Avoidant/restrictive food intake disorder (ARFID) is an eating disorder (ED) which involves being avoidant or restrictive in the food that is consumed. It was introduced in the DSM in 2013, with prevalence estimates of 16% in children and adolescents (Gonçalves et al., 2019) and up to 4% in adults (Chua et al., 2022).
In contrast to other EDs, like anorexia or bulimia, restriction around food intake in ARFID is not due to a drive for thinness or a fear of weight gain (Seetharaman & Fields, 2020). Instead, restriction is due to a fear of aversive consequences after eating food, sensory sensitivities, or a lack of interest in food or eating (Kambanis et al., 2024). At present, there has been some cross-sectional research in support of these different ARFID ‘profiles’ (e.g., Norris et al., 2018; Reilly et al., 2019; Zickgraf et al., 2019), but there are no longitudinal studies. Prospective longitudinal studies are important in research, as they can they follow the same individuals over time, eliminating sources of bias and allowing us to track the course of a disease as it happens. Studies like these are needed in the context of ARFID, including how these different profiles predict ARFID symptoms and progression. As such, Kambanis et al. (2024) aimed to evaluate the course and outcomes of ARFID over a 2-year period in a sample of young people.
ARFID is different to other eating disorders; and is often due to a fear of aversive consequences after eating food, sensory sensitivities, or a lack of interest in food or eating.
Methods
This was a prospective, longitudinal study which followed participants for up for 2 years. By prospective, we mean a type of study design which follows people over time rather than examining what has happened to them in the past (retrospective). Young people with full or subthreshold ARFID symptoms were recruited either from local hospitals or community advertisements. Individuals were excluded if they had any other ED, a substance/alcohol use disorder, or demonstrated any suicidal ideation or clinically disordered eating or exercise behaviours over the last 28 days.
At baseline, 1-year and 2-year follow-up, participants completed two measures to confirm either full or subthreshold ARFID symptoms (PARDI; Bryant-Waugh et al., 2019) and to rule out other feeding or ED diagnoses (EDA-5; Sysko et al., 2015). These measures were collected via clinical interviews conducted by research assistants and doctoral-level psychologists; when clinical interviews were not possible during follow-up, medical records were reviewed where possible.
Results
One hundred participants (49% female) between the age of 9–23 years (mean age = 15.89) took part in this study. Just over one third of the sample had received prior ARFID treatment and a range of participants reported current comorbid disorders, including: depressive or bipolar-related disorders (11%), anxiety, obsessive-compulsive or trauma-related disorders (42%), or neurodevelopmental, disruptive, or conduct disorders (21%).
1-year and 2-year follow-up data was collected for 92% (78% from clinical interviews) and 85% (74% from clinical interviews) of participants respectively.
The longitudinal course of ARFID across 2-years
- 44% of the sample persisted with their original ARFID diagnosis across both follow-up timepoints.
- 6% retained their original ARFID diagnosis at 1-year but had remitted by the 2-year follow-up; in contrast, 11% had remitted from the original ARFID diagnosis by 1-year but had relapsed at 2-years.
- A further 12% achieved remission at 1-year which was sustained at 2-years.
- Of those who had subthreshold symptoms of ARFID at 1-year, 5% had developed full ARFID symptoms by 2-years.
- Of those who had full symptoms of ARFID at 1-year, 2% had transitioned to subthreshold ARFID symptoms by 2-years.
- Of the 12 participants (12%) who presented with subthreshold ARFID at baseline, 3% transitioned to full ARFID at 1-year and 4% at 2-years.
Diagnostic crossover
Three participants (3%) experienced a diagnostic shift during the 2-year follow-up to a restricted form of Anorexia Nervosa (ANr), which was present at 1-year follow-up and maintained at 2-years for all 3 participants.
Predictors of outcome
Using a logistic regression, the authors found that greater baseline severity in food sensitivity (OR = 1.68, 95% CI [1.05 to 2.69], p = .239) and lack of interest in food/eating (OR = 1.59, 95% CI [1.06 to 2.38], p = .25) predicted greater ARFID persistence at 1-year.
Furthermore, a fear of aversive consequences at baseline did not predict ARFID persistence at 1-year (OR = 0.58, 95% CI [0.30 to 1.12], p = .104); in fact, at 2-years this was associated with ARFID remission (OR = 0.42, 95% CI [0.20 to 0.86], p = .019). Although age of participants was not found to be a predictor of ARFID outcomes (p = .653), remission rates were found to be numerically lower in older participants.
