3.1. Eligibility criteria

In order to be included in this review, the study must:

1. 1. Have used the full 23-item Intuitive Eating Scale-2 in English [2]. Studies that used non-English translations, other versions of the Intuitive Eating Scale [13,17] or that only used certain subscales were not included.
2. 2. Have been available in full-text
3. 3. Have used the correct scoring procedure for the IES-2 items
4. 4. Have provided information about at least one of the following psychometric properties:
   1. a. Scale and/or subscale means
   2. b. Subscale intercorrelations
   3. c. Relationship to other variables associated with psychological or physical health (correlational and/or longitudinal)
   4. d. Sex differences
   5. e. Construct validity, consisting of:
      1. i. Convergent validity – whether the measure correlates with other variables that theoretically should be related to it
      2. ii. Discriminant validity – whether the measure is unrelated to variables that are not expected to be associated with it
      3. iii. Known-groups validity – whether the measure is sensitive enough to detect known differences between certain groups
   6. f. Reliability (e.g., test-retest, split half, internal consistency (α or Ω))
   7. g. Factor analysis and/or factor structure

Systematic reviews, meta-analyses, literature reviews, unpublished theses and qualitative studies were not eligible to be included in the review, nor were any studies that did not present original empirical results. It is noted that the COSMIN guidelines [26] recommend that studies that only include the measure of interest as an outcome measurement should not be included in systematic reviews of patient-reported outcome measures as they only provide indirect evidence about measurement properties. However, in this instance preliminary searches suggested that there were insufficient studies that specifically focused on evaluation of the IES-2 and consequently it was necessary for the eligibility criteria to be widened to include all studies that had used the IES-2 and reported at least one psychometric property of interest.

Eligible studies published in languages other than English were not removed until full-text screening because Covidence records a reason for the exclusion of each paper at this stage which enabled us to keep a record of papers that were excluded for this reason. However, they were excluded at this stage and data were not extracted as there was not scope within the review’s time-frame or funding to allow for translation of non-English reports. Studies published in English but which used non-English translations of the IES-2 were also excluded from the review because of the potential for psychometric differences between the original and translated scales that were artefacts of the translation processes and quality rather than representing true cross-cultural variation [27,28].

3.2. Information sources

Searches for peer-reviewed journal articles were performed in Scopus, Web of Science, PsycINFO and MEDLINE. In addition, citation searches for the IES-2’s development paper [2] were performed in Scopus and Web of Science.

Due to the limited time-frame and funding of the review, it was decided that grey literature and non-indexed studies would not be included, and that authors would not be contacted about the possibility of unpublished studies or about any missing data in papers included in the review.

3.3. Search strategy

Initially, a citation search for the IES-2’s development paper [2] was performed. Papers were limited to those published from 2013 onward as this is when the IES-2 was published. While there were other scales of intuitive eating prior to this, the review is specifically focused on the IES-2 and therefore the searches were run from its year of publication. Keyword searches were structured as follows:

1. “intuitive eating” OR
2. “intuitive eating scale-2” OR
3. “intuitive eating scale 2” OR
4. “IES-2” OR
5. IES2

The search strategy was developed alongside a research librarian from the University of Leicester and was reviewed by the research team prior to implementation. The full search terms for each of the databases are detailed in Supplementary File 4.

3.4. Data management

Literature search results were uploaded to Covidence [29], an online systematic review program. Zotero reference management software was also used to manage references and to check for duplicates. In order to avoid “double counting”, author names, study dates, sample sizes and other characteristics and variables assessed were compared to ascertain whether multiple publications had arisen from the same study.

