¹, dietary fat reduction⁸², and an adapted form of the algorithm created for smoking behaviour³.

The findings of the present review are limited in that we used a subjective method to categorise and evaluate each study individually, as well as evaluate the body of literature as a whole. While we strove to use a systematic approach and clear criteria for the assessment of this literature, it remains a qualitative process that relies upon human judgment. While we tried to include all of the relevant literature, it is also possible that an appropriate study was inadvertently omitted from the present review.

We make the following recommendations for future research in the application of the TTM to dietary behaviour:

1. (1) The use of previously tested and validated forms of a staging algorithm is desirable when possible, as well as the clear identification of the staging algorithm used. Correctly staging individuals is the cornerstone of all TTM applications. It would enhance the validity of the application of the TTM to dietary behaviour if consensus could be reached for the staging algorithm used with each dietary behaviour. This would be particularly useful for fruit and vegetable consumption and dietary fat reduction, as they appear to be the most often-studied behaviours.

2. (2) The application of the entire TTM is important in designing and evaluating interventions, rather than just the stage-of-change construct. For example, it would be very useful to know which processes of change facilitate forward stage movement for each dietary behaviour, and how these vary between populations. Similarly, understanding how decisional balance, situational temptations and self-efficacy affect stage progression would also be useful in designing effective interventions.

3. (3) More experimental studies are needed to determine if TTM-based dietary interventions are more effective than other kinds of interventions. The current evidence suggests that they are effective for dietary fat reduction and may be for fruit and vegetable consumption, but a larger body of evidence is needed to state this conclusively. Experimental studies are completely lacking for TTM-based interventions related to other dietary behaviours, and these are also warranted.

The progress that has been made in the body of research on the application of the TTM to dietary behaviour is encouraging. Several of Horwath's⁴ concerns and recommendations for future research have been addressed; however, more research is needed to determine with confidence if employing the TTM to describe dietary behaviour and plan dietary interventions is the most effective way to promote healthier dietary intake for many populations.