Psychometric evaluation of the Persian version of the adult Nine Item ARFID screen (Persian- NIAS)

Abstract

Background

Avoidant/restrictive food intake disorder (ARFID) is an eating disorder characterized by restrictive eating that leads to weight loss or failure to grow, nutritional deficiency, supplement dependence, and/or psychosocial impairment. The aim of the present study was to culturally adapt and investigate the psychometric properties of the Persian version of the adult Nine Item Avoidant/Restrictive Food Intake disorder screen (NIAS), among general population of Iranian adults.

Methods

This methodological cross—sectional study was conducted among 936 Persian-speaking adults, living in Isfahan, Iran. Translation of the NIAS was performed using forward–backward method. Intra Class Correlation (ICC) and Cronbach’s α were used to assess test–retest reliability and internal consistency, respectively. Construct validity was investigated by using exploratory factor analysis (EFA), and confirmatory factor analysis (CFA). Convergent and divergent validity was determined using BMI, adult eating behavior questionnaire (AEBQ), and Hospital Anxiety and Depression Scale (HADS).

Results

Persian version of the NIAS showed excellent test–retest reliability in all domains (ICCs: 0.989–0.995). The computed Cronbach’s alpha coefficients for dimensions of Persian-NIAS were in the range good to excellent (alpha: 0.803–0.879). Persian-NIAS showed good known-group validity and differentiated persons with different levels of BMI from each other (P < 0.05). Construct validity evaluated by EFA led to extraction of three factors (picky eating, appetite, and fear), and the CFA confirmed the adequacy of extracted construct from EFA (CFI = 0.979, TLI = 0.958, SRMR = 0.037 and RMSEA = 0.077). Convergent and divergent Validity revealed significant positive or negative correlations between Persian-NIAS dimensions and different dimensions of the AEBQ, and HADS.

Conclusions

The Persian version of the NIAS is a reliable and valid instrument with applicability in a broad range of the population of Persian-speaking adults for assessing AFRID eating disorders in research projects.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40337-026-01555-4.

Plain language summary

Avoidant/restrictive food intake disorder (ARFID) is an eating disorder characterized by restrictive eating that leads to weight loss or failure to grow, nutritional deficiency, supplement dependence, and/or psychosocial impairment. Due to the lack of ARFID assessment tools in Persian, the aim of our study was to culturally adapt and investigate the psychometric properties of the Persian version of the adult Nine Item Avoidant/Restrictive Food Intake disorder screen (NIAS), among general population of Iranian adults. Our findings suggest that the Persian version of the NIAS is a reliable and valid instrument with applicability in a broad range of the population of Persian-speaking adults for assessing AFRID eating disorders in research projects.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40337-026-01555-4.

Introduction

Avoidant/restrictive food intake disorder (ARFID) is an eating disorder characterized by restrictive eating that leads to weight loss or failure to grow, nutritional deficiency, supplement dependence, and/or psychosocial impairment [1]. The restrictive eating associated with ARFID is characterized by aversive emotional/physiological responses to food, the act of eating, and/or the short-term consequences of eating as opposed to more distal consequences (i.e., weight gain). Three presentations of ARFID restrictive eating are described in the Diagnostic and Statistical Manual of Mental Disorders (DSM-5-TR) and documented in clinically referred and community-based samples [2]. These involve food avoidance driven by food selectivity, poor appetite or a lack of interest in eating, and/or fear of aversive consequences of eating such as choking, vomiting, allergic reaction, or gut pain [3, 4].

There is limited information about the prevalence and incidence of this disorder. When DSM-5 was introduced, no published data supported the existence of the three distinct patterns of eating in ARFID or that these could lead to the ARFID impairment symptoms [5]. Since the introduction of ARFID in 2013, the three presentations have been identified in clinical samples of children, adolescents, and adults presenting for eating disorder care [6, 7]. While much of this research has been conducted in North America, data from eating disorder treatment centers in Mexico and Japan has also reproduced the three ARFID eating restrictions [8–10]. Clinical descriptions of ARIFD patients seeking intensive treatment suggest that while the drivers of ARFID are found cross-culturally, there may be cultural differences in how they manifest, e.g., selective eaters’ typical food preferences [11]. There also appear to be developmental differences in the clinical presentation of ARFID: for example, adults may be less likely than children to meet impairment criteria because of extreme food selectivity alone or because of choking and vomiting fears and more likely to report restriction due to functional gastrointestinal symptoms [6–9, 12].

Further support for the cross-cultural consistency of the three drivers of ARFID eating restrictions comes from psychometric studies of self-report symptom measures conducted in general population and convenience samples around the world. The most widely translated measure of ARFID symptoms is the nine-item ARFID screen (NIAS) [13]. In the original English, the NIAS has shown factorial, convergent/divergent, known groups, predictive validity, child/caregiver agreement in self-reporting youth and adults [13], and measurement invariance by food security status and autism diagnosis in adults [14]. Translations of the NIAS have been validated in numerous cultural contexts, including Mexico [15], Poland [16], Sweden [17], Lebanon [18], the Gulf Cooperation Council countries [19], Türkiye [20], and China [21], with evidence of factorial validity across translations, and support for convergent/divergent validity reported in most publications. Another instrument, the Eating Disorders in Youth Questionnaire, also measures the three ARFID eating restrictions and has been validated in Germany [22] and Switzerland [23]. Further, analyses of large datasets collected in routine clinical practice have also identified the three ARFID eating restrictions in national pediatric surveillance studies from the UK [24] and Canada [25].