3.5. Selection process

Titles and abstracts yielded by the search were screened against the pre-defined eligibility criteria. Full reports were obtained for any titles that seemed to meet the criteria or any where the eligibility was uncertain. The full-text reports were then screened and a decision was made as to whether the inclusion criteria had been met. If studies were excluded, reasons for exclusion were recorded. If there was still uncertainty about study eligibility once the full-text report had been screened, study authors were contacted for clarification. All studies were screened by the first author (RL) and a subset of studies (10%) were reviewed independently by another member of the research team (the third author, SH) at each stage of the screening process. There was no disagreement between reviewers about the exclusion or inclusion of any of the studies, likely due to the tightly-defined eligibility criteria.

3.6. Data collection process

Standardised forms were created on Covidence and used to extract data from included studies. All data were extracted by the first author and any concerns discussed between all authors.

3.7. Data items

The following data items were extracted:

1. 1. Publication details: Title, authors, publication date, journal published in, article metadata (volume, issue, pages, digital object identifier (DOI)), database(s) indexed in
2. 2. Study characteristics: Type of study (e.g., intervention, cross-sectional), sample size, country/ies in which study took place, study date and duration, method of data collection (e.g., online, face-to-face, postal survey)
3. 3. Participant characteristics: Sample type, gender distribution, age (mean, SD, range), ethnicity
4. 4. Measurement properties (for the full scale and individual subscales):
   1. a. Scale and/or subscale means
   2. b. Subscale intercorrelations
   3. c. Relationship to other variables (correlational and/or longitudinal) that pertain to physical or psychological health (these could not be established a priori as the aim of this element of the review was to explore the constructs that intuitive eating is associated with rather than assess its relationship to a limited number of pre-specified variables)
   4. d. Sex differences
   5. e. Construct validity (e.g., convergent, discriminant and/or known-groups validity)
   6. f. Reliability
   7. g. Factor analysis and/or factor structure

3.8. Data synthesis

Due to the exploratory nature of this review, the populations, outcomes and other characteristics of the studies included were deemed too heterogeneous for a meta-analysis to be feasible. Data were instead synthesised narratively as this approach can accommodate difference and allows for the descriptive exploration of heterogeneity [30].

3.9. Risk of bias

The risk of bias of each study was assessed using two domains from the COSMIN Risk of Bias checklist [25,26], specifically those relating to internal consistency and hypothesis testing for construct validity. The other COSMIN domains were not relevant to this review because the checklist is designed for studies that specifically evaluate the measurement properties of an instrument whereas the majority of the studies in this review only used the IES-2 as an outcome measure and did not report on its development. For each domain, studies were given a rating of either 1) Very good; 2) Adequate; 3) Doubtful; or 4) Inadequate quality. The study was then given an overall rating using the same scale, commensurate with the lowest domain rating it had received.

4.1. Search results

After removing duplicates, the search strategy yielded 2,379 articles published between the release of the IES-2 in 2013 and June 2025, when the final updated searches were conducted. Title and abstract screening led to the exclusion of 1,645 studies that were clearly irrelevant based on their titles or abstracts. An additional 644 studies were excluded after full-text review (see Fig 1 for a detailed overview of the study selection process). The most common reason for exclusion was the use of a non-English version of the IES-2 (224 studies). Other reasons included not using the IES-2 at all (96 studies), being an unpublished thesis (87 studies), or employing a different English-language measure of intuitive eating (64 studies), such as the original Intuitive Eating Scale [17] or the first version developed by Tracy Tylka [13]. The original IES-2 development paper was also excluded, as the review focused on studies that used the measure post-development. In total, 90 studies were included in the final review.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA diagram depicting study selection.

https://doi.org/10.1371/journal.pone.0349590.g001

4.2. Quality assessment

Of the 90 studies included in the review, 66 (73.3%) were rated as ‘Very good’ and 24 (26.7%) were rated as ‘Inadequate’. The predominant reason for a study receiving an ‘Inadequate’ rating was for failing to calculate an internal consistency statistic for each subscale of the IES-2 (15 out of the 22 studies rated ‘Inadequate’ received this rating solely for this reason) – see Section 5.3 for further discussion.

