Futsal is a team sport involving a complex range of high-intensity locomotor activities, requiring both aerobic and anaerobic fitness to cope with the multiple requirements of the match (Ribeiro et al., 2020). Research studies investigating this sport modality have increased significantly over the past two decades (Barbero-Alvarez et al., 2008; Castagna et al., 2009; De Oliveira Bueno et al., 2014; Caetano et al., 2015; Nakamura et al., 2020; Ribeiro et al., 2020), contributing to better understanding of the physical and skills requirements during the futsal match and organization of the training contents. Time motion analysis studies reported that during a single match, futsal players cover, on average, a total distance of 3000–4000 m, of which 8–13 and 5–9% are performed at high-speed running (HSR; >15.5 km/h) and sprinting (>18.3 km/h), respectively (Barbero-Alvarez et al., 2008; Castagna et al., 2009; De Oliveira Bueno et al., 2014). Repeated sprint sequences (RSS) with up to 2 to 3 sprints interspersed with 15 s of recovery are also frequently (∼80 occurrences per match) performed in futsal (Caetano et al., 2015). Similarly, the execution of deceleration and acceleration actions constitutes a critical part of the futsal players’ work-rate (Ribeiro et al., 2020). Thus, HSR episodes, sprints, RSS, changes of directions (COD), and acceleration and deceleration actions are among the main types of activities that players should be prepared to perform efficiently. This information may be of practical relevance for the design of suitable training programs in order to enhance the main physical capacities related to successful performance in futsal.

The training intensity is among the first training variables to be manipulated in most physical conditioning programs for athletes (Buchheit and Laursen, 2013). The definition of number and duration of sets performed per training session is also a key component to determine the total training volume (i.e., total work duration performed) (Buchheit and Laursen, 2013). Thus, any manipulation in these variables, in an isolated or combined manner, will be decisive to determine the magnitude of adaptations to training (Buchheit and Laursen, 2013). High-intensity interval training (HIIT) and repeated-sprint/sprint interval training (RST/SIT) are currently among the most frequently used training models in physical conditioning programs of team sports players (Buchheit et al., 2008; Ferrari-Bravo et al., 2008; Buchheit and Laursen, 2013; McGinley and Bishop, 2016). The current evidence comparing RST/SIT vs. HIIT models demonstrates conflicting results, with some studies showing reduced gains (Buchheit et al., 2008) and others superior gains (Ferrari-Bravo et al., 2008) in performance after RST/SIT compared to HIIT models. Other studies have examined the effects of two work-matched HIIT models performed at different training intensities (McGinley and Bishop, 2016; Viaño-Santasmarinas et al., 2018). For instance, Viaño-Santasmarinas et al. (2018) did not find any improvements in aerobic fitness indices and RSA performance after two work-matched, HIIT models, performed at 85 and 95% of the final speed obtained in the 30–15 Intermittent Fitness Test (V_IFT) in professional handball players. This finding is in line with other research reporting no effect of training intensity on RSA performance outcomes following HIIT models (McGinley and Bishop, 2016). Thus, these two studies concluded that distinct training intensities during HIIT models produce similar performance enhancements when the same total exercise duration (i.e., isotime) is applied. While this isotime approach is a methodological strategy used in several studies to investigate the isolated effect of training intensity (McGinley and Bishop, 2016; Teixeira et al., 2018, 2019; Viaño-Santasmarinas et al., 2018), few studies have addressed the potential effects of distinct HIIT models varying in exercise intensity and total work duration on the athletic performance of team sport players (Ferrari-Bravo et al., 2008; Maggioni et al., 2019). Prior research studies investigating the integrative effects of these two key variables on the adaptive responses to distinct HIIT formats are limited to individual endurance sports athletes (Seiler et al., 2013).

Considering the multidirectional running pattern during futsal matches, HIIT strategies are usually composed of shuttle-runs in order to increase the specificity of these HIIT drills (Akenhead et al., 2014; Teixeira et al., 2018, 2019). In this sense, the simultaneous manipulation of the exercise intensity and total work duration during shuttle run HIIT models will have a direct influence on the magnitude and total number of acceleration and deceleration actions performed for each COD during a training session, respectively (Buchheit and Laursen, 2013; Akenhead et al., 2014). For instance, shuttle-run drills with a greater frequency of COD elicit an increased metabolic, cardiovascular, neuromuscular, and perceptual response in team sport athletes (Akenhead et al., 2014). Of interest, the number of COD performed during a short-term training period (5–6 weeks) was shown to be crucial to improve a variety of physical performance parameters in female team sport players (Sanchez-Sanchez et al., 2018; Teixeira et al., 2019). This research topic still needs further investigation, since the same effects observed in female athletes have not been noticed in male athletes (da Silva et al., 2015; Attene et al., 2016). Therefore, comparative studies examining the effects of two HIIT models at submaximal and maximal intensities with a contrasting total work duration in male futsal players could address practical questions on whether shuttle run HIIT programs, taking into account the physiological consequences of COD (Akenhead et al., 2014; Teixeira et al., 2018), would be more effective if performed at lower intensities with longer sets or at higher intensities with shorter set durations.