Taken together, there is robust evidence that ARFID exists cross-culturally and is characterized by selective eating, poor appetite/low interest in food, and/or fear of aversive consequences of eating. ARFID is a serious illness, which can manifest with weight loss and malnutrition comparable to that caused by anorexia nervosa, often with an early onset and long duration of symptoms [26–29]. Even when it does not result in serious weight loss or nutritional compromise, ARFID is associated with significant social, family, and emotional impairment, and has high levels of comorbidity with mood and anxiety disorders and suicidality [30, 31]. Despite its seriousness, ARFID is relatively common compared to other restricting eating disorders. Although published prevalence estimates vary widely (e.g., 1.5–64%), a recent meta-analysis estimated the population-based prevalence—using data mostly collected in North America—to be 2.5–4.5% [32]. Given the seriousness of this illness and its cross-cultural relevance, there is a need for validated assessment tools in more languages.

Although studies on eating disorders and eating patterns have been conducted in Iran, ARFID is understudied in the Iranian context, likely because there is no standard self-report tool for AFRID screening studies in the country. Recently, a Persian translation of the Youth NIAS was validated for ages 10–19, but there no instrument currently available to measure ARFID symptoms in Persian-speaking Iranian adults [33]. The availability of a validated instrument will enable physicians, mental healthcare providers, and patients themselves to evaluate three common causes of disordered eating that often go unrecognized, particularly in adults. Our review indicated that the NIAS questionnaire is a simple, comprehensive, and valid tool in this area. Considering the cultural, linguistic, racial, and geographical differences among residents of different countries, which can affect the way questionnaires are completed and the validity of the results obtained from them, it is necessary to translate and conceptually adapt the various items of this questionnaire in order to use a scale such as NIAS for the Iranian adult population, and then examine the reproducibility and validity of this tool in the Iranian adult population. Therefore, the present study aimed to translate the NIAS questionnaire into Persian and evaluate the psychometric properties, including construct and convergent/divergent validity, internal consistency, and test–retest reliability of its Persian version. We also report descriptive statistics for subgroups of the population including gender, marital status, education attainment, health behaviors, psychiatric diagnosis history, and physical health conditions. The NIAS has shown consistent factorial validity across translations [16, 18, 20, 21]. We expected to reproduce the three-factor structure in this setting. Measures of anxiety and depression were selected for divergent validity; these symptoms are frequently comorbid with ARFID, so we expected moderate correlations with all NIAS scales [13]. The Adult Eating Behavior Questionnaire measures appetitive traits that are strongly linked to picky eating (PE) and appetite ARFID symptoms; we expected the NIAS PE scale to be relatively more strongly related to AEBQ Food Fussiness than the other subscales, the PE and Appetite subscale to be more strongly related to AEBQ Enjoyment of Eating than the Fear subscale, and the Appetite subscale to be more strongly related to AEBQ Slow Eating, Satiety Responsiveness, Emotional Under-Eating, sand Hunger (inversely) than the other two subscales [13]. In the original NIAS validation sample, the Fear subscale was unexpectedly correlated with AEBQ Emotional Over-Eating [13]. The authors hypothesized that this might suggest a role for restraint (e.g., effortful avoidance of desirable food) in some or all manifestations of fear ARFID; however, fear ARFID is heterogeneous and the inherent aversiveness of food and eating may depend on the specific feared outcome and how distal it is from the act of eating [13]. Therefore, we did not make hypotheses about the relationship between NIAS Fear and AEBQ EOE. Associations between ARFID symptoms and BMI are variable in non-clinical samples; BMI analyses were exploratory.

Method

This methodological cross-sectional study was conducted from September 2019 to April 2020 among 936 Persian-speaking individuals in Isfahan, the largest city in central Iran. Participants were chosen from urban health centers in Isfahan using a multistage cluster sampling method. Initially, the health centers, serving as the primary clusters, were selected from Isfahan Health Centers I and II, the two key clusters. We randomly chose 6 and 9 urban health centers from Isfahan health centers I and II as the second-stage clusters, respectively. Next, at each chosen health center, individuals who met our inclusion criteria were identified. The study's inclusion criteria required participants to be 18 years or older, able to read and write Persian, and permanent residents of Isfahan city. The exclusion criteria included the following: being pregnant, experiencing psychological and cognitive problems, suffering from a physical illness during the interview, and not completing all questionnaires. We explained the study's objectives to all eligible participants and then invited them to participate in our research study. A trained interviewer performed the interviews. All participants received sufficient information about the study and provided informed consent.

Procedures

The Nine Item ARFID screen (NIAS)

The NIAS is a self-report questionnaire comprising nine items designed to assess avoidant/restrictive eating patterns [13]. The NIAS consists of three subscales: the picky eating subscale measures aversion to food or food selectivity (e.g., “I dislike most foods that other people eat”), the appetite subscale measures low appetite or limited interest in eating or food (e.g., “Even when I am eating foods I really like, I struggle to consume a sufficient volume during meals”), and the fear subscale measures fears of negative consequences related to eating (e.g., “I avoid or delay eating because I fear gastrointestinal discomfort, choking, or vomiting”). Participants answer each question using a scale from 1 (Strongly Disagree) to 5 (Strongly Agree). Each subscale receives a score ranging from 3 to 15, where higher scores reflect increased levels of factors like picky eating, lack of interest, and fear. The validity and reliability of this tool have been previously confirmed [13].

