Over the past two decades evidence has shown that childhood overweight and obesity and their co-morbidities are increasing worldwide(Reference Cole, Bellizzi and Flegal1–Reference Sinha, Fisch and Teague6). Like many other South and East Mediterranean countries, Tunisia is experiencing an advanced nutrition transition followed by the increase of overweight and obesity prevalence in adolescent boys and girls aged 15–19 years that has reached 17·4 and 20·7 %, respectively(Reference Aounallah-Skhiri, Ben Romdhane and Traissac7, Reference Aounallah-Skhiri, El Ati and Traissac8). Insufficient physical activity (PA) has been identified as the fourth leading risk factor for death and disability associated with obesity and its co-morbidities(9). Moreover, regular PA early in life is associated with major health benefits, including decreased adiposity, lower blood lipids, lower blood pressure and greater bone mineral density(Reference Andersen, Harro and Sardinha10–Reference Strong, Malina and Blimkie12), and could prevent or delay the onset of chronic diseases later in life(Reference Telama, Yang and Laakso13, Reference Twisk, Kemper and van Mechelen14). There is dearth of information related to lifestyle factors especially PA in Tunisia(Reference Aounallah-Skhiri, Traissac and El Ati15).
Thus, prevention wise, the understanding of the different dimensions and patterns of PA among youth in this context states the need for accurate assessment instruments. A variety of PA measurement methods do exist such as PA questionnaires (PAQ), direct observation, mechanical or electronic monitoring, or doubly labelled water(Reference Valanou, Bamia and Trichopoulou16). Each method has its own utilities and limits in terms of accuracy, cost, versatility and practicality in free-living situations for the evaluation of average daily energy expenditure (EE) and dose–response associations with health outcomes based on recommended PA or EE cut-off points(Reference Pate17, Reference Pate, Pratt and Blair18). Although simplified and relatively cheap versions of the heart-rate monitoring (HRM) technique have been devised to be used in epidemiological investigations(Reference Kurpad, Raj and Maruthy19, Reference Wicks, Oldridge and Nielsen20), a specific usefulness of the PAQ method is that it also provides detailed quantitative and qualitative assessment of the entire spectrum of activity and inactivity(Reference Staten, Taren and Howell21–Reference van der Ploeg, Merom and Chau26) allowing a variety of analysis purposes, e.g. habitual PA patterns(Reference Gordon-Larsen, McMurray and Popkin27).
Several questionnaires have been developed for children and adolescents, such as that of the Behavioral Risk Factor Surveillance System (BRFSS)(Reference Yore, Ham and Ainsworth28), the International Physical Activity Questionnaire (IPAQ)(Reference Craig, Marshall and Sjostrom29) or the Arab Teens Lifestyle Study (ATLS) PA questionnaire(Reference Al-Hazzaa, Al-Sobayel and Musaiger30), and can be used for specific purposes such as ranking, grouping or categorizing PA levels, and possibly assessing some aspects of low, moderate and vigorous PA(Reference Kohl, Fulton and Caspersen31). However, it would seem that more comprehensive self-report questionnaires aiming at assessing all forms of activity performed during the day are essential. These questionnaires are specifically important for the assessment of the pattern and mode of PA in preadolescent and adolescent Tunisians in different cultural settings. Also, previous research has indicated that the reliability and validity of PAQ varies according to ethnicity(Reference Moore, Hanes and Barbeau32).
The objectives of the present study were thus to develop and validate a Physical Activity Frequency Questionnaire (PAFQ) specific to Tunisian adolescents.
Methods
Participants
The validation sample size was based on the recommendation that 100–200 subjects should be interviewed(Reference Willett and Lenart33). Participants were 150 volunteers from two elementary schools and two high schools in the Greater Tunis region (the most economically developed and urbanized area including the capital city). Data were collected between March and June 2008, a period when the climate allows a mix of indoor and outdoor activities which cover the whole range of EE.
Ethical approval
The study was conducted according to the guidelines laid down in the Declaration of Helsinki. The protocol was approved by the Ethic Committee on Human Research of the National Institute of Nutrition. After the participants’ parents had been thoroughly informed about the purpose, requirements and procedures, all participants gave free informed verbal consent which was witnessed and formally recorded.
Anthropometry
Anthropometric measurements were taken at the beginning of the study following standard procedures(Reference Lohman, Roche and Martorell34); all volunteers wore light clothing but no shoes or socks. Height was measured to the nearest 0·1 cm with a wall-mounted stadiometer (Person-check®; Kirchner & Wilhelm, Germany). Weight was measured to the nearest 0·1 kg on a calibrated scale (Detecto, USA).