The dose-response relationship between the accumulated training load (TL) and performance adaptations is another relevant topic in the field of team sports, which deserves attention from coaches and sport scientists. The majority of studies examining the dose-response relationship have been primarily conducted with soccer and rugby players (Taylor et al., 2018; Daniels et al., 2019; Rabbani et al., 2019; Ellis et al., 2020). To date, few studies have investigated the dose-response relationship between TL and performance adaptations in futsal (Oliveira et al., 2013; Nakamura et al., 2020). Oliveira et al. (2013) did not find any association between TL and heart rate variability (HRV) and aerobic-anaerobic performance changes during seasonal training phases in professional futsal players. Differently, Nakamura et al. (2020) showed that accumulated TL negatively affected the physical and physiological adaptations of elite futsal players. Thus, further studies are still warranted to better understand the potential consequences of accumulated TL during the preseason phase on the subsequent changes in performance in futsal teams, especially in youth players.

The current study aimed to compare the effects of two shuttle run HIIT models performed at 86% (HIIT₈₆) and 100% (HIIT₁₀₀) of peak speed derived from the Futsal Intermittent Endurance Test (FIET, PS_FIET) with a total work duration of 16 and 8 min, respectively, implemented over a period of 10 weeks, on the HRV, aerobic fitness, RSA, and neuromuscular performance of young male futsal players. A second aim of this study was to examine the dose-response relationships between accumulated TL and changes in physical and physiological measures. Based on previous studies (Sanchez-Sanchez et al., 2018; Teixeira et al., 2018, 2019) our hypothesis was that the HIIT model with more COD and longer set duration (HIIT₈₆) would induce superior improvements on the selected physiological and physical measures than the model with less COD and shorter set duration performed at a higher intensity (HIIT₁₀₀).

Descriptive statistics of observed data (mean ± standard deviation [range]) for aerobic, RSA, sprint, and vertical jump performances before (pre-) and after (post-training) the training period (10 weeks) in each HIIT model are presented in Table 1.

Changes for the parameters determined during the FIET and treadmill incremental tests are displayed in Figure 2. The HIIT₁₀₀ model showed clear beneficial effects (i.e., the full 90% CI boundaries out of ROPE) for VO₂peak and PS_TREADMILL measurements. For the HIIT₈₆ model, clear beneficial changes occurred in the VO₂peak, PS_TREADMILL, VT₂, and VT₁. For the HIIT₁₀₀ model, VT₂ and VT₁ changes were well supported within the ROPE, nevertheless, these effects are inconclusive because they overlapped the lower and upper ROPE boundaries. For PS_FIET, both HIIT models presented no clear improvements, although the probabilities were high for improvement (> 89%), low for trivial (< 8%), and negligible for impairment (< 2.6%). Between HIIT models comparisons showed more favorable probabilities in favor of HIIT₈₆ than HIIT₁₀₀ in all parameters; however, clearly more favorable changes for the HIIT₈₆ model compared to the HIIT₁₀₀ model were observed only for PS_TREADMILL, VT₂, and VT₁ (probabilities > 96%).

Pre to post changes for RSA_MEAN, RSA_BEST, 15-m sprint, CMJ, and SJ measures are summarized in Figure 3. The HIIT₁₀₀ model showed clear beneficial effects for RSA_MEAN and CMJ measurements, considerable probability of improvement for 15-m sprint time (84%), and inconclusive effects for RSA_BEST and SJ. The HIIT₈₆ model showed clear beneficial effects for all anaerobic running measurements, except the 15-m sprint time, where a high probability of improvement (92.1%), small/moderate of being trivial (7.1%), and negligible of being harmful (0.8%) were observed. Between HIIT models comparisons showed clearly more favorable changes for the HIIT₈₆ than HIIT₁₀₀ model in the RSA_BEST and SJ measures (probabilities > 96%). All other effects between model were deemed inconclusive.

Pre to post change in the HRV revealed clear beneficial changes in both training models, with probabilities > 98.1% (Figure 4). However, no evidence of superiority of one HIIT model over the other was observed.