Adult eating behaviour questionnaire

The AEBQ consists of 35 items that assess eight appetitive traits based on a 1–5 Likert scale for agreement or disagreement scale [34]. The eight scales consist of three to five items each, with sample items included for every scale in parentheses. Food responsiveness refers to the desire to eat when one sees or smells appealing food. Hunger manifests as light-headedness when meals are delayed. Enjoyment of eating is characterized by a love for food. Emotional overeating occurs when one consumes more food when upset, while emotional undereating means eating less under the same circumstances. Satiety responsiveness indicates that one often feels full before finishing a meal. Slow eating is when someone typically takes longer to complete a meal than others. Food fussiness describes the tendency to decide one dislikes a new food before even trying it. Both previous validation studies raised concerns about the eight-factor structure of the AEBQ, with each group noting a minor enhancement in fit when either removing the Hunger items or permitting them to be associated with Food Responsiveness scale. [34]. A freely downloadable PDF copy of the original AEBQ is available from the developers at http://www.ucl.ac.uk/iehc/research/behavioural-science-health/resources/questionnaires/eating-behaviour-questionnaires. The AEBQ measure's validity and reliability have been evaluated and validated in prior studies (alpha values range from 0.75 to 0.90, and ICCs range from 0.73 to 0.91) [34, 35].

Hospital anxiety and depression scale (HADS)

A 14-item self-report screening tool, the Hospital Anxiety and Depression Scale (HADS), was employed to evaluate anxiety and depression levels among participants. In the Iranian population, the questionnaire demonstrated an internal consistency with a Cronbach’s alpha coefficient of 0.78. It utilizes a 4-point Likert scale, where scores range from 0 (not present) to 3 (considerable). Participants’ scores for anxiety or depression can vary from 0 to 21 points (0–7: normal; 8–21: indicates mild, moderate, or severe disorder) [36].

Translation and content validity

Permission was obtained from the initial developer. The approach suggested by Beaton et al. was employed to translate the NIAS from English into Persian [37]. Two separate professional translators translated the items into Persian (forward translation). One translator was experienced with translating texts related to nutrition, whereas the other did not work in this area. Subsequently, the researchers of the current study (Z.H. and A.F.) and both translators created a unified forward version. This questionnaire was then translated back into English by two bilingual translators to assess its conceptual equivalence with the original. After a comprehensive evaluation by the researchers (Z.H. and A.F.), significant modifications were implemented, leading to the creation of the provisional Persian version of the NIAS questionnaire. Overall, the translation process encountered no challenges. As a result, we evaluated content validity by determining the Content Validity Index (CVI) and Content Validity Ratio (CVR). The Content Validity Index (CVI) evaluates the simplicity, relevance, and clarity of each item concerning the construct assessed by the scale. We enlisted ten professionals (nine nutritionists (familiar with eating disorders) and one biostatistician) to rate the simplicity, relevance, and clarity of the Persian NIAS items on a four-point scale. The experts evaluated the relevance of the items using the following categories: (1) not relevant; (2) slightly relevant; (3) relevant; and (4) completely relevant. CVI is calculated by averaging the number of experts who give each item a 3 or 4, with a score of 1 reflecting unanimous agreement that the item is relevant or completely relevant; A minimum acceptable CVI score for each item was established at ≥ 0.79. The CVR measures the necessity of each item. To calculate the CVR, ten experts rated the essentiality of the Persian NIAS items on a three-point scale: 1 for unnecessary, 2 for useful but unnecessary, and 3 for necessary; CVR is scored in the same way as CVI. A CVR score of ≥ 0.62 was deemed satisfactory for each item [38, 39]. Face validity was quantitatively assessed by gathering the opinions of 20 adults (which selected from the general adult population) on the importance of each item and its alternatives using a 5-point Likert scale ranging from “not important at all” (score = 1) to “very important” (score = 5). Subsequently, each item's impact score was calculated by multiplying the importance coefficient by its relative frequency; items that scored ≥ 1.5 were retained in the questionnaire. Finally, the Persian version of the NIAS was developed and utilized to evaluate its psychometric properties.

Validity

Construct validity

The NIAS factor structure was examined through EFA, and CFA on 936 general adults. Using the cross-validation method [40], we randomly divided our sample into two subsamples. Two random samples were selected using random number generation from a uniform distribution (between 0 and 1) using software. A cut-off 0.5 was used to halve the resulting values. EFA was conducted on the first subsample (training sample) utilizing the principal component extraction method, orthogonal Varimax rotation and oblique Promax rotation. We retained factors for further analysis based on eigenvalues and the Scree plot. We considered factor-item loading values above 0.40 and factors with eigenvalues exceeding 1 as thresholds to yield more interpretable factors that explain adequate overall variation. We also used the criterion proposed in Norman and Streiner's book to determine meaningful factor loadings. The first subsample includes 468 participants, so factor loadings greater than CV = 5.152/sqrt(N-2) = 0.24 are retained [41]. The Kaiser–Meyer–Olkin (KMO) measure of sample adequacy (values above 0.8) and Bartlett’s test of Sphericity (P < 0.05) were employed to assess the data viability for factor analysis [40]. We also used parallel analysis, Minimum Average Partial Test, and the Empirical Kaiser test to determine the number of factors [41]. The free “psych”, “EFA.dimensions” packages of R software were used for this purpose. Based on the loaded items for each factor, we assigned labels to the extracted factors. We calculated the score for each sub-scale (factor) by summing the items multiplied by their respective loadings and then assigned these scores to each participant. Next, we performed a Confirmatory Factor Analysis (CFA) on the validation sample to verify the factor structure obtained from the Exploratory Factor Analysis (EFA). We used the Comparative Fit Index (CFI) ≥ 0.9, Tucker-Lewis Index (TLI) > 0.9, and Root Mean Square Error of Approximation (RMSEA) < 0.08 to assess the goodness of fit of the CFA [42]. Lower values of RMSEA, along with higher values of TLI and CFI, indicate superior model fit.