Development of the physical activity frequency questionnaire
The PAFQ was designed and developed in several steps. As insufficient data were available concerning the PA profile of adolescents in Tunisia, a new inventory was made using a series of 24 h PA recalls (24h-R)(Reference Cale35) carried out on a sample of forty volunteer Tunisian adolescents (10–19 years old), diversified by area (urban/rural) and gender to cover a range of different situations. Each participant answered the 24h-R three times, i.e. on Sunday (rest day), Monday (full day of school) and Saturday (half day of school). Each day was divided into 5-min blocks and the respondent was prompted to report the type of the activity performed during each block as the day progressed using time cues. Information was also collected about specific characteristics of each activity while it was being performed, e.g. the body position (sitting, standing, lying down or in locomotion) and intensity (carrying a load, or not).
The energy cost of each activity was calculated by multiplying the time spent by the corresponding metabolic equivalent of task (MET) value, taken from the compendium of physical activities for youth(Reference Ridley, Ainsworth and Olds36) modelled on the previous adult compendium of physical activities(Reference Ainsworth, Haskell and Whitt37). For each participant, the total daily energy expenditure (TEE) was estimated as the sum of all EE over the three 24 h periods divided by three, and the mean overall TEE was defined as the average of TEE over the forty participants. The relative contribution of each activity to TEE was computed, and activities that contributed up to 95 % of mean overall TEE were included in the questionnaire. Activities that were not significant with respect to the whole sample but that contributed ≥10 % to one individual's TEE were also included(Reference Bernstein, Sloutskis and Kumanyika38).
The final list consisted of fifty-four different physical activities classified in eight sections: (i) at home (personal care, nursing, household chores, cooking); (ii) meals; (iii) school time; (iv) transport; (v) non-sport leisure activities; (vi) sports; (vii) prayer; and (viii) sleeping. A 7 d interview questionnaire, which took both school and rest days into account, was then designed in Arabic for use among the adolescents. For each activity, the frequency and length of time spent were detailed. Weekly frequency was scored from 0 (never) to 7 (every day). Possible durations ranged from 0 to 8 h/d with a precision of 15 min. A pilot test was conducted among ten adolescents (10–19 years old) to assess whether the PAFQ was easy for them to read and understand. Minor changes were made including the use of age-appropriate language and a more appealing layout for young people. The final version of the questionnaire was designed to be interviewer-administered and answering all the questions took about 30 min (see online supplementary material for a version translated from French/Arabic to English).
After data collection and acquisition, for each activity, frequency and duration were converted into minutes per day. If the amount of time declared was less than 24 × 60 = 1440 min/d, TEE was completed by the resting energy expenditure estimated with the Schofield-HW equations(Reference Schofield39) using sex, height and weight, as they are recommended for mixed populations of obese and non-obese children and adolescents(Reference Rodriguez, Moreno and Sarria40); if it was more than 1440 min/d, the energy cost of each activity was proportionately adjusted so that the corrected time spent added up to one day. Then EE for each activity was then derived using the compendium of physical activities as described above.
For data analysis and presentation purposes, the fifty-four activities from the eight categories were arranged in twenty-five groups and twenty-four groups (with the exclusion of sleep) were classified according to MET values into low- (MET < 3), moderate- (MET 3–6) and high-intensity activities (MET ≥ 6)(41, Reference Pate, Blair and Eddy42).
Validation study
To validate the PAFQ, PA estimations based on the questionnaire were compared with those derived from both an objective instrument, HRM, and a subjective one, 24h-R, on the same individuals.