4.3. Description of studies included

See Table 1 for a summary of the participant groups, countries of recruitment and largest ethnic groups of the reviewed studies and Table 2 for a summary of each paper included in the review. Mean participant ages ranged from a low of 15.14 years (SD = 1.43 years) [24], in one of only two studies to include adolescents, to a high of 68.8 years (SD = 6.3 years) [23] in a study focused on older adults. Nearly half of the studies had a mean age of 25 and under, likely due to the large proportion of university/college student samples. Twenty-four out of 90 studies (26.7%) recruited only women, two recruited only men and in all but seven of the remaining studies that reported the gender distribution of their participants, women made up more than 50% of the sample.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Summary of study characteristics.

https://doi.org/10.1371/journal.pone.0349590.t001

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Summary of published studies using the Intuitive Eating Scale-2 included in the review.

https://doi.org/10.1371/journal.pone.0349590.t002

4.4. Mean IES-2 scores

Forty out of 90 studies (44.4%) provided mean scores for the IES-2 total and each of the four IES-2 subscales. A further 37 studies (41.1%) provided mean scores for just the IES-2 total and two more studies provided mean scores for just the IES-2 subscales. Highest and lowest scores for the IES-2 total and all subscales are provided in Table 3. Notably, the studies that account for many of the results used clinical samples of people who are either of a higher weight or who have a current or historical eating disorder. In fact, whilst the vast majority of mean scores fell between 3.00 and 3.99, those studies with mean scores below 3 (demonstrating that participants did not generally engage in intuitive eating behaviours) were almost all within people of a higher weight or within eating disorder or disordered eating populations. If the 17 studies that are focused on weight or eating disorders [18,19,24,31–33,35,38,41,44,45,47,66,73,91,69,111] are not taken into account, the sample characteristics are slightly more varied and scores tend more towards the centre of the scale’s 1–5 range (see the second part of Table 3). There were very few scores that fell at 4.00 or above and no commonalities were identified between populations where this occurred.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Highest and lowest mean scores for IES-2 total and subscales.

https://doi.org/10.1371/journal.pone.0349590.t003

4.5. IES-2 inter-correlations

Eleven out of 90 studies (12.2%) provided correlation estimates between total IES-2 scores and the four subscales. A further six studies provided correlation estimates between just the four subscales. Statistically significant inter-correlations ranged between −0.49 and 0.92; see Table 4 for the highest and lowest statistically significant correlations between IES-2 total and subscale scores. Correlations that were not statistically significant are not presented.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Highest and lowest correlations between IES-2 total and subscale scores.

https://doi.org/10.1371/journal.pone.0349590.t004

4.6. Construct validity

The review found a large amount of evidence for the convergent construct validity of the IES-2, in terms of it being associated with measures of constructs that would be expected to be related to intuitive eating e.g., [113–116]. Twenty-one out of 90 studies found the IES-2 to be negatively associated with BMI, although, interestingly, Tylka et al. [14] also found a small-to-medium positive association (r = 0.28, p < .001) between the IES-2 and BMI in their subsample of male participants. Several studies found the IES-2 to be positively associated with body appreciation, self-compassion and self-esteem and negatively associated with depression and anxiety, disordered eating behaviour/eating disorder symptomology and rigid dietary control. Intuitive eating as measured by the IES-2 appeared in general to not be statistically significantly related to age [14,51,88,109,40,75], other than in one study of older adults [23] which found a moderate association (r = 0.32, p < .01).

A full list of all constructs that were statistically significantly associated with the IES-2 can be found in the S2 Table. None of the studies included in the review provided evidence for the discriminant construct validity of the scale.

