Validating CRAVED: Can We Make Food Addiction Easier to Spot in Clinical Practice?

In the first three blogs in this series, I explored what our Liberate research may mean for people experiencing ultra-processed food addiction. We looked at whether the programme was feasible and acceptable, what happened to binge eating symptoms, and why addictive-like eating may be a missing conversation in type 2 diabetes care.

This fourth blog turns to a practical question: how can healthcare professionals recognise that addictive-like eating may be part of the picture in the first place?

Research into food addiction has grown considerably, but identifying it in everyday practice remains difficult. The Yale Food Addiction Scale 2.0 (YFAS 2.0) is the most widely used research measure. It is comprehensive, but its 35 questions and scoring process can make it difficult to use as a quick screen during a routine appointment.

CRAVED was developed to help bridge that gap. It is a brief six-question tool designed to identify behaviours that may point towards an addictive relationship with particular foods.

Our latest research asked a simple question: does CRAVED identify the same kinds of problems as established measures of food addiction and binge eating?

What is CRAVED?

CRAVED consists of six simple yes-or-no questions. Importantly, these questions were not chosen simply because they sounded like common difficulties with food. The tool was developed by adapting the World Health Organization’s ICD-10 criteria for dependence syndrome, the established framework used to recognise addiction to substances, and applying them to a person’s relationship with certain foods.

The ICD-10 framework describes six core features of dependence: a strong desire or compulsion; difficulty controlling use; withdrawal when use is reduced or stopped; tolerance, meaning that more is needed to achieve the same effect; increasing neglect of other interests; and continued use despite clear evidence of harm.

CRAVED translates those six addiction criteria into brief, food-related questions. Each letter represents one of these features:

• Compulsion: powerful urges or persistent thoughts about consuming particular foods
• Reaching: needing increasing amounts to achieve the same effect
• Activities neglected: food taking priority over activities, relationships or responsibilities that once mattered
• Volume: eating more than intended or finding it difficult to stop
• Exclusion: experiencing withdrawal-like effects when particular foods are reduced or removed
• Damage: continuing to eat those foods despite recognising negative consequences

The result is a score from 0 to 6. The aim is not to diagnose food addiction in a few minutes. It is to help a practitioner recognise when a person's experience may warrant a fuller, non-shaming conversation or further assessment.

That distinction is important. A screening tool identifies people who may be experiencing a problem. It does not provide a clinical diagnosis.

Why did CRAVED need to be validated?

Before a screening tool can be recommended for wider use, researchers need to investigate its psychometric properties. Psychometrics simply means examining how well a questionnaire or measurement tool performs.

Researchers ask questions such as:

• Does the tool appear to measure what it is intended to measure?
• Does it identify people who are more likely to meet established criteria?
• Does it relate meaningfully to other measures of similar behaviours?
• Can it distinguish between people who do and do not appear to have the problem being screened for?

This study provided the first preliminary psychometric evaluation of the English-language CRAVED tool.

What did we do?

We carried out a secondary analysis of anonymised baseline data from 117 adults who had joined Liberate, our community-based intervention for people experiencing addictive-like eating behaviours.

Participants completed three measures:

• CRAVED: the six-question screening tool
• YFAS 2.0: the most widely used research measure of food addiction
• Binge Eating Scale (BES): a questionnaire measuring the severity of binge eating symptoms

We then examined whether CRAVED scores were associated with scores on the two established measures, and how well CRAVED distinguished between participants who did and did not meet YFAS food addiction criteria.

What did we find?

People with higher CRAVED scores also tended to report more food addiction symptoms on the YFAS. The correlation was r = 0.55 and was statistically significant.

A correlation tells us whether two measures tend to move together. In this case, higher CRAVED scores tended to occur alongside higher YFAS symptom scores. The relationship was moderate: meaningful enough to suggest that the tools are detecting related experiences, but not so strong that they should be treated as interchangeable.

CRAVED scores were also positively associated with Binge Eating Scale scores, with a correlation of r = 0.56. Participants reporting more addictive-like eating behaviours therefore also tended to report more severe binge eating symptoms.

This does not mean that food addiction and binge eating are the same condition. It supports what many people and practitioners observe: the two can overlap, while still representing distinct experiences that may require different questions and different forms of support.

Could CRAVED identify people meeting YFAS criteria?

To explore this, we used a Receiver Operating Characteristic analysis, usually shortened to ROC analysis. This looks at how well a screening tool distinguishes between people who do and do not meet a particular set of criteria.