Convergent and divergent validity

To evaluate convergent and divergent validity of the Persian NIAS subscales (Picky Eating, Appetite, Fear), we specified on a priori, theory-driven framework linking each NIAS domain to particular AEBQ dimensions, HADS subscales, and BMI. Picky eating was hypothesized to converge most strongly with food fussiness and to show inverse associations with enjoyment of food and food responsiveness; Appetite was hypothesized to converge with satiety responsiveness, slowness in eating, and reduced enjoyment of food (reflecting low hedonic drive and slower intake); Fear was expected to show modest convergence with behavioral proxies of aversive interoception (operationalized here via satiety responsiveness and food fussiness) and exploratory links to emotional eating dimensions. For divergent validity we selected HADS scale and BMI, reasoning that ARFID-related symptom domain may co-occur with general distress and weight indices but should not exhibit strong correlation with them; we therefore pre-specified modest convergent correlations (approximately r≈0.2–0.4) and conservative divergence thresholds (|r|> 0.3–0.4). Because NIAS subscales are theoretically inter-correlated (rs = 0.2–0.3), we planned to report both zero-order and partial Pearson correlations (each NIAS subscale controlling for the other two), emphasize effect size magnitude and relative differences across subscales rather than sole reliance on statistical significant. In this study, convergent and divergent validity was conducted with the participation of 220 adults.

Reliability

To assess internal consistency and test–retest reliability, we recruited 50 participants. They completed the Persian-NIAS measure on two separate days with a 10-day interval between tests. We estimated the test–retest reliability using the ICC coefficient at a 95% confidence level, applying a two-way mixed model. An ICC value above 0.70 was considered excellent reliability. Additionally, we evaluated internal consistency with Cronbach’s α coefficient, rating values between 0.70 and 0.95 as satisfactory [43]. We also evaluated the degree of "ceiling and floor effects” by examining the distribution of Persian-NIAS scores in the studied population.

Other statistical analysis

In this paper, quantitative and qualitative variables were expressed as mean(SD) and number (percent), respectively. We applied analysis of variance (ANOVA) to compare the mean total score of the Persian-NIAS and its dimensions across different sub-groups including gender, marital status, education attainment, health behaviors, psychiatric diagnosis history, and physical health conditions. The skewness, kurtosis, q-q-plot, and Kolmogorov–Smirnov test were used to assess the normality of the data distribution, and the Levene test was used to assess the homogeneity of variance. Appropriate transformations were used for non-normal data. In all statistical analyses P-value < 0.05 was considered as significant level. All analyses were conducted using SPSS software (version 16; SPSS Inc., Chicago, IL, USA).

Results

Participant characteristics

A total of 936 general adults, including 603 (64.4%) females, participated in the current study. The mean (SD) age was 40.03 ± 12.8 years. Almost 50% of the participants were obese or overweight. Nearly, 80% of the participants had never smoked. About 7% of the participants had a history of malnutrition. History of psychological and physical problems was reported by 57 (6.1%) and 217 (23.2%) of participants (Table 1).

Table 1.

Participant`s characteristics

Age(years)		40.03 ± 12.8
Gender	Female	603(64.4)
	Male	333(35.6)
Educational	Illiterate	22(2.4)
	Under diploma	172(18.4)
	Diploma	282(30.1)
	College	460(49.1)
Marital status	Single	115(12.3)
	Married	800(85.5)
	Widow	21(2.2)
Income status	Inadequate	233(24.9)
	Middle	515(55.0)
	Adequate	188(20.1)
History of malnutrition	Yes	65(6.9)
	No	871(93.1)
History of psychological problems	Yes	57(6.1)
	No	879(93.9)
History of physical problems	Yes	217(23.2)
	No	719(76.8)
Smoking	Current Smoker	85(9.1)
	In past	76(8.1)
	Never	775(82.8)
BMI	Under weight	18(1.9)
	Normal	415(44.3)
	Overweight	390(41.7)
	obese	113(12.1)

Values are mean ± SD or frequency (percentage)

Content and face validity

The expert panel checked the difficulty, relevance, and ambiguity of the wording and phrasing of the Persian NIAS items. The CVI ranged between 0.9 and 1.00 for all items of the Persian NIAS. In addition, the CVR ranged between 0.8 and 1.00 for all items. Consequently, no items were deleted (Table 2).

Table 2.

Relevance, simplicity, clarity, Item Content Validity Index (I-CVI), and Content Validity Ratio (CVR) values of the Persian version of Nine Item Avoidant/Restrictive Food Intake disorder screen (Persian—NIAS)

Items	CVR	Simplicity	Clarity	Relevance	I-CVI	Impact score
1. I am a picky eater 1. من هر غذایی را نمی‏خورم (بد غذا هستم)	1	0.9	0.9	1	0.93	2.25
2. I dislike most of the foods that other people eat 2. من بیشتر غذاهایی که دیگران می‌خورند را دوست ندارم	1	1	1	1	1	3.86
3. The list of foods that I like and will eat is shorter than the list of foods I won't eat 3. فهرست غذاهایی که دوست دارم یا می‌خورم کوتاه‌تر از فهرست غذاهایی‌ است که نمی‌خورم یا نمی‏خواهم بخورم	1	0.9	0.9	1	0.93	4.36
4. I am not very interested in eating; I seem to have a smaller appetite than other people 4. خیلی اهل خوردن نیستم؛ به نظر می‌رسد نسبت به دیگران اشتهای کمتری دارم	1	1	1	1	1	2.48
5. I have to push myself to eat regular meals throughout the day, or to eat a large enough amount of food at meals 5. باید خودم را وادار کنم که در طول روز به طور منظم وعده‌های غذایی را میل کنم، و در هر وعده به مقدار کافی غذا بخورم	1	1	1	1	1	2.96
6. Even when I am eating a food I really like, it is hard for me to eat a large enough volume at meals 6. حتی وقتی غذایی را که واقعاً دوست دارم مصرف می‌کنم، برایم دشوار است که در یک وعده مقدار زیادی غذا بخورم	1	1	1	1	1	3.74
7. I avoid or put off eating because I am afraid of GI discomfort, choking, or vomiting 7. من از خوردن غذا اجتناب می‌کنم و یا آن را به تأخیر می‌اندازم زیرا می‌ترسم دل‌درد بگیرم، خفه شوم و یا بالا بیاورم	1	1	1	1	1	3.61
8. I restrict myself to certain foods because I am afraid that other foods will cause GI discomfort, choking, or vomiting 8. خودم را به خوردن غذاهای بخصوصی محدود می‌کنم زیرا می‌ترسم غذاهای دیگر باعث شوند دل‌درد بگیرم، خفه شوم یا بالا بیاورم	0.8	1	1	1	1	2.81
9. I eat small portions because I am afraid of GI discomfort, choking, or vomiting 9. من غذایم را به صورت لقمه ‏های کوچک می‌خورم زیرا می‌ترسم دل‌درد بگیرم، خفه شوم و یا بالا بیاورم	1	1	1	1	1	2.96