4.7. Sex differences

Thirteen out of 90 studies tested for sex differences in levels of intuitive eating. Of these, six found statistically significant differences between total IES-2 scores for men and women, with men consistently scoring more highly [51,84,88,101,55,61]. Effect sizes ranged from d = 0.28 [84] to d = 0.42 [88] across the five studies that reported them. One study found statistically significant sex differences in ‘Unconditional permission to eat’ subscale scores [101], with men again scoring more highly and a reported effect size of d = 0.30. Seven studies found statistically significant sex differences in ‘Eating for physical rather than emotional reasons’ subscale scores [54,81,84,86,88,101,61], with men consistently scoring more highly. Effect sizes ranged from d = 0.29 [84] to d = 0.80 [101] across the four studies that reported them (although the latter effect size should be interpreted with caution as this is far higher than would be expected). Two studies found statistically significant sex differences in ‘Reliance on hunger and satiety cues’ subscale scores [84,88], with men again scoring more highly and reported effect sizes between d = 0.29 [84] and d = 0.31 [88]. Murray et al. [88] also found statistically significant sex differences in ‘Body-food choice congruence’ subscale scores, with men again scoring higher than women and a reported effect size of d = 0.22.

4.8. Reliability

The COSMIN methodology for systematic reviews of patient-reported outcome measures [117] recommend that internal consistency (Cronbach’s alpha (α) or McDonald’s omega (Ω)) should be at least 0.7. Out of 90 studies, 67 reported an internal consistency statistic for the IES-2 overall, with 66 reporting an internal consistency statistic of 0.7 or higher. Twenty-nine studies provided internal consistency statistics for the ‘Unconditional permission to eat’ subscale, with 21 reporting alphas of 0.7 or higher. Twenty-eight studies reported on the internal consistency of the ‘Eating for physical rather than emotional reasons’ subscale, all of which were 0.7 or higher. Twenty-eight studies reported on the internal consistency of the ‘Reliance on hunger and satiety cues’ subscale, with all but one of these reporting alphas of 0.7 or higher. Finally, 30 studies reported on the internal consistency of the ‘Body-food choice congruence’ subscale, with all but one of these reporting alphas of 0.7 or higher.

None of the studies included in the review provided evidence for any other types of reliability (e.g., test-retest or split half).

4.9. Dimensionality

Ten studies reported on the factor structure of the IES-2. Five of these [19,33,36,90,104] tested the factor structure found in the original development and validation paper for the IES-2 [2] of four latent factors loaded onto a higher order intuitive eating factor. Babbott et al. [33] found that, in their sample of people seeking eating disorder treatment, this model had negative residual variance, suggesting that the model was misspecified. Swami et al. [104] found this model to be a reasonable-to-good fit to the data in their general population sample (CFI = .96, TLI = .93, RMSEA = .08, SRMR = .03) but only when the variances between the negatively worded items were allowed to correlate. Martin-Wagar and Heppner [19], Parsons et al. [90] and Bazo Perez et al. [36] found this model to be a poor fit to the data in their samples of eating disorder patients (CFI = 0.89, TLI = 0.87, RMSEA = 0.10, SRMR = 0.10), married male sexual minority couples (CFI = 0.69, TLI = 0.67, RMSEA = 0.10, SRMR = 0.13) and Hispanic-majority young adults (CFI = 0.79, TLI = 0.76, RMSEA = 0.13, SRMR = 0.13) respectively.

