Autism spectrum disorder (ASD) is a neurodevelopmental disorder associated with difficulties in social interaction and communication, repetitive behaviour and restrictive interests(1). Studies investigating nutritional aspects in ASD have found a high prevalence of feeding difficulties, which are associated with different factors, such as feeding behaviour problems and sensory processing alterations(Reference Leader, Tuohy and Chen2–Reference Alibrandi, Zirilli and Loschiavo5). These factors result in selective or excessive food consumption, impacting daily dietary repertoire and energy and nutrient intake(Reference Castro, Faccioli and Baronio6,Reference Nadeau, Richard and Wallace7) .
In studies assessing food consumption, estimating usual intake (i.e. intake over a long period) of energy and nutrients is essential for investigating associations between diet and health. However, the method of data collection and day-to-day variability in consumption must be considered(Reference Willett8). A recent systematic review highlighted the need for more detailed information on the methods used to assess dietary intake in individuals with ASD, as well as the inclusion of adjustments for estimating usual intake in their analyses(Reference de Souza Silva, Castro and Valle9).
Understanding the variation in food consumption from day-to-day for each individual (within-person variability) and between individuals (between-person variability) is required to obtain the necessary adjustments in analyses aimed at identifying the prevalence of inadequacies in intake(Reference Willett8,Reference Nel, Steyn and Senekal10,Reference Verly Junior, Cesar and Fisberg11) . Studies have shown that the variability of energy, macronutrient and micronutrient intake varies according to age, sex and country(Reference Caswell, Talegawkar and Siamusantu12–Reference de Castro, Verly and Fisberg14).
FFQ, 24-hour recalls (24HR) and food records (FR) are the most used dietary instruments for assessing usual intake. All these instruments have strengths and biases discussed in the literature(Reference Willett8). Short-term instruments, such as the 24HR and FR, when applied correctly, have a lower systematic error. However, they do not capture within-person variation and do not allow the usual intake to be estimated. Also, although this limitation may be mitigated by repeating the application of the instrument over several days, the number of repetitions required differs according to within and between-person variances, which are inherent to the nutrients of interest in the study population(Reference Willett8,Reference Nel, Steyn and Senekal10,Reference Padilha, França and da Conceição15) .
The number of studies investigating intake variability in typically developing children and adolescents is relatively limited, and until now, no study has reported these findings for individuals with ASD(Reference Caswell, Talegawkar and Siamusantu12,Reference de Castro, Verly and Fisberg14–Reference Ollberding, Couch and Woo16) . The particularities of the diagnostic and difficulties associated with eating in ASD directly affect food acceptance and refusal and, consequently, may contribute to lower dietary variability in these individuals(Reference Chistol, Bandini and Must4,Reference Bandini, Curtin and Phillips17,Reference Sharp, Postorino and McCracken18) . The absence of data on dietary intake variability in this population represents a significant gap in the design of studies involving the collection and interpretation of dietary data. Additionally, factors including the heterogeneity observed among individuals with ASD, the higher prevalence of boys in the samples and the different age groups are rarely considered in analyses of nutritional studies. As a result, the impact of these characteristics on the variation in food consumption remains unclear(Reference Bicer and Alsaffar19,Reference Tsujiguchi, Miyagi and Nguyen20) .
In this context, the main objective of this study was to estimate the within-person and between-person variability and the number of days of 24HR needed to estimate the usual intake of energy and nutrients in a sample of children and adolescents with ASD from southern Brazil. To our knowledge, this study constitutes the first investigation on nutrient variability within a population of children and adolescents with autism. Examining these aspects of dietary intake offers novel insights into our understanding of dietary aspects in this group, enriching the existing body of knowledge on nutrient variability and dietary assessment methods.
Methods
Design and study sample
A cross-sectional study was conducted with baseline data collected between July 2021 and April 2024, from a larger study with patients diagnosed with ASD from 2 to under 19 years, who were assisted at the public neuropediatric health service of a public university in the city of Pelotas, Southern Brazil. Two paediatric neurology professors, responsible for periodic care at this service, conducted the diagnosis of ASD. Out of 311 participants, twenty-four were excluded from the analyses for using dietary supplements, and another three were excluded due to missing dietary data, resulting in a final sample of 284.