Heart-rate monitoring
The HRM technique is based on the linear relationship between heart rate (HR) and oxygen consumption(Reference Pate43, Reference Spurr, Prentice and Murgatroyd44) and has been shown to be a valid indicator of EE and activity levels in youth(Reference Kohl, Fulton and Caspersen31). HR was recorded in free-living conditions over a continuous 3 d period (including one weekday and two weekend days) by an electrode-belt worn around the chest that detected and stored the mean HR at 5-min intervals (S610i; Polar Instruments, Kempele, Finland). Data were downloaded to a computer for storage using a Polar propriety interface and software. All non-physiological values (>215 or <45 beats/min) were replaced by the average of the adjacent value, and individual files were discarded if the total aberrant values exceeded 3 % of the file(Reference Welk and Corbin45). Before wearing a HR monitor, each participant's maximum oxygen uptake and maximum heart rate (HRmax) were assessed using the Léger test(Reference Leger and Lambert46). The value used as the resting heart rate (HRrest) was the average HR during sleep, and the heart rate reserve (HRR) was calculated as the difference between HRmax and HRrest. The Karvonen formula(Reference Karvonen, Kentala and Mustala47) was then used to calculate the daily physical activity intensity (PAI) expressed as a percentage of HRR and based on both maximum and resting heart: PAI (%) = [(HRdaily – HRrest)/HRR] × 100. To estimate TEE, we assumed the percentage of HRR and VO2 reserve were equivalent, and used the energy equivalent to convert the rate of oxygen consumption into an EE rate(Reference Schutz, Weinsier and Hunter48). Since the EE estimate derived from the PAFQ did not include the thermic effect of food (TEF), it was assumed that 10 % of the TEE was due to TEF and this was subtracted from the TEE assessed by HRM. Moreover, contrary to what happens in younger children, in the age group under study energy for growth is weak (i.e. between 0 and 1·5 % of TEE)(Reference Anderson, Dewey and Frongillo49), and it was decided not to take this into account. The different activity intensities were classified based on percentage usage of the individual's personal HRR: low (<40 % HRR), moderate (40–70 % HRR) and high intensity (>70 % HRR)(Reference Karvonen, Kentala and Mustala47).
24 h Physical activity recall
A 24h-R protocol(Reference Tudor-Locke, Ainsworth and Adair50) was used simultaneously to validate the PAFQ. The adolescents were asked to note down their activities while wearing the HR monitor. The 24h-R was prompted by trained interviewers asking about segments of the day starting when the HR monitor was first worn and ending with its retrieval. The adolescents were encouraged to account for the entire time they wore the monitor.
Reliability study
The reliability of the PAFQ was evaluated using a test–retest design(Reference Kohl, Fulton and Caspersen31). The PAFQ was interviewer-administered once on the day following the end of the HRM sequence and once one month later.
Data processing and analysis
Data entry screens including quality checks as well as validation by double entry used Epi-Data software version 3·1. Data management and analysis were performed using the statistical software packages Stata version 11 and SPSS 15·0 for Windows. Results are presented as estimates and their standard deviations or 95 % confidence intervals.
Validation
We assessed relationships between the PAFQ and HRM or 24h-R by scatter diagrams and computing Pearson's correlation coefficients. Agreement between the PAFQ and HRM or 24h-R was assessed using a plot of the differences between the two methods v. the average of the measurements according to the Bland–Altman technique(Reference Bland and Altman51); the magnitude and direction of the bias of the PAFQ was assessed as the mean of individual differences and precision as the standard deviation of those differences. In addition, Spearman correlation coefficients were used to examine the associations between self-reported time spent in various activity intensities and data from both HRM and 24h-R.
Reliability
Agreement between the two replicates of PAFQ from the test–retest procedures was assessed by within-subject intra-class correlation coefficients (ICC) derived using one-factor ANOVA and using weighted kappa (κ w) statistics (for interpretation, κ w ≤ 0·09 indicated poor agreement, κ w = 0·10–0·20 slight agreement, κ w = 0·21–0·40 fair agreement, κ w = 0·41–0·60 moderate agreement, κ w = 0·61–0·80 substantial agreement, and κ w = 0·81–1·00 indicated almost perfect agreement(Reference Landis and Koch52)). The test–retest reliability was assessed for TEE and also for EE in each of the eight categories of PA in the questionnaire.
Results
Participants
Among the 150 originally recruited participants, eight did not complete all the stages of the study, so that the validation was finally based on a sample of 142 volunteers (seventy-one girls, seventy-one boys) aged 10–19 years (mean 14·5 (sd 3·0) years); most of them were urban (93 %), 43 % attended elementary school and 57 % secondary school.