Six studies [33,37,54,79,61,71] tested a factor structure of four latent factors with the 23 IES-2 items as indicators but no hierarchical intuitive eating factor. In Belon et al.’s [37] sample of college women, this model showed an acceptable fit to the data, with the exception of the RMSEA (CFI = .92; TLI = .91; RMSEA = .10; SRMR = .08). In Jackson et al.’s [61] general population sample this model fitted well (CFI = .98, TLI = .98, SRMR = .05, RMSEA = .04) but only with modifications (two items were removed as their factor loadings were less than 0.5 and errors with modification indices greater than 10 that were theoretically similar were allowed to correlate). In Babbott, Mitchison, et al.’s [33] sample of people seeking eating disorder treatment, this model did not fit unless co-variances were added to similarly phrased items (22 & 23, 13 & 14, 7 & 8), after which it showed acceptable fit to the data (CFI = .93, SRMR = .10, RMSEA = .07). Both Khalsa et al. [71] in their sample of racial minority and/or low-income parents and Saunders et al. [79] in their sample of Hispanic American college students found this model to be an inadequate fit to the data. Ge et al. [54] stated that this model showed acceptable fit to the data in their general population sample (CFI = 0.90, TLI = 0.88, RMSEA = 0.08, SRMR = 0.09) although arguably the TLI and SRMR estimates suggest less-than-adequate model fit. Finally, Swami et al. [104] also tested a bifactor model which contained four factors with individual items as indicators as well as an overall intuitive eating factor with individual items as indicators. This gave the best fit to the data in their general population sample (CFI = .97, TLI = .94, RMSEA = .08, SRMR = .02), again with correlation allowed between the variances of the negatively worded items.

Four studies performed subsequent exploratory factor analyses after testing the original factor structure and finding it to be a poor fit to their sample [19,79,90,71]. In their sample of eating disorder patients, Martin-Wagar and Heppner [19] found a four-factor, 22-item model fit their data best. This utilised the original four factors found by Tylka and Kroon Van Diest [2] but with the items distributed differently across the ‘Unconditional permission to eat’ and ‘Eating for physical rather than emotional reasons’ factors and with item 11 removed due to low factor loading. They also found that those scoring highly on most of the items in the ‘Eating for physical rather than emotional reasons’ factor were low-scoring on items 4 and 5 (“If I am craving a certain food, I allow myself to have it” and “I allow myself to eat what food I desire at the moment”), suggesting that therefore these items should be reverse-scored in eating disorder samples. Parsons et al. [90] found that, in their sample of married male sexual minority couples, a five-factor, 23-item model fit their data the best, utilising the original four IES-2 factors with an additional factor of ‘Eating not as coping’. In their sample of racial minority and/or low income parents, Khalsa et al. [71] found that a six-factor, 23-item model fit their data best, although a subsequent confirmatory factor analysis on a novel sample was not performed. This model utilised all of the items in the IES-2 but arranged them into the factors of (1) Avoiding forbidden foods; (2) Permission to eat; (3) Avoiding emotional eating; (4) Avoiding food-related coping strategies; (5) Reliance on hunger and satiety cues; and (6) Body-food choice congruence. Finally, Saunders et al. [79] found that in their sample of Hispanic American college students, a three-factor, 11-item model fit their data best. This was achieved by splitting their dataset and performing an exploratory factor analysis on one subsample, and then performing a subsequent confirmatory factor analysis on the other subsample. This model excluded the ‘Unconditional permission to eat’ subscale and retained the other three subscales from the original scale, and was found to have excellent fit to their data (CFI = 0.99, TLI = 0.98, RMSEA = 0.05). This model was also tested by Bazo Perez et al. [36] in their Hispanic-majority sample of young adults, and was similarly found to fit their data well (CFI = 0.91, TLI = 0.93, RMSEA = 0.08, SRMR = 0.08).

5.1. Psychometric properties

The vast majority of studies that reported the internal consistency of the overall IES-2 and/or its subscales achieved the cut-off of 0.7 or higher recommended by the COSMIN methodology [117], although it is notable that a smaller proportion of studies met this criteria for the ‘Unconditional permission to eat’ subscale as compared to overall IES-2 scores and the other three subscales. These findings support the internal consistency reliability of the IES-2 and at least three of the four subscales. Unfortunately, none of the studies included in the review provided evidence for any other types of reliability of the IES-2 (such as test-retest or split half).