The total weekly TLs for both HIIT₈₆ (black bars) and HIIT₁₀₀ (gray bars) models during each training week are presented in Figure 5A. The total accumulated TLs derived from (i) all training sessions and matches (#), (ii) all training sessions and matches without HIIT sessions ($), and (iii) HIIT sessions (^∗) are presented in Figure 5B. For the entire training period, the total accumulated TL from all training sessions and matches showed a mean difference [90% CI] of −1836 [−8279, 4340] arbitrary units (a.u.) between HIIT₈₆ and HIIT₁₀₀ models, with a probability of 55.9% of the TL being lesser in the HIIT₈₆ than HIIT₁₀₀. The total accumulated TL of all training sessions and matches without HIIT was lesser in the HIIT₈₆ than HIIT₁₀₀ (−2671 [90% CI; −4137, 9004] a.u.), with a probability of TL being lesser in the HIIT₈₆ of 76.5%. Contrarily, as expected, the total accumulated TL over the 8 HIIT sessions was two-fold higher in the HIIT₈₆ than the HIIT₁₀₀ model (difference: 778 [90% CI; 609, 941] a.u.), with 100% probability of being higher in the HIIT₈₆ model.

Since the players followed the same training routine, with the exception of the HIIT sessions, the dose-response relationship between RPE-based TL and Δ rMSSD with changes in aerobic, RSA, 15-m sprint, and jump performances was carried out, adding HIIT models as a covariate in the final model. The regression outputs between total weekly TL (10-week average) and ΔrMSSD in addition to HIIT type with changes in performance measures are displayed in Figures 6, 7, respectively. Total weekly TL and HIIT model accounted for 25 to 87% of the variance (i.e., Bayesian R²) in aerobic fitness, RSA, and power-speed-related performance changes. The explained variance derived from regression models using ΔrMSSD and HIIT type as covariates ranged from 26 to 72%.

The current study aimed to compare the effects of two shuttle run-based HIIT models of varying intensity and total work duration (HITT₈₆: 16 min vs. HITT₁₀₀: 8 min) on aerobic, HRV, RSA, and neuromuscular performance outcomes in junior male futsal players. The dose-response relationship between RPE-based TL and changes in performance was also examined. The main findings of this study showed that after 10-weeks of futsal training: (i) the HIIT₈₆ model was clearly more effective at improving the PS_TREADMILL (Δ = 5.4%), VT₂ (Δ = 10.0%), VT₁ (Δ = 13.0%), RSA_BEST time (Δ = −3.5%), and SJ height (Δ = 13.7%) than the HIIT₁₀₀; (ii) RPE-based TL in association with HIIT type explained 71% to 87% of the inter-individual variation in RSA performance changes, while the explained variance for the other parameters was smaller (25–59%); and (iii) changes in HRV along with HIIT type accounted for 72% of inter-individual variance in VO₂peak changes following the training period.

Comparative studies examining the effectiveness of different training models are increasingly needed and recommended to help guide decision-making of strength and conditioning coaches during the planning of training programs with futsal players (Buchheit et al., 2008; Ferrari-Bravo et al., 2008; Sanchez-Sanchez et al., 2018). Training intensity and total work duration are two key variables commonly altered in order to increase the physical capacity of athletes (Buchheit and Laursen, 2013) and, therefore, they need to be well managed during HIIT programming. In the current study, the improvements in physical fitness indices were specific to training type. Our data indicated that the HIIT₈₆ model clearly improved (i.e., full 90% CI out of ROPE) the majority of the physical and physiological parameters measured (9 out of 11 parameters; Figures 2, 3, 4) compared to the HIIT₁₀₀ model (5 out of 11 parameters; Figures 2, 3, 4). In addition, the HIIT₈₆ model induced larger improvements in aerobic, RSA, and neuromuscular performance outcomes than the HIIT₁₀₀ model. The results presented herein suggest that the HIIT₈₆ training, comprising longer sets at a lower intensity, was more effective to enhance performance than the HITT₁₀₀ composed of shorter sets and more intense running efforts. Similarly, Buchheit et al. (2008) also reported greater performance adaptations after HIIT models performed at lower running intensities (i.e., close to 90–95% V_IFT) in male adolescent handball players. Seiler et al. (2013) also showed that training intensity and total work duration influenced the magnitude of adaptive responses following distinct HIIT models (4 × 4 min; 4 × 8 min; 4 × 16 min) in trained cyclists. On the other hand, current studies published in the literature suggests that HIIT models of different training intensities induced similar performance enhancements in male adults (McGinley and Bishop, 2016; Viaño-Santasmarinas et al., 2018). Some differences between these studies should be addressed. For instance, the total work duration in our training design varied between training groups, while prior studies used an isotime approach (i.e., matched-work) only varying exercise intensity (Buchheit et al., 2008; Seiler et al., 2013; McGinley and Bishop, 2016; Viaño-Santasmarinas et al., 2018). In addition, the reference running speed used for training prescription also differed between the cited studies (Buchheit et al., 2008; Viaño-Santasmarinas et al., 2018). Thus, comparison between the current results and those observed in prior studies should be interpreted with caution due to differences in sample characteristics (e.g., chronological age), methodological issues, sport modality, training demands, and performance levels.