Validation
Scatter diagrams (Fig. 1) and correlation coefficients indicated a strong linear relationship between the participants’ TEE estimated by the PAFQ and both HRM (r = 0·70; 95 % CI 0·62, 0·76) and 24h-R (r = 0·81; 95 % CI 0·77, 0·84), despite some outliers. The paired data points were somewhat away from the diagonal line, indicating a not perfect individual agreement between TEE assessed with the PAFQ and 24h-R, or especially HRM. Indeed, mean overall TEE (Table 1) estimated by the questionnaire tended to be higher (+11·3 %) than by HRM, but lower (−2·8 %) than by 24h-R. Also, the Bland–Altman plot (Fig. 2(a)), which assessed the agreement between participants’ TEE estimated by PAFQ v. HRM, showed some amount of bias: the mean of individual PAFQ – HRM differences was +1106·2 kJ/d (95 % CI 845·8, 1366·6 kJ/d). On the other hand, the bias for TEE estimated by PAFQ v. 24h-R as represented in the corresponding Bland–Altman plot (Fig. 2(b)) was reversed and of smaller absolute value: mean of individual PAFQ – 24h-R differences was −272·7 kJ/d (95 % CI −490·6, −57·4 kJ/d). On both plots, there was no obvious pattern of differences which were scattered around the bias; precision (as assessed by upper and lower limits of agreement on the Bland–Altman plots) of estimation of TEE by PAFQ was better for 24h-R (−2900·2 and +2353·0 kJ/d) than for HRM (−2033·1 and +4245·4 kJ/d). The Spearman correlation coefficients between the individuals’ time spent in low-, moderate- and high-intensity activities estimated by PAFQ and HRM were significant and moderate (r = 0·55–0·72) except during light activity (r = 0·29), while the PAFQ and the 24h-R showed significant and moderate correlations (r = 0·67–0·75) across all ranges of intensity (Table 1).
Fig. 1 Validation of the Physical Activity Frequency Questionnaire (PAFQ): scatter diagrams of daily total energy expenditure estimated by the PAFQ v. (a) heart-rate monitoring (HRM) or (b) 24 h physical activity recall (24h-R) among Tunisian preadolescent and adolescent children aged 10–19 years (n 142), March–June 2008. Solid line represents perfect agreement (line of equality, y = x). Pearson correlation coefficient, r (95 % CI): 0·70 (0·62, 0·76) for PAFQ v. HRM; 0·81 (0·77, 0·84) for PAFQ v. 24h-R
Table 1 Daily total energy expenditure (TEE) estimated by the Physical Activity Frequency Questionnaire (PAFQ), heart-rate monitoring (HRM) and 24 h physical activity recall (24h-R) methods, and Spearman correlations between the time spent in different activity levels estimated by the PAFQ and each reference method, among Tunisian preadolescent and adolescent children aged 10–19 years (n 142), March–June 2008
Fig. 2 Validation of the Physical Activity Frequency Questionnaire (PAFQ): Bland–Altman plots of the differences in daily total energy expenditure estimated by the PAFQ and (a) heart-rate monitoring (HRM) or (b) the 24 h physical activity recall (24h-R) v. the average of the two methods among Tunisian preadolescent and adolescent children aged 10–19 years (n 142), March–June 2008. The horizontal solid line represents the mean of the differences (bias) and the dashed lines the mean of the differences ±2 sd of the differences (lower and upper limits of agreement)
Reliability
Overall mean TEE over the 142 participants for the two administrations of the PAFQ was 9797·1 (sd 2112·9) kJ/d and 9538·1 (sd 1890·9) kJ/d; the mean of individual differences was 259·0 kJ/d (95 % CI 121·3, 396·8 kJ/d), showing no systematic departure from the null hypothesis (paired t test P = 0·152). Within-subject ICC for the two replicates of the PAFQ was 0·86 (95 % CI 0·71, 0·99) for TEE and ranged between 0·79 and 0·93 for EE by PA category. The κ w value for TEE over test–retest was 0·61 (95 % CI 0·50, 0·71) and κ w coefficients by category of PA ranged from 0·47 to 0·78 (Table 2).
Table 2 Test–retest reliability of the Physical Activity Frequency Questionnaire (PAFQ): weighted kappa coefficients (κ w) and within-subject intra-class correlation coefficients (ICC) of estimated energy expenditure according to the eight physical activity categories among Tunisian preadolescent and adolescent children aged 10–19 years (n 142), March–June 2008
Energy expenditure by activities assessed by the physical activity frequency questionnaire
Participants showed low daily EE in high-intensity (16·8 min/d, i.e. 1·9 % of waking time) and moderate-intensity (101·1 min/d, i.e. 11·5 % of waking time) PA like sports, home activities or walking (Table 3); the EE spent sitting, standing, studying, using television/computer, playing board games, eating and using motorized transport (760·2 min/d, i.e. 86·6 % of waking time) was higher. The contribution to the estimated daily TEE was 21 % for sleeping, 6 % for high-, 15 % for moderate- and 58 % for low-intensity activities.