Data collection and study variables
Data were collected in three interviews carried out by research nutritionists trained for this study. A standard questionnaire including socio-demographic, educational, clinical and anthropometric questions was applied during previously scheduled face-to-face interviews.
The caregivers were asked to answer three non-consecutive days of the 24HR, with a 1-week interval between each collection, regarding the dietary intake of the subject under assessment. The first and third 24HR were administered face-to-face and covered weekdays. The second recall was collected by telephone interview and corresponded to a weekend day. Caregivers were instructed to report all foods and beverages consumed the day before the interviews, including meal and snack times and quantities consumed.
To improve data quality, the multi-pass method was used during the application of the 24HR(Reference Johnson, Driscoll and Goran21). In addition, a photo book of commonly used measurements was used to support the reporting of portions(Reference Vitolo22). Regarding school meals, caregivers provided information about the food sent from home for consumption at school, as well as any leftovers. A total of 232 caregivers completed 3 days of 24HR at the end of the study, while thirty-six completed 2 days and sixteen completed 1 day. The energy, macronutrient, micronutrient and fibre composition of the reported foods was calculated using the Brazilian Table of Food Composition (TBCA, version 7.2, Universidade de São Paulo, Food Research Center, 2023)(23).
The anthropometric data of weight (kg) and height (cm) were measured using a digital scale with a capacity of 150 kg and accuracy of 100 g (TRENTIN, RS, Brazil) and a stadiometer coupled to the scale (maximum measurement of 213 cm and accuracy of 0·1 cm). For children with heights under 100 cm, a horizontal stadiometer was used (maximum measurement of 100 cm and accuracy of 0·1 cm). These data were used to evaluate nutritional status using the Z-score of body mass index for age, and overweight was assessed according to the cutoff of +1 Z-score, established by the World Health Organization(24,Reference de Onis, Onyango and Borghi25) .
The following variables were used to describe the sample: age categorised by years (2–5, 6–9, 10–18); sex (female, male); skin colour (white, black, brown or indigenous); per capita family income (US$) (<1/2, 1/2–< 1, ≥ 1 minimum wage), considering the Brazilian minimum wage according to the years of data collection; primary caregiver (mother, others); primary caregiver’s years of schooling (≤ 8, ≥ 9); use of antipsychotic medication (yes, no) and overweight (yes, no).
Data analysis
The data were double entered into the EpiData® version 3.1 programme (EpiData, version 3.1; The EpiData Association, Odense, Denmark, 2003–2005). All data were imported to STATA® statistical software version 15.1 (Stata Statistical Software, version 15.1) for analysis.
The presence of outliers in energy and nutrient estimates was investigated by graphical visualisation of the distribution and review of extreme values for each dietary variable. The identified inconsistencies, primarily related to errors in household measurements, were addressed and corrected without excluding any participants.
The standardised within-person (S2w) and between-person (S2b) variances, variance ratio (VR = S2w /S2b), the distribution of the measured intake and the usual intake of energy and nutrients were obtained through analyses conducted in the Multiple Source Method ® version 1.0.1 (MSM) web application (Department of Epidemiology at the German Institute of Human Nutrition Potsdam-Rehbrücke) for each dietary variable investigated(Reference Harttig, Haubrock and Knüppel26). MSM is a statistical method developed to remove within-person random error in estimates of usual consumption distribution from data collected by short-term dietary assessment instruments. Details of the method are available in the publication by Haubrock et al. (Reference Haubrock, Nöthlings and Volatier27).
The equation proposed by Black et al.(Reference Black, Cole and Wiles28) was used to estimate the number of days needed to estimate usual intake: d = (r 2/1 – r 2) × (VR) (equation 1):
where d is the number of days, r is the correlation coefficient required between measured intake and usual intake and VR corresponds to the ratio between S2w and S2b. Correlation coefficients of 0·7, 0·8 and 0·9 were considered acknowledging that the higher the r, the greater the correlation between the information collected and usual intake(Reference Padilha, França and da Conceição15,Reference Verly Junior, Fisberg and Cesar29) .
The analyses were stratified according to age group and sex.
Ethical considerations
This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the Research Ethics Committee of the Faculty of Medicine of the Federal University of Pelotas (CAEE: 94253518·0·0000·5317). Written informed consent was obtained from all those responsible for the participants.