Construct validity

EFA with Varimax and Promax rotations extracted three factors from the NIAS measure which were labeled as “picky eating”, “fear” and “appetite” accounting for 28.6, 26.6 and 23% of the total variance. A KMO value of 0.848 and P < 0.05 for Bartlett’s test confirmed the sample size adequacy and data factorability, respectively. Table 3 provides the factor loadings of three extracted factors from EFA on the 9 items of the NIAS measure. Also, all three methods of Parallel Analysis, Minimum Average Partial Test, and the Empirical Kaiser test estimated the number of factors to be 3 (Supplementary materials). The results obtained from the CFA indicated a good fit according to the goodness of fit indices as follows: CFI = 0.979, TLI = 0.958, SRMR = 0.037 and RMSEA = 0.077; also, all items loaded significantly on their respective factors (Table 3 and Fig. 1).

Table 3.

Factor loadings of Persian version of Nine Item Avoidant/Restrictive Food Intake disorder screen (Persian—NIAS)

	EFA a			EFA b			CFA
	Factor 1 (Picky eating)	Factor 2 (Fear)	Factor 3 (Appetite)	Factor 1 (Picky eating)	Factor 2 (Fear)	Factor 3 (Appetite)	Factor 1 (Picky eating)	Factor 2 (Fear)	Factor 3 (Appetite)
1. I am a picky eater	0.896			0.963			0.733
2. I dislike most of the foods that other people eat	0.898			0.955			0.758
3. The list of foods that I like and will eat is shorter than the list of foods I won't eat	0.763			0.749			0.861
4. I am not very interested in eating; I seem to have a smaller appetite than other people			0.673			0.659			0.811
5. I have to push myself to eat regular meals throughout the day, or to eat a large enough amount of food at meals			0.794			0.864			0.673
6. Even when I am eating a food I really like, it is hard for me to eat a large enough volume at meals			0.828			0.911			0.650
7. I avoid or put off eating because I am afraid of GI discomfort, choking, or vomiting		0.843			0.862			0.813
8. I restrict myself to certain foods because I am afraid that other foods will cause GI discomfort, choking, or vomiting		0.889			0.933			0.764
9. I eat small portions because I am afraid of GI discomfort, choking, or vomiting		0.799			0.823			0.736
Variance explained* (%)	28.6	26.6	23

^a Exploratory factor analysis with Varimax rotation; ^b Exploratory factor analysis with Promax rotation; * Variance explained resulted from factor analysis

Values are factor loadings. EFA: exploratory factor analysis; and CFA: confirmatory factor analysis

Fig. 1.

Confirmatory factor analysis testing the extracted construct from EFA on the Persian-NIAS items

Convergent and divergent validity

The zero-order and partial Pearson correlations between all dimensions of the adult eating behavior questionnaire (AEBQ) and the Persian-NIAS subscales is reported in Table 4. The three dimensions of “Enjoyment of food”, “Emotional over-eating”, and “Food responsiveness” showed a negative and significant correlation with all dimensions of the Persian-NIAS questionnaire (P < 0.05). In contrast, the four dimensions of “Emotional under-eating”, “Food fussiness”, “Slowness in eating”, and “Satiety responsiveness” showed a positive and significant relationship with all dimensions of the questionnaire (P < 0.05). Depression and anxiety showed a positive and significant relationship with the total score of the questionnaire and all its dimensions except the “Fear” dimension (P < 0.05). Participants' BMI showed a negative and significant correlation with the total score of the questionnaire and all its dimensions except the “Fear” dimension (P < 0.05). When partial correlations were used to control for shared variance among the NIAS subscales, picky eating (PE) was independently related to AEBQ Food Fussiness, and Appetite to Emotional Under Eating, Satiety Responsiveness, and Slow Eating (at a trend level) and inversely related to Enjoyment of Eating and Food Responsiveness (Table 4).

Table 4.

Correlations of the sub-scales score of Persian-NIAS questionnaire with BMI, psychological problems and AEBQ sub-scales to assess the convergent and divergent validity