Table 3 Time spent and energy expenditure estimated by the Physical Activity Frequency Questionnaire for the twenty-five groups of activities among Tunisian preadolescent and adolescent children aged 10–19 years (n 142), March–June 2008
MET, metabolic equivalents of task.
*In each group activities are sorted by decreasing mean daily energy expenditure.
Discussion
In a context of increasing prevalence of overweight and obesity, understanding PA behaviours is of major importance prevention wise. We developed a PAFQ, which lists fifty-four different activities and then classified into twenty-five groups and eight categories, to document in depth the various dimensions and patterns of PA in free-living Tunisian preadolescents and adolescents. We estimated its validity both v. one objective instrument (HRM with individual calibration) and one subjective instrument (a 24h-R of daily activities) and we assessed its reliability by a test–retest procedure at a 1-month interval.
Development of the physical activity frequency questionnaire
Development of the questionnaire including a pilot study and use of age-appropriate language resulted in a final version of the questionnaire which was understandable among the target population of preadolescents and adolescents. Nevertheless, given the complexity, especially for young adolescents, to estimate with precision the duration and intensity of their activities(Reference Montoye, Kemper and Saris53), and particularly in the Tunisian context, the PAFQ was designed as an interviewer-administered rather than a self-report questionnaire. Answering all the questions took about 30 min which seemed acceptable for the participants and compatible with use of the PAFQ in large-scale studies.
Validity
As for the validity, the PAFQ was strongly correlated with both HRM and 24h-R and featured acceptable agreement with both criterion measures, with a better result v. 24h-R than v. HRM. Indeed, validity analysis using Bland–Altman plots revealed an acceptable moderate amount of bias (although larger for HRM than 24h-R), despite the fact that a few outlying data points were included in the analysis while their exclusion would have improved our results (detailed data not shown). The HRM method is well recognized as not being a good predictor of EE at low compared with higher levels of PA(Reference Livingstone54) and this could explain the lower validity of the PAFQ v. the HRM method in low-activity individuals. Because of methodological issues (including inappropriate use of correlation coefficients for quantification of validity in many studies(Reference Bland and Altman51)) it is not always possible to make direct comparisons with the findings of other studies, but the characteristics of the PAFQ v. HRM or 24h-R appeared to be comparable with other instruments. For example, in a sample of forty-four middle- and high-school students in South Carolina, Weston et al.(Reference Weston, Petosa and Pate55) reported correlation coefficients of 0·88, 0·77 and 0·53 between relative EE from a previous day recall of PA and a pedometer, Caltrac counts and percentage HR range, respectively. In addition, several validation studies, reviewed by Kohl et al.(Reference Kohl, Fulton and Caspersen31), which examined among children and adolescents self- or interviewer-administered recall of PA using HRM as criterion measure, revealed correlation coefficients ranging from 0·25 to 0·58.
Reliability
Regarding reliability during the test–retest procedure the ICC was 0·86 and not less than 0·79 for EE by PA categories, indicating strong agreement between the two assessments; also values of κ w indicated moderate to substantial test–retest reliability according to the components of the estimated TEE. Variability of PA habits or technical errors may negatively affect reliability although it is not possible to distinguish between both. The quite satisfactory level of reliability could have been due to both the comprehensiveness of various types of PA in the questionnaire and the good understanding of questions due to careful testing at the development stage. Furthermore, test–retest reliability is known to improve with time when self-report techniques are used(Reference Bernstein, Sloutskis and Kumanyika38, Reference Baranowski, Dworkin and Cieslik56, Reference Sallis, Buono and Roby57). Also as the PAFQ was an interviewer-administered questionnaire, a trained interviewer may also have an important role in the strength of the reliability results.
Methodological strengths and limitations of the study
A validation sample composed of volunteers who were mostly urban and attending schools from around the capital city could be a limit of the present study, especially regarding large-scale use of the PAFQ. But levels of TEE among the surveyed participants indeed showed a wide range of values and the mean overall level of PA was coherent with results observed in the same context on a much larger sample (although with a questionnaire adapted from one developed for adults)(Reference Aounallah-Skhiri, El Ati and Traissac8). Also in the context of socio-economic changes and socio-cultural evolution there seems to be a trend towards convergence of lifestyles (including PA behaviours with more time spent on sedentary activities) whatever the area of residence(Reference Popkin, Adair and Ng58). Although the questionnaire was developed in Tunisia, it likely has broader applicability to other North African countries; indeed, these countries share with Tunisia both a number of socio-cultural features linked to the Arab–Muslim culture and the common consequences on lifestyle factors such as diet and PA of socio-economic and societal changes due to globalization and modernization.