Results
The mean age of the participants was 7·4 (sd 3·5) years, and 75·7 % were children between 2 and 9 years old. Most of the participants were male (82·8 %), white (77·8 %) and from families with a per capita income of under US$247·00 (88·4 %). Almost half were using antipsychotic drugs (54·2 %) and 66·7 % were overweight. Approximately 73 % of caregivers had 9 years or more of schooling and, for the greater part of the participants, mothers were the main caregivers (91·9 %) (Table 1).
Table 1. Socio-economic and demographic characteristics of children and adolescents with autism spectrum disorder. Pelotas, Brazil, PANA study (n 284)
* Other caregivers include fathers, grandparents, foster parents and other family members.
† Missing data for three participants.
‡ Assessed based on BMI for age. Missing data for twenty participants.
Distribution of energy and nutrient intake
Table 2 shows the distribution of energy and nutrient intakes according to the crude and usual estimates. The mean intakes were similar in both estimates for all dietary variables, with a reduction in the standard deviation and differences in the distribution of intakes. In the analyses stratified by age group and sex, higher mean intakes were observed in the groups of boys and the older age group (10–18 years) for most of the nutrients (online Supplementary material 1 and 2).
Table 2. Distribution of crude and usual estimates of energy, macronutrient and micronutrient intake in children and adolescents with autism spectrum disorder. Pelotas, Brazil, PANA study (n 284)
Values were estimated using the Multiple Source Method® programme. P = percentile.
Variance components and days required for calculating usual intakes
In general, S2w was higher than between-person variance (S2b) for most of the dietary variables, resulting in VR values of > 1. Exceptions were observed for fibre, retinol, riboflavin, Ca, Fe, phosphorus and Cu (VR < 1). The highest VR values were observed for cholesterol (1·64), MUFA (1·59) and SFA (1·56), while the lowest values were obtained for Ca (0·52) and riboflavin (0·79) (Figure 1). The number of days of 24HR required to estimate usual intake, considering the correlation coefficient of 0·8, ranged from two to three for most nutrients, except Ca and riboflavin, for which a single day may be sufficient for obtaining an estimate of usual intake. For the 0·7 and 0·9 correlation coefficients, the range varied from 1 to 2 days, and 2 to 7 days, respectively (Table 3).
Figure 1. Variance ratio of energy, macronutrients and micronutrients in children and adolescents with autism spectrum disorder. Pelotas, Brazil, PANA Study (n 284).
Table 3. Variances of within-person and between-person variability and the number of days of 24-hour dietary recall needed to estimate usual energy, macronutrient and micronutrient intake in children and adolescents with autism spectrum disorder. Pelotas, Brazil, PANA study (n 284)
Variability values were estimated using the Multiple Source Method® programme. Results for the number of days were obtained using the formula proposed by Black et al.(Reference Black, Cole and Wiles28), considering the correlation coefficients of 0·7, 0·8 and 0·9. r = correlation coefficient. S2w = within-person variance. S2b = between-person variance. VR = S2w/S2b. VR, variance ratio.
Analyses conducted with male participants showed similar results to the total sample, with VR ≥ 1 for most of the variables. For girls, the VR was predominantly < 1. There were differences between sexes in the values of S2w, S2b and VR; however, no pattern was observed directed at a specific sex. The group of girls, compared with the boys, had lower VR values for most of the dietary variables. Higher VR values were obtained for SFA (1·86), MUFA (1·72) and potassium (1·71) for boys, and for niacin (2·27), cholesterol (2·07) and Zn (1·29) in the group of girls (online Supplementary material 3). Based on the correlation coefficient of 0·8, most of the variables required more days of 24HR in the group of boys (1–3 days) compared with the group of girls (1–4 days) (Table 4).
Table 4. Variances of intra- and inter-individual variability, variance ratios and number of days needed for 24-hour recall to estimate usual energy, macronutrient and micronutrient intake according to sex of children and adolescents with autism spectrum disorder. Pelotas, Brazil, PANA study (n 284)
Variability values were estimated using the Multiple Source Method® programme. Results for the number of days were obtained using the formula proposed by Black et al.(Reference Black, Cole and Wiles28), considering the correlation coefficients of 0·7, 0·8 and 0·9. r = correlation coefficient. S2w = within-person variance. S2b = between-person variance. VR = S2w/S2b. VR, variance ratio.