	Picky eating		Appetite		Fear		NIAS Score
	Zero-order	Partial a	Zero-order	Partial a	Zero-order	Partial a
Enjoyment of food	− 0.125**	− 0.057	− 0.206**	− 0.213**	− 0.095**	0.049	− 0.177**
P-Value	< 0.001	0.405	< 0.001	0.002	0.004	0.467	< 0.001
Emotional over-eating	− 0.081*	0.096	− 0.148**	− 0.227**	− 0.071*	0.083	− 0.125**
P-Value	0.013	0.156	< 0.001	0.001	0.029	0.221	< 0.001
Emotional under-eating	0.165**	− 0.025	0.151**	0.236**	0.081*	− 0.076	0.168**
P-Value	< 0.001	0.714	< 0.001	0.000	0.013	0.262	< 0.001
Food fussiness	0.296**	0.456**	0.195**	0.056	0.202**	0.029	0.288**
P-Value	< 0.001	< 0.001	< 0.001	0.406	< 0.001	0.668	< 0.001
Food responsiveness	− 0.151**	0.044	− 0.262**	− 0.189**	− 0.158**	0.028	− 0.234**
P-Value	< 0.001	0.522	< 0.001	0.005	< 0.001	0.685	< 0.001
Hunger	0.033	0.171*	− 0.111**	− 0.002	− 0.041	0.014	− 0.045
P-Value	0.307	0.011	0.001	0.973	0.210	0.834	0.169
Slowness in eating	0.071*	− 0.067	0.072*	0.129	0.059	0.037	0.085**
P-Value	0.029	0.320	0.027	0.057	0.070	0.587	0.010
Satiety responsiveness	0.254**	0.101	0.288**	0.449**	0.224**	− 0.049	0.315**
P-Value	< 0.001	0.136	< 0.001	0.000	< 0.001	0.469	< 0.001
anxiety	0.196**	0.126	0.204**	0.140*	0.090	− 0.023	0.214**
P-Value	0.003	0.062	0.002	0.039	0.183	0.734	0.001
depression	0.148*	0.088	0.186**	0.145*	0.044	− 0.053	0.168*
P-Value	0.028	0.195	0.006	0.033	0.518	0.432	0.012
BMI	− 0.112**	− 0.069	− 0.095**	− 0.171*	0.048	0.254**	− 0.076*
P-Value	0.001	0.308	0.004	0.011	0.145	< 0.001	0.019

^**Correlation is significant at the 0.01 level; *Correlation is significant at the 0.05 level;

AEBQ: Adult eating behavior questionnaire; ^a Partial correlations: each NIAS subscale controlling for the other two

Reliability analyses

The reliability and descriptive statistics for the three Persian-NIAS scales are shown in Table 5. The ICC coefficient for the total score of the Persian-NIAS suggests strong test–retest reliability (ICC = 0.995, 95% CI: 0.993 to 0.996; P < 0.001). The ICC coefficients for the extracted subscales including “Picky eating”, “Appetite”, and “Fear” were estimated to be more than 0.9.

Table 5.

Descriptive statistics and reliability data for the Persian-NIAS total items and its subscales

	Mean (SD)	Cronbach’s α	ICC (%95CI)	Floor (%)	Ceiling (%)
Picky eating	7.40 (3.7)	0.879	0.995 (0.993, 0.995)	0.1	5.9
Appetite	7.30 (3.3)	0.803	0.993 (0.988, 0.994)	0.1	2.0
Fear	5.22 (2.7)	0.850	0.989 (0.983, 0.993)	0.2	1.1
NIAS Score	19.91 (7.91)	0.878	0.995 (0.993, 0.996)	0.1	0.5

ICC, Intra Class correlation Coefficient

Cronbach’s alpha coefficient to indicate item internal consistency reliability for each scale is presented in Table 5 and all scales showed satisfactory results (varied from 0.803 to 0.879). The Cronbach’s alpha coefficient for the total score of the Persian-NIAS suggests strong internal consistency (Cronbach’s α = 0.878). The percentage of respondents scoring at the highest level (i.e., ceiling effect) was between 0.5 to 5.9% for all sub-scales, while the percentage of participants scoring at the lowest level (i.e., floor effect) was minimal for all sub-scales.

Descriptive statistics

We compared mean score of Persian-NIAS’s subscales between different sub-groups (Table 6). Mean ± SD of the “Picky eating” subscale and total score of the Persian-NIAS was significantly lower in people with university education compared to other educational groups (P < 0.05). Total score of the Persian-NIAS was significantly lower in married participants (P = 0.011). For participants with a history of malnutrition, the “Picky eating” score and the total score of the questionnaire were significantly higher (P < 0.05). For individuals with underweight-range BMI (people with a BMI of less than 18.5 kg/m²), the mean score of “Picky eating” was significantly higher than other groups (P = 0.033). However, no significant difference was observed in terms of the mean value of the Persian-NIAS subscales in gender, employment status, income status, history of physical/psychological problems groups (Table 6).

Table 6.

Comparison of the sub-scales score of Persian-NIAS questionnaire between different characteristics of participants (known-group validity)