Of the 90 studies included in the review, 78 reported mean scores for either the total IES-2 or at least one of the subscales. In a five-point scale such as that used for IES-2 responses, mean scores would be expected to fall within the 2−4 range [118]. When all studies were included in the analysis, this was the case for ‘Unconditional permission to eat’ scores but not for total IES-2 scores or the ‘Eating for physical rather than emotional reasons’, ‘Reliance on hunger and satiety cues’ and ‘Body-food choice congruence’ subscales. However, when the 17 studies that were focused on weight or eating disorders were excluded, mean scores tended more towards the centre. In this context, only total IES-2 scores and the ‘Reliance on hunger and satiety cues’ subscale fell outside of the 2−4 range, and even this was only slightly (with a highest mean score of 4.04 for the IES-2 total and 4.1 for the ‘Reliance on hunger and satiety cues’ subscale). This suggests that in samples focused on weight or eating disorders, mean scores may tend more towards the extremes of the scale.

Convergent construct validity of the scale was well supported, as the IES-2 was found to be negatively associated with several constructs it would be expected to, including BMI, depression and anxiety, disordered eating/eating disorder symptomology and rigid dietary control. Similarly, it was also positively associated with body appreciation, self-compassion and self-esteem. Unfortunately, none of the studies included in the review provided evidence for the IES-2’s discriminant construct validity.

In terms of known-groups construct validity, 13 studies tested for sex differences in intuitive eating scores and six found statistically significant differences in total IES-2 scores, with men scoring higher. Men have generally been found to have higher levels of intuitive eating than women [e.g., 2] so this supports the known-groups construct validity of the scale, i.e., that certain groups of people might be expected to score differently to others and therefore the scale should be sensitive enough to detect this [119]. Seven of the 13 studies also reported sex differences in the ‘Eating for physical rather than emotional reasons’ subscale and differences were also found in the ‘Unconditional permission to eat’ and ‘Body-food choice congruence’ subscales (one study each, respectively), each time with men scoring more highly. Finally, two studies found statistically significant sex differences in the ‘Reliance on hunger and satiety cues subscale’, again with men scoring more highly. These findings show a consistent pattern of the IES-2 behaving as it should do, both in terms of its association with constructs that it should have an association with and in terms of it detecting known sex differences in intuitive eating levels. This suggests that the instrument is indeed measuring the construct that it purports to measure, i.e., intuitive eating [115].

5.2. Dimensionality

Inter-correlations between total IES-2 scores and its subscales were reported in 11 out of the 90 studies in the review. Correlations between the subscales and total IES-2 scores were mixed, but were particularly low for the ‘Unconditional permission to eat’ and ‘Body-food choice congruence’ subscales. Both of these subscales also correlated negatively with some of the other subscales. This suggests that the ‘Unconditional permission to eat’ and ‘Body-food choice congruence’ subscales may be less strongly related to overall intuitive eating than the other subscales.

Ten studies reported on the factor structure of the IES-2. Five tested the original factor structure of four factors with a hierarchical intuitive eating factor, which was not promising in terms of fit to the data in any of the studies. Six tested a four-factor model without a hierarchical intuitive eating factor, but only one study found this model to have acceptable fit [37]; notably, this was using a sample of college women and therefore was very similar to the samples in which the four-factor conceptualisation of the IES-2 was originally developed and validated in.

Three of the studies that did not find the four-factor model to fit at all were using predominantly non-white samples [36,79,71], which suggests that the construct of intuitive eating as represented by Tylka and Kroon Van Diest’s [2] four-factor model may possibly be representative of Western or Caucasian eating styles that don’t necessarily reflect what is considered normal, healthy eating in other cultures [104]. This is in keeping with the findings of other studies [e.g., 43], who have also highlighted differences in intuitive eating according to socioeconomic status and race.