There were differences in S2w and S2b values between age groups, with higher VR values for most of the variables in 10–18 age group (0·93–3·69), except for lipids, SFA, MUFA, thiamine, niacin, phosphorus, Na and potassium, whose VR values were higher in the 6–9 age group (0·82–3·61). For the group between 2 and 5 years old, the VR ranged from 0·38 to 1·34, and the S2b values were higher than the S2w values for 15 of the 23 variables analysed (VR < 1), with the exception of proteins, MUFA, PUFA, cholesterol, niacin, vitamin C, Na and Zn. In the other age groups, VR > 1 was obtained for almost all the nutrients; however, the opposite was observed for retinol, vitamin C, Ca and Fe for children between 6 and 9 years old, and only for Ca in the older age group (Table 5) (online Supplementary material 4).
Table 5. Variances of within-person and between-person variability of energy, macronutrient, and micronutrient according to age group of children and adolescents with autism spectrum disorder. Pelotas, Brazil, PANA study (n 284)
Variability values were estimated using the Multiple Source Method® programme. S2w = within-person variance. S2b = between-person variance. VR = S2w/S2b. VR, variance ratio.
A correlation coefficient of 0·8 indicates that 1 to 2 days of 24HR were needed to estimate usual intake in children aged 2–5 years, while 1 to 6 days were required for older children (6–9 years). In the group of adolescents (10–18 years), the range of days, using the same coefficient, varied from 2 to 7 days (Table 6).
Table 6. Number of days of 24-hour dietary recall required to estimate the usual energy, macronutrient and micronutrient intake according to the age group of children and adolescents with autism spectrum disorder. Pelotas, Brazil, PANA study (n 284)
Results for the number of days were obtained using the formula proposed by Black et al.(Reference Black, Cole and Wiles28), considering the correlation coefficients of 0·7, 0·8 and 0·9. r = correlation coefficient.
Discussion
The present study had three main findings. Firstly, it was identified that children and adolescents with ASD have higher S2w values for most dietary variables. The second finding indicated differences in S2w and S2b for different nutrients when stratified by sex and age group, with higher S2b values observed in children up to 5 years old and an increase in S2w for older children and adolescents. However, it was not possible to identify the direction of these alterations according to sex. Finally, based on the obtained VR, 2 to 3 days of 24HR are sufficient to obtain food consumption with a good correlation with usual intake (0·8). To our knowledge, this is the first study to examine nutrient variability and the number of days needed to estimate the usual intake of children and adolescents with autism and provide evidence on how sex and age group affect the variance.
Previous studies have identified S2w and S2b variability of energy and nutrients in samples of typically developing children and adolescents in different countries, with observed VR values above and below one(Reference Nel, Steyn and Senekal10,Reference Caswell, Talegawkar and Siamusantu12,Reference Jahns, Carriquiry and Arab13,Reference Padilha, França and da Conceição15,Reference Erkkola, Kyttälä and Takkinen30,Reference Huybrechts, De Bacquer and Cox31) . In this study, S2w was higher than S2b for most nutrients, in analyses not stratified by sex or age. Additionally, the VR values obtained in our sample were lower than those observed in other studies(Reference Caswell, Talegawkar and Siamusantu12,Reference Verly Junior, Fisberg and Cesar29) . Several factors may contribute to these results. We believe that behavioural fluctuations, medication use and variations in the quantities of food consumed, despite a restricted feeding repertoire, could reflect a higher S2w.
The assessment of food consumption during childhood presents challenges due to the use of reporting by caregivers and day-to-day variation, as quantity and variety of foods consumed change throughout different stages of growth and development(Reference de Castro, Verly and Fisberg14,Reference Livingstone, Robson and Wallace32) . In our sample, different S2w and S2b values were obtained according to age group, with higher S2b values in the 2- to 5-year-old group and increased S2w in subsequent age groups. These findings may be explained by the reduced variety of foods consumed by children and the increase of the food repertoire as age progresses(Reference Erkkola, Kyttälä and Takkinen30,Reference Cooke and Wardle33,Reference Lanigan, Wells and Lawson34) . Furthermore, evidence indicates that individuals with ASD often have more limited food repertoires compared with control groups(Reference Chistol, Bandini and Must4,Reference Castro, Faccioli and Baronio6,Reference Bandini, Anderson and Curtin35) . However, studies on feeding difficulties in ASD have not comprehensively investigated the influence of age on food refusal(Reference Chistol, Bandini and Must4,Reference Bandini, Curtin and Phillips17,Reference Curtin, Hubbard and Anderson36) .