There are a number of advantages in using the present PAFQ in comparison with many other PAQ designs related to both its detailed content and the proposed mode of administration. First, its interviewer-administered format ensured the quality of the data collected. Second, PAFQ produces a high level of detail for PA behaviours across different domains of free-living teenagers, which may be useful in identifying common activities that could be appropriate targets for behavioural interventions to improve PA at a large scale. Third, the assignment of different intensity categories and body postures for each activity can reduce the magnitude of errors in intensity level reported. Fourth, a distinction between school and rest days over the habitual week such as developed in our PAFQ on one hand, and the closed questions and specific time slots on the other hand, may be able to minimize duration reporting errors. Indeed, reducing errors of the two basic elements of dose (intensity and time spent) may be an important approach to reduce overall measurement errors in self-report data(Reference Matthews, Moore and George59).
Validation of the PAFQ v. both a subjective instrument (the 24h-R) and an objective one (the HRM method) was also a strength of the study. Nevertheless, a limitation of assessing PA based on HR data is the confounding effect of factors other than energy demand on the HR response to exercise; for example, high ambient temperature, time of day, emotional state, dehydration, food and caffeine intake, smoking, body position, muscle groups used and illness(Reference Livingstone54, Reference Maas, Kok and Westra60, Reference Maffeis, Pinelli and Zaffanello61). However, when applied to groups of individuals, the HRM technique provides an acceptable estimate of TEE and associated patterns of PA(Reference Spurr, Prentice and Murgatroyd44). Objective measures of PA fitness with multiple motion sensors or HRM have often been used as a criterion measure in field studies validating other subjective PA instruments, particularly self-report instruments(Reference Valanou, Bamia and Trichopoulou16). Some recent questionnaire validation studies have used an accelerometer as a criterion(Reference Corder, Ekelund and Steele62–Reference Telford, Salmon and Jolley64) and some others used combined HRM with motion sensors, e.g. pedometers or accelerometers(Reference Bassett, Ainsworth and Swartz65, Reference Sobngwi, Mbanya and Unwin66); indeed, it is assumed that combining HR data with motion sensors may provide more valid representation of PA, but several studies have investigated the validity of self-report questionnaires by using the same type of methods as used in the present study(Reference Valanou, Bamia and Trichopoulou16, Reference Ridley, Ainsworth and Olds36, Reference Sallis, Buono and Roby57, Reference Booth, Okely and Chey67, Reference Florindo, Romero and Peres68). Furthermore, the percentage HRR is assumed to give a reasonably accurate estimation of the TEE estimated by the gold standard methods of doubly labelled water(Reference Spurr, Prentice and Murgatroyd44, Reference Van den Berg-Emons, Saris and Westerterp69) and from oxygen uptake measured in a laboratory or from Caltrac(Reference Ballor, Burke and Knudson70). It is nevertheless recognized that the doubly labelled water technique is the most accurate for the estimation of daily EE and hence as the reference method to check the validity of questionnaires designed to evaluate the PA level(Reference Montoye, Kemper and Saris53). But, its application is constrained by cost and technical complexity.
Conclusion
In Tunisia, in the context of the nutrition transition and the increasing burden of overweight and obesity resulting from specificities of that critical period of life, adolescents are especially prone to the related changes in diet and PA. In the present study we have developed a comprehensive and detailed PAFQ specific to Tunisian preadolescents and adolescents. This questionnaire was validated v. two reference methods and shown to be reproducible. The developed frequency questionnaire could be useful for large-scale description, analysis and surveillance of patterns of PA or evaluation of interventions to promote PA among Tunisian adolescents(Reference Harrabi, Maatoug and Gaha71, Reference Holdsworth, El Ati and Bour72).
Acknowledgements
Source of funding: This study was financed by the CORUS programme of the French Ministry of Overseas and European Affairs, as part of the ‘Obe-Maghreb’ research project (Contract Corus 6028-2); the Institut de Recherche pour le Développement (IRD), France; and the National Institute of Nutrition and Food Technology (INNTA), Tunisia. CORUS programme, IRD and INNTA had no role in the design, analysis or writing of this article. Conflicts of interest: The authors declare they have non-financial competing interests. Authors’ contributions: All authors participated in the study conception; collection, analysis and interpretation of the data; and writing of the manuscript.