In a sample of 100 young people with ARFID, almost half (44%) remained with this diagnosis throughout the 2-year follow-up period. 12% achieved remission at 1-year follow-up and maintained this at 2-years.
Conclusions
Kambanis et al. (2024) is the first study to look at the course of ARFID longitudinally in a prospective, naturalistic way. Given the large percentage of participants experiencing a consistent diagnosis of ARFID throughout the 2-year period and the small number experiencing a crossover to a different diagnosis, these findings suggest that ARFID is both a persistent and distinct ED diagnosis.
The results of this study, including the large percentage of participants retaining a diagnosis over a 2-year period, highlights ARFID as a distinct and persistent eating disorder.
Strengths and limitations
This study had considerable strengths, including:
- A prospective longitudinal design meant the authors were able to look at the course and profiles of ARFID over time. This is advantageous to previous cross-sectional or retrospective studies which have limited causal inferences. As such, this design was less prone to sources of bias and other confounding variables, increasing its reliability and validity.
- A naturalistic design, which increased its ecological validity. Participants with comorbidities were not excluded, nor was inclusion dependent on previous treatment status. This provides a more realistic look at the course of ARFID as it is in the real world, which is therefore more insightful when thinking of real-world practice and treatments.
- Use of clinical interviews with strong psychometric properties increases the certainty we can have in the diagnoses given throughout this study, subsequently increasing the reliability of the conclusions drawn. Further, the decision to supplement data collection with information collected from medical records also meant follow-up rates and data retention was increased, which reduces bias in the study results.
However, the results must be viewed with consideration of the study’s limitations, such as:
- The modest sample size, with only 100 participants in total. Larger sample sizes can increase statistical power, which reduces the margin of error and results in more reliable results. Therefore, a modest sample size such as this may increase the risk of finding either false-positive or false-negative results.
- Lack of sample diversity. Whilst the sample has almost an equal split in terms of gender, over 90% of participants were White, and the oldest participants in this study were 23 years old. These results therefore cannot add to our knowledge or allow us to generalise these results about ARFID to different age or ethnic groups.
- Breadth of age range. This study also combined the analysis of participants from a broad age range (9-23 years). Considering that older participants in this study were found be less likely to enter remission, there may be differences in the predictors and course of ARFID across different age demographics. By combining all ages together, we are unable to dig deeper into the effect of age.
- Short follow-up period. Participants were only followed up for 2-years, which is shorter than other longitudinal studies looking at the course of other EDs. This limits our understanding of the course of the disorder beyond this point, which has implications for treatment due to the lack of evidence for how the disorder may progress.
- Quality of follow-up data. Whilst the use of medical records aided in increasing data retention, the use of notes might have impacted study outcomes, due to the authors needing to rely on quality of notes to ascertain outcomes (compared to the use of clinical interviews for other participants).
The authors of this study increased the rate of follow-up by using medical records to supplement missing data where possible. Whilst this potentially increased the power of the study, it is not as reliable as clinician interviews, which impacts the robustness of the study.
Implications for practice
The results of this study provide a much-needed insight into the longitudinal course of ARFID, showing it to be not only pervasive, but also diagnostically distinct from other EDs. Up until now, ARFID as an ED diagnosis has largely been neglected in both research and in clinical practice; in February 2024, BEAT (the UK’s leading ED charity) reported that the rise in calls they were experiencing for those with ARFID had risen by 7x (Campbell, 2024). As such, the authors of this paper sum up the need for changes in practice regarding ARFID care and support, highlighting the need for clinicians to “intervene on ARFID with the same urgency and dedication that they demonstrate when treating other eating disorders”. This should include efforts towards early detection and intervention for those with ARFID, particularly considering the results of this study where remission rates were more likely in younger participants.
The pervasive nature of the disorder, with this study showing just less than 50% of those with ARFID continuing for the entire 2-year period, also highlights the need for more effective evidence-based treatments for ARFID. Previous research indicates a need for more robust treatment trials for ARFID to be conducted (Archibald & Bryant-Waugh, 2023). Considering the results of this study, these should now be seen as essential.
Given the general neglect in research about ARFID up until now, this paper is much needed. However, with its limitations regarding sample heterogeneity and size, and length of follow-up, the results can only tell us so much. Little is currently known about the epidemiology and prevalence of ARFID across different demographic groups, particularly marginalised communities (Goel et al., 2022). There is now a need for further research in this area to expand upon the results of this study using samples with greater representation across longer periods of time.