		Picky eating	Appetite	Fear	NIAS Score
Gender	Female	7.26 (3.61)	7.38 (3.29)	5.2 (2.63)	19.83 (7.94)
	Male	7.64 (3.80)	7.15 (3.33)	5.27 (2.74)	20.07 (7.87)
	P-Value	0.13	0.313	0.682	0.663
Educational status	Illiterate	8.95 (5.05)	7.5 (4.52)	6.23 (4.33)	22.68 (12.4)
	Under diploma	7.78 (4.50)	6.92 (4.06)	5.38 (3.32)	20.09 (10.6)
	Diploma	7.75 (3.42)	7.78 (3.29)	5.37 (2.58)	20.9 (7.35)
	College	6.97 (3.36)	7.13 (2.88)	5.02 (2.31)	19.11 (6.62)
	P-Value	0.002	0.022	0.068	0.007
Marital status	Single	7.82 (3.78)	8.03 (3.15)	5.37 (2.57)	21.23 (7.55)
	Married	7.31 (3.65)	7.15 (3.29)	5.18 (2.66)	19.63 (7.87)
	Widow	8.57 (3.88)	9 (3.95)	6.1 (3.46)	23.67 (9.88)
	P-Value	0.128	0.001	0.243	0.011
Employment status	Employed	7.22 (3.55)	7.07 (3.19)	5.1 (2.53)	19.39 (7.35)
	Housekeeper	7.39 (3.71)	7.38 (3.44)	5.21 (2.72)	19.97 (8.30)
	Unemployed	7.83 (3.87)	7.62 (3.19)	5.54 (2.83)	20.99 (8.05)
	P-Value	0.218	0.177	0.228	0.104
Income status	adequate	7.12 (3.34)	7.35 (3.05)	5.16 (2.49)	19.63 (6.88)
	Middle	7.29 (3.45)	7.25 (3.10)	5.24 (2.59)	19.76 (7.42)
	inadequate	7.87 (4.35)	7.37 (3.90)	5.24 (2.98)	20.48 (9.58)
	P-Value	0.071	0.864	0.936	0.447
Smoking status	Current Smoker	7.88 (4.03)	6.85 (3.52)	4.85 (2.86)	19.58 (8.07)
	In past	8.56 (4.07)	7.49 (3.45)	5.63 (2.83)	21.68 (8.21)
	Never	7.23 (3.58)	7.33 (3.27)	5.23 (2.63)	19.78 (7.85)
	P-Value	0.005	0.386	0.183	0.128
History of physical problems	Yes	7.7 (4.02)	7.58 (3.68)	5.38 (2.94)	20.65 (9.12)
	No	7.31 (3.57)	7.21 (3.18)	5.18 (2.58)	19.69 (7.51)
	P-Value	0.175	0.158	0.336	0.12
History of psychological problems	Yes	7.89 (4.28)	7.75 (4.10)	5.58 (3.22)	21.23 (10.42)
	No	7.37 (3.64)	7.27 (3.25)	5.2 (2.63)	19.83 (7.72)
	P-Value	0.294	0.282	0.299	0.196
History of malnutrition	Yes	8.38 (4.70)	7.85 (4.24)	5.75 (3.49)	21.98 (11.05)
	No	7.33 (3.58)	7.26 (3.22)	5.18 (2.60)	19.76 (7.61)
	P-Value	0.025	0.166	0.096	0.029
BMI	Under weight	8.22 (4.33)	8.89 (3.36)	6.39 (2.95)	23.50 (9.80)
	Normal	7.71 (3.97)	7.44 (3.52)	5.03 (2.60)	20.19 (8.28)
	Overweight	7.23 (3.38)	7.18 (3.11)	5.24 (2.60)	19.65 (7.27)
	Obese	6.72 (3.34)	6.91 (3.10)	5.66 (3.02)	19.24 (8.25)
	P-Value	0.033	0.074	0.036	0.142

Values are mean (SD); *P-value resulted from independent samples t-test or ANOVA

Discussion

Given the lack of validated instruments to measure adult ARFID symptoms in Iran, our aim was to contribute to the literature on eating disorders by validating the Persian version of the NIAS in a large sample of Persian-speaking adults (N = 936) from Isfahan (a city in central Iran). As expected, analyses showed that the Persian NIAS for Adults exhibited a three-factor solution with excellent levels of internal consistency and test–retest reliability. The measure also demonstrated convergent validity with the AEBQ, a measure of appetitive traits related to food approach and avoidance, symptoms of depression and anxiety which are often co-occurring with ARFID. In addition, the mean score of the Persian NIAS dimensions showed differences based on education level and history of malnutrition, as well as good correlation with other measures. Finally, a pattern of relationships between NIAS subscales and BMIs that has been previously demonstrated in a sample of young adults from China was replicated, with small but significant negative correlations between BMI and the picky eating and appetite subscales. Given these findings, we recommend the NIAS as a measure of dimensional ARFID eating restrictions among Persian-speaking adults. Providing this psychometrically sound Persian version of the NIAS could help provide accurate epidemiological data on ARFID in Persian-speaking countries, increase awareness of ARFID screening and diagnosis in Persian-speaking settings, and aid in the development of culturally appropriate and evidence-based interventions.

We used a cross-validation approach to examine the factor structure of the adult Persian NIAS. Based on this approach, both EFA and CFA provided evidence for goodness-of-fit indices for the three-factor model (i.e., picky eating, appetite, and fear) for the 9-item version of the Persian NIAS, which supports the subdomains proposed in the original scale [13]. Other linguistic validations of the NIAS, including the US (original development and validation population) [13], Mexico [15], Turkye [20], Poland [16], and China [21], and most recently in Iranian youth, were also able to replicate the main factor structure. Across all of these samples, the three-factor structure has been supported with goodness-of-fit measures ranging from RMSEA = 0.036–0.092 and CFI = 0.97 to 0.99. In EFA, the factors have consistently been shown to explain 58–78% of total variance across the three factors. All items measure their intended latent trait well, with CFA loadings ranging from 0.47 to 0.99

Furthermore, adequate internal reliability was obtained for the Persian NIAS, with Cronbach’s alpha for the total score and all three subscales ranging from 0.803 to 0.879. Similarly, good reliability of the NIAS was demonstrated in initial validation [13] and subsequent validations (e.g., Cronbach’s alpha 0.81–0.91 for the Turkish version [20], 0.73–0.86 in Chinese students [21], 0.74–0.84 in Mexican youth [15], or 0.84–0.99 for the Polish version [16]). We also examined external reliability using the ICC index and found acceptable external reliability for the Persian version of the NIAS. Several previous validation studies have assessed test–retest reliability, indicating moderate to acceptable reliability indicative of stable measurement properties over time ((e.g. ICC of 0.62–0.72 in the original version [13], ICC of 0.7–0.9 in Mexican youth [15]).

Similar to the initial validation [13], as well as the Turkish version [20], we used the AEBQ dimensions to examine convergent and divergent validity. As expected,, when partial correlations were used to control for shared variance among the NIAS subscales, PE was independently related to AEBQ Food Fussiness, and Appetite to Emotional Under Eating, Satiety Responsiveness, and Slow Eating (at a trend level) and inversely related to Enjoyment of Eating and Food Responsiveness. Partial correlations supported the specificity of relationships between the PE and Appetite subscales and AEBQ scales measuring related constructs, although there were some differences in the pattern of findings between this and the US validation sample. Specifically, in the current sample, PE was not independently related to eating enjoyment, and Appetite was not inversely related to Hunger.