CRAVED produced an Area Under the Curve, or AUC, of 0.75. In simple terms, this indicated an acceptable ability to distinguish between participants who did and did not meet YFAS food addiction criteria.

This is encouraging, but it is not perfect discrimination. CRAVED should help practitioners decide when to explore further, not decide the outcome of that exploration.

Why does a score of three matter?

The CRAVED tool was originally designed with a threshold of three or more. In this study, that threshold showed high sensitivity, identifying around 94% of participants who met YFAS food addiction criteria.

Sensitivity tells us how good a screening tool is at picking up people who may genuinely have the problem being considered. High sensitivity is useful at the start of a clinical conversation because missing someone who needs further support may have greater consequences than asking a few additional questions.

A score of three or more should not be interpreted as a diagnosis. It is a prompt to explore the person's experience in more detail, consider other possible explanations, and decide whether a fuller assessment or referral may be helpful.

What might this look like in practice?

Imagine someone living with obesity or type 2 diabetes who has repeatedly been advised to change their diet. They understand what they are being asked to do and genuinely want to improve their health, but they also describe:

• intense cravings or constant thoughts about particular foods
• eating more than they intended once they start
• repeated, unsuccessful attempts to cut down
• neglecting activities or responsibilities because of eating
• feeling distressed when they try to remove certain foods
• continuing to eat them despite harm to their physical health, mental wellbeing or relationships

Without the right questions, this can be interpreted as a lack of knowledge, motivation or willpower. CRAVED offers a structured way to ask whether addictive-like eating may be contributing to the difficulty.

The next step is not to attach a label. It is to listen. Which foods feel difficult to control? What happens when the person tries to stop? Are there signs of binge eating or another eating disorder? Is the person safe, adequately nourished and receiving appropriate clinical support? What has helped or made things worse in the past?

Used in this way, a brief screen can change the tone of an appointment from judgement to curiosity.

Why this matters

Food addiction is not currently recognised as a standalone diagnosis in major diagnostic systems, and debate about its classification continues. Yet people are already presenting to healthcare services describing cravings, loss of control, repeated relapse and continued eating despite harm in relation to ultra-processed foods.

If healthcare professionals do not ask about these experiences, they may miss an important reason why conventional dietary advice has repeatedly failed.

CRAVED gives practitioners a quick way to open that conversation. It may be especially useful in time-limited settings such as primary care, diabetes services and weight-management clinics, where a comprehensive research questionnaire may not be practical.

What are the limitations?

This was an initial validation study, so the findings need to be interpreted with care.

The 117 participants had already sought help for addictive-like eating through Liberate. This means the sample had high levels of the symptoms CRAVED is designed to identify and does not represent the wider population. The sample was also relatively small, and the analysis used existing data collected at a single timepoint.

CRAVED's internal consistency was modest. Internal consistency looks at how closely the questions within a questionnaire relate to one another. However, CRAVED intentionally asks about several different aspects of addictive-like eating rather than repeatedly measuring one narrow behaviour. Its brief yes-or-no format may also affect this measure.

These explanations are plausible, but they do not remove the need for further research. CRAVED should now be tested in larger and more diverse groups, against other clinical assessments, and in the settings where healthcare professionals may actually use it.

So, what can we say?

This study does not show that CRAVED can diagnose food addiction.

What it does provide is preliminary evidence that six simple questions identify patterns that relate meaningfully to established measures of food addiction and binge eating. CRAVED showed acceptable discrimination and high sensitivity at its established threshold, while remaining quick and straightforward to score.

That matters because clinical appointments are short. Sometimes the first step does not need to be a diagnosis. Sometimes it simply needs to be the right question.

Where next?

Future research should test CRAVED in:

• primary care
• diabetes services
• weight-management services
• eating disorder settings
• general population samples
• different countries and cultural groups

It will also be important to understand how CRAVED performs when healthcare professionals use it prospectively in routine clinical practice, and what forms of assessment and support should follow a positive screen.

Conclusion

Food addiction can be difficult to talk about. People may fear being judged, dismissed or told once again that they simply need more willpower.

But if healthcare professionals never ask about cravings, loss of control, withdrawal-like experiences and unsuccessful attempts to cut down, people living with these difficulties may remain invisible within services.

CRAVED was designed to make that first conversation easier.

Six questions. A few minutes. And potentially a very different clinical conversation.

Reference

Bennett E, Magdi HM, McEnaney F, Lycett D, Unwin J. Validating CRAVED: a brief screening instrument for ultra-processed food addiction. Frontiers in Public Health. 2026;14:1871164. doi:10.3389/fpubh.2026.1871164.