The influence of age on the increase in VR values in typically developed children has been previously observed in the literature(Reference Nel, Steyn and Senekal10,Reference Jahns, Carriquiry and Arab13,Reference de Castro, Verly and Fisberg14,Reference Erkkola, Kyttälä and Takkinen30,Reference Huybrechts, De Bacquer and Cox31) . However, discrepant results compared with our study were reported by Ollberding et al. (Reference Ollberding, Couch and Woo16), who found that the highest VR values were observed in the group of children (6–11 years), compared with the group of adolescents (12–17 years). In Brazil, studies have been carried out on different age groups; however, they are also limited to samples of typically developing children and adolescents and from different regions of the country, primarily from the northeast and southeast. Padilha et al.(Reference Padilha, França and da Conceição15) (13–32 months) and Salles-Costa et al.(Reference Salles-Costa, Barroso and Mello37) (6–30 months) found VR < 1 for most of the nutrients analysed, except for PUFA, energy, fibre(Reference Padilha, França and da Conceição15) and vitamin C(Reference Padilha, França and da Conceição15,Reference Salles-Costa, Barroso and Mello37) . Conversely, a study with preschool children (1–6 years) obtained VR > 1 values, including data from participants across all five macro-regions of Brazil (south, southeast, central west, north and northeast)(Reference de Castro, Verly and Fisberg14). However, the inclusion of a few cities per region affects the representativeness of the data.
As for adolescents, studies conducted in São Paulo and Rio de Janeiro found higher S2w values compared with S2b for most nutrients (VR > 1)(Reference Verly Junior, Fisberg and Cesar29,Reference Costa, Takeyama and Voci38,Reference Pereira, Araujo and Lopes39) . Different results were presented in the study by Lima et al.(Reference Lima, Lyra and Sena-Evangelista40), carried out with individuals aged 10–19 years residing in the city of Natal (northeast Brazil), where higher S2b values were obtained (VR < 1).
Our study sample had a higher proportion of boys, consistent with the latest prevalence estimate of ASD from the Centers for Disease Control and Prevention(Reference Maenner, Warren and Williams41). Differences in nutrient intake variability have also been observed in studies with typically developed children and adolescents. Caswell et al. (Reference Caswell, Talegawkar and Siamusantu12) found lower VR in samples of Zambian girls (4–8 years old) compared with boys, corroborating our findings. In contrast, other studies have reported higher VR for females when analysing nutrient intake variability(Reference Erkkola, Kyttälä and Takkinen30,Reference Pereira, Araujo and Lopes39,Reference Lima, Lyra and Sena-Evangelista40) . Other studies reported a lack of consistent findings that could indicate possible discrepancies in within-person and between-person variances among boys and girls(Reference Jahns, Carriquiry and Arab13,Reference Ollberding, Couch and Woo16,Reference Huybrechts, De Bacquer and Cox31) .
The use of within and between-person variance to determine the number of days of 24HR is widely recognised and recommended(Reference Willett8,Reference Black, Cole and Wiles28,Reference Nelson, Black and Morris42) . Results observed in samples of Brazilian children and adolescents presented intervals ranging from 1 to 16 d for younger children (aged 1–6 years)(Reference de Castro, Verly and Fisberg14,Reference Padilha, França and da Conceição15,Reference Erkkola, Kyttälä and Takkinen30) , from 4 to 14 d for older children (6–11 years)(Reference Ollberding, Couch and Woo16) and 2 to 16 d for adolescents (12–18 years)(Reference Ollberding, Couch and Woo16,Reference Pereira, Araujo and Lopes39) , for correlation coefficients ≥ 0·8. Intervals of 1–5 d were obtained considering coefficients of 0·9 for samples of children and adolescents(Reference Salles-Costa, Barroso and Mello37,Reference Lima, Lyra and Sena-Evangelista40) . However, most studies have estimated a greater number of days for coefficients ≥ 0·9(Reference de Castro, Verly and Fisberg14,Reference Padilha, França and da Conceição15,Reference Verly Junior, Fisberg and Cesar29–Reference Huybrechts, De Bacquer and Cox31,Reference Costa, Takeyama and Voci38,Reference Pereira, Araujo and Lopes39) .