Our findings also showed that appetite and picky eating, but not fear, were positively correlated with anxiety and depression. Zickgraf and Ellis [13] reported a positive correlation between appetite subscale with depression and also between fear with anxiety in the US validation sample. In addition, He et al. [21] showed that fear and appetite subscales were each independently associated with psychological distress in the Chinese validation sample., whereas picky eating was not. Finally, He and colleagues found that both picky eating and appetite were weakly but significantly negatively related to BMI, whereas in the original US validation sample there was no relationship between picky eating and BMI. A meta-analysis based primarily on data from samples from the US, Canada, Europe, and Australia reported no relationship between picky eating and BMI, although there was notable heterogeneity, with some null, positive, and negative correlations reported across studies [44]. Somewhat unexpectedly, in the current sample Fear was positively related to BMI when controlling for other NIAS subscales; Appetite showed the expected negative relationship.

The mixed findings suggest the presence of unmeasured moderators, one of which may be related to cross-cultural differences between adults in the US, Asia, and the Middle East, perhaps related to differing food cultures and risk of overweight and obesity.

The impact of ARFID eating restrictions on weight gain likely depends on the food environment and cultural practices around food and eating. For example, in food environments characterized by an abundance of highly palatable, highly caloric foods that are more affordable than less-processed whole foods (as in the United States), appetite ARFID symptoms may be relatively protective against overweight, whereas in food environments where whole foods are more accessible or in cases of food insecurity characterized by hunger, it may be linked to risk of underweight [21, 45]. Relatedly, picky eating (PE) is associated with a preference for highly palatable foods, particularly those associated with childhood [10, 11]. Because “childhood” foods are culture-specific, there may be cultural differences in the association between PE and risk of obesity, with, for example, lower risk in settings where children’s diets involve less calorically-dense/highly palatable foods [21]. Conversely, the highly processed foods that are part of selective eaters’ diets in the US tend to also be nutritionally fortified; in settings where processed food is less abundant or less fortified, PE may be a greater risk factor for malnutrition than it is in the US. Further, symptoms of PE and Appetite ARFID are correlated, and these eating restrictions appear to co-occur at clinically significant levels at higher frequency than other combinations of ARFID presentations [12, 24, 46, 47]. When PE co-occurs with Appetite ARFID symptoms, it is robustly associated with low weight (and in children, poor growth) and involves limited intake from a low-variety diet. The combination of PE and poor appetite is associated with greater risk for chronic under-nutrition compared to either ARFID eating restriction alone. The finding that NIAS Fear was moderately positively related to BMI was unexpected and, to our knowledge, inconsistent with other cross-cultural validation studies that have reported this result [12, 24]. We hope that the availability of a Persian NIAS for adult respondents will allow for more research on this association; with more effects from new samples, the moderately strong relationship observed in our sample may prove to be an outlier or reflective of another unmeasured moderator in the relationship between Fear ARFID and BMI specific to our sample or the Iranian context.

Strengths and limitations

In this study, the validity and reliability of the NIAS questionnaire were comprehensively and extensively evaluated and confirmed from various aspects. The large sample size of this study is also one of its strengths. Despite these advantages, there are limitations. This study was conducted on a sample only from Isfahan city, which affects the generalizability of its results to all Persian speakers living in Iran. The sample was randomly selected from the entire population. However, after randomly selecting clusters (city-level health centers), convenience sampling was performed within each cluster. Given that women tend to visit health centers more frequently than men and that individuals with higher education are usually more inclined to respond to surveys, the distribution of demographic variables in the present study is not balanced. To assess discriminant validity (using ROC analysis), we did not have access to a sample of patients with ARFID. Therefore, it is suggested that future studies conduct research in this area to determine the appropriate cutoff point for this questionnaire to identify individuals with this disorder. Finally, cross-cultural validation was not conducted in the present study.

Conclusion

The results of this study support the validity and reliability of the Persian language NIAS questionnaire, in individuals over 18 years of age in Isfahan province. This concise and comprehensible questionnaire can be utilized to assess this disorder in Persian-speaking adult populations at health centers and across various research studies.

Abbreviations

ARFID

Avoidant/restrictive food intake disorder

NIAS

Nine item ARFID screen

ICC

Intra class correlation

EFA

Exploratory factor analysis

CFA

Confirmatory factor analysis

AEBQ

Adult eating behavior questionnaire

HADS

Hospital anxiety and depression scale

CVI

Content validity index

CVR

Content validity ratio

KMO

Kaiser–Meyer–Olkin

CFI

Comparative fit index

TLI

Tucker-Lewis index

RMSEA

Root mean square error of approximation

Author contributions

SB: Software, Formal analysis, Writing – original draft; ZH: Conceptualization, Methodology, Investigation, Writing – review & editing, Funding acquisition, Supervision; AF: Conceptualization, Investigation, Writing – review & editing; HZ: Investigation, Writing – review & editing.

Funding

This article is the result of a student thesis that was written on the approved project of Isfahan University of Medical Sciences with research project No: 198021, ethical approval No: IR.MUI.RESEARCH.REC.1398.128. However, this thesis was not funded.

Data availability

Data sharing is not applicable to this article as it contains personally identifiable information, which could compromise individual privacy. Data supporting the findings is reported in the article.

Declarations

Ethics approval and consent to participate

All participants received enough information about the study and also provided informed consent to participate in our study. The protocol of the study was ethically approved by the Isfahan University of Medical Sciences (Research project No: 340009, ethical approval No” IR.MUI.RESEARCH.REC.1400.053).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.