Four studies performed exploratory factor analyses [19,79,90,71] to discover whether there was a factor structure that represented their data better. All found an alternative factor structure that fit their data better than the factor structured posited in the original development and validation paper; two of these retained all 23 of the IES-2 items but arranged them into an alternative factor structure, one retained 22 of the items but reverse-scored two of the remaining items and arranged them into a four-factor structure, and the other only retained 11 of the original IES-2 items and arranged them into a three-factor model. Interestingly, this latter study, which was conducted in a Hispanic-majority sample, excluded the ‘Unconditional permission to eat’ subscale and its items, which this review has highlighted as being potentially problematic from a psychometric perspective. This factor structure has also recently been validated in another Hispanic-majority sample [36], suggesting that the original factor structure of the IES-2 may not perform as well in samples that differ from the Western college student sample that the scale was developed in, and that the ‘Unconditional permission to eat’ subscale in particular may not be as applicable to other samples or may be especially sensitive to cultural differences in eating styles or other contextual factors.

5.3. Strengths, limitations and future research directions

Although there have been systematic reviews that examine the psychosocial correlates of intuitive eating, this study is the first to specifically systematically review the psychometric performance of the IES-2 in its original English form since its publication twelve years ago. The review provides evidence for the measure’s internal consistency, as well as its convergent and known-groups construct validity, and has provided a comprehensive account of other psychological and physical health constructs with which it is associated. This review has also made novel contributions in identifying that studies relating to weight or eating disorders often report mean intuitive eating scores that tend towards the extremes of the scale, suggesting that studies of this kind may need to be considered separately in future reviews, particularly when statistical aggregation of scores or meta-analyses are being performed. This study has also identified that the four-factor structure of the IES-2 may not be representative of the construct of intuitive eating in samples that differ greatly from those in which the IES-2 was developed (i.e., Western college students).

One potential drawback to this review was that the samples of the studies included tended to be majority female and white, and mainly took place in the USA or other countries in the Global North. With the exception of gender distribution, this is likely a corollary of the study’s eligibility criteria, which stipulated that studies must be available in English to be included. In particular, the search process identified a considerable number of studies undertaken in France, Turkey and several South American and Middle Eastern countries that were excluded due to language but suggestive of a research corpus that this review was unable to access. It was also decided to exclude studies published in English but that used non-English translations of the IES-2 as it was felt that it would be difficult to discern between true cross-cultural variation and variation brought about by differences in translation processes and quality [27,28], particularly as some translations of the IES-2 do not even have the same number of items as the original English version [e.g., 120, 121]. However, future reviews would benefit from being conducted in languages other than English, incorporating translation of non-English texts to diversify the samples included, and including studies using non-English translations of the IES-2 in order to allow for the assessment of the scale’s cross-cultural validity. The authors also acknowledge that grey literature and non-indexed sources were not included in this review due to limited time and financial resources, and authors were not contacted regarding unpublished studies. As such, there is a risk of publication bias. However, this review does not report pooled effect size estimates; instead, it presents ranges across studies to reflect variability and provide a broader perspective on the findings. The authors agree that future reviews could benefit from including grey literature and non-indexed sources to further reduce the risk of bias and better capture the full scope of available evidence.

5.4. Conclusion

In conclusion, the findings of this review demonstrate that the IES-2 has good convergent and known-groups construct validity but that there are several other important areas where the IES-2 underperforms psychometrically, raising doubts about the reliability and validity of the measure. Notably, there were problems with response distribution for some of the subscales, and correlations between the subscales and the total scale score were very low for ‘Unconditional permission to eat’ and ‘Body-food choice congruence’ (raising questions about how strongly they are related to intuitive eating as an overall concept). Importantly, the original four-factor structure of the IES-2 was also found to not be a good fit in many of the studies that assessed this and an alternative factor structure that excluded the ‘Unconditional permission to eat’ subscale was identified as a viable alternative. Other than studies focusing on weight or eating disorders, there was also only one clinical sample included in the review, which means there is insufficient evidence to infer how the measure would perform in medical populations. As a result of this and the other psychometric issues identified in this review, it is therefore recommended that researchers who wish to continue to use the IES-2 validate the scale in the population of interest before using it in novel clinical samples.