In the current study sample, a range of 1–3 d of 24HR was required to estimate the intake of most nutrients to obtain correlations of 0·7 and 0·8 with the usual intake. In addition, it highlighted the necessity to consider the characteristics of the target population in estimating the required number of 24HR days. The increase in the number of days, proportional to higher VR values, observed for older children and adolescents, as well as variations between girls and boys in different nutrients, has been previously reported in studies with children and adolescents without a diagnosis of ASD(Reference Erkkola, Kyttälä and Takkinen30,Reference Huybrechts, De Bacquer and Cox31,Reference Pereira, Araujo and Lopes39) .
Although the application of several days of 24HR increases the precision of food consumption estimates, it results in higher costs and requires greater availability from the interviewees to report the data(Reference Willett8,Reference Lanigan, Wells and Lawson34) . The findings of this study have relevant implications for the design and interpretation of data from research on children and adolescents with ASD. Considering that validation studies of FFQ developed for this age group generally found correlations ≤ 0·7, the results indicate that 1–2 d of 24HR may provide similar estimates to those obtained with FFQ commonly used in epidemiological studies with samples with and without ASD diagnoses(Reference Rockett, Breitenbach and Frazier43–Reference Saravia, Miguel-Berges and Iglesia45).
The presented results provide support for the planning of methodologies for dietary data collection for children and adolescents with ASD. It also provides data for adjustments in analyses of the distribution and adequacy of food consumption in this population. The use of external dietary variability data is supported by previous studies, which indicate the applicability of previously collected values of S2w in order to correct estimates of the distribution of usual intake in studies with only one day of 24HR, on the condition that these data come from a sample with characteristics similar to those of the investigated population(Reference Verly Junior, Cesar and Fisberg11,Reference Jahns, Arab and Carriquiry46) . However, the application of repeated 24HR is still the most recommended method(Reference Luo, Dodd and Arnold47,Reference French, Arsenault and Arnold48) .
Our study has limitations inherent to the dietary method applied, such as the potential for underestimation or overestimation of food portions which can impact dietary intake estimates. To minimise this bias, we used a photo book of household measurements, and all nutritionists were trained to identify potential inconsistencies during the 24HR interviews. Additionally, sample restrictions should be considered, as the results were derived from children and adolescents attending a public health service in a city in Southern Brazil, which restricts the external validity of the study. Therefore, the interpretation and use of current findings require caution, as dietary habits and variability are influenced by multiple socio-demographic factors and may differ between populations(Reference Willett8). Furthermore, there is a need for similar investigations with the adult population with ASD, as evidence suggests the presence of feeding problems in this age group as well(Reference Demartini, Nisticò and Bertino49).
Conclusion
In this study, the within-personal variance was predominantly higher than the between-person variance for most of the nutrients analysed, resulting in estimates of 2 to 3 days of 24HR needed to assess usual intake, considering the correlation coefficient of 0·8. The findings differed according to age and sex, suggesting the importance of including these aspects in the design and interpretation of results related to food consumption in ASD. Future studies may analyse how socio-economic characteristics, behavioural and clinical factors may influence food consumption and variability of nutrient intake in individuals with ASD.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114525000200
Acknowledgements
The authors thank the families who took part in the study, staff and personnel for their assistance in the recruitment and schedule the appointments. The authors acknowledge the essential work and dedication of all research assistants during the data collection of this study.
Financial support
This study was financed in part by the Coordination for the Improvement of Higher Education Personnel (CAPES) - Finance Code 001, and the Brazilian National Research Council (CNPq) - Finance Code 308213/2021–1 and 407237/2021–6.
Competing interests
The authors declare that there are no conflicts of interest.
Authorship
The author’s responsibilities were as follows: E. S. participated in the designing of the study, statistical analysis, interpreting of data and writing the manuscript. E. V. J. supported the plan of statistical analysis and interpretation of data. L. H., K. C., S. V. and J. S. V. participated in designing the study and interpretation of data. All authors contributed to the writing or revising of the manuscript and approved the final version.