It is well known and accepted that aerobic fitness and RSA performance are two discriminant physical qualities of the competitive level in futsal (Álvarez et al., 2009; Pedro et al., 2013; Ayarra et al., 2018). Thus, training strategies targeting the development of these physical qualities simultaneously are essential. Our findings indicating the superiority of HIIT₈₆ over HIIT₁₀₀ at improving aerobic fitness, RSA, and vertical jump performance suggest that this training type (submaximal runs at 86% PS_FIET and longer sets) should be preferentially used with young futsal players. The specific adaptations in physical performance following the HIIT₈₆ and HITT₁₀₀ models could potentially be related to differences in total work duration between the models (Teixeira et al., 2018, 2019). Due to its higher total running volume (16 vs. 8 min), the HIIT₈₆ implies a greater number of COD performed in a single session (96 vs. 48 turns), increasing the total time that athletes spent accelerating per running bout compared to the HIIT₁₀₀ model. Akenhead et al. (2014) showed that the number of turns and time spent accelerating (>± 1 m⋅s^–2) are linearly related during shuttle run drills. Prior research has indicated that accumulated individual acceleration load is positively associated with changes in aerobic fitness and neuromuscular measures in professional soccer players (Clemente et al., 2019). Given that acute and chronic responses to training are dependent on acceleration load accumulated during multidirectional drills (Akenhead et al., 2014; Clemente et al., 2019), practitioners and coaches can increase or decrease the acceleration load accumulated by athlete during shuttle run-based HIIT varying either set duration as in this study or COD frequency per running bout (Sanchez-Sanchez et al., 2018; Teixeira et al., 2018, 2019). To date, to the best of our knowledge, this is the first study demonstrating that a HIIT model with more directional changes leads to superior gains in physical performance in a sample of aerobically well-trained male futsal players (VO₂peak > 55 mL/kg/min at baseline). Previous studies published on this research topic did not show any further performance gain after different shuttle run training models varying the number of COD required per running bout in male soccer and basketball players (da Silva et al., 2015; Attene et al., 2016). Of interest, the same HIIT₈₆ model used here was also applied in a sample of female futsal players (VO₂peak: 47–49 mL/kg/min at baseline) (Teixeira et al., 2018, 2019). The current and prior studies (Teixeira et al., 2018, 2019) showed a greater improvement in PS_TREADMILL and RSA performance after HIIT₈₆ model. This demonstrates the consistency and effectiveness of this training model (HIIT₈₆) to improve these physical qualities in age-matched male and female futsal players. At the same time, male and female futsal athletes of similar ages can display distinct neuromuscular performance adaptations (i.e., changes in SJ and CMJ height) following HIIT₈₆, with male athletes in the current study being more responsiveness (Δ = 13–17%) than female athletes (Δ = 8–9%) in the study of Teixeira et al. (2018). Therefore, it should be highlighted that any generalization of our findings to other samples of futsal (male or female; young or adult) or other indoor team sports (e.g., basketball and handball) would be a hasty inference, since differences related to age, gender and sport demand in terms of workload and training content distribution during a typical training period could influence the effectiveness of this HIIT₈₆ model in these other sports scenarios.

From the present results on RPE-based TL and changes in physical performance, it is possible to make inferences about the dose-response relationship during the training process. The present study suggests that TL in addition to training type (HIIT₈₆ and HIIT₁₀₀) accounted for a large portion of the inter-individual variance in RSA (71–87%), 15-m Sprint (46%), and vertical jump (40–47%) performance changes. A positive linear dose-response relationship between TL and changes in these performance measures was found for the HIIT₈₆ model, while no (RSA performance) or negative (sprint and vertical jump) relationships were identified for the HIIT₁₀₀ model. These findings are of practical relevance for practitioners and coaches. First, players in both HIIT₈₆ and HIIT₁₀₀ models with a similar total TL displayed a distinct adaptive response (especially for RSA_BEST and SJ performance), highlighting that the quality/specificity of the training stimuli is the most relevant component of the training process (Sanchez-Sanchez et al., 2018). In this case, the increased number of COD in the HIIT₈₆ model (longer sets) may have been decisive to induce superior gains in performance. Second, players who accumulated higher training loads in the HIIT₈₆ model demonstrated the largest improvements in RSA, 15-m Sprint, and vertical jump performance, while the opposite was observed for the HIIT₁₀₀ model. Although these results cannot be easily explained from the data analysis employed in our study, it is possible to suggest that the training loads derived from other training strategies (technical-tactical, strength-power, friendly and official matches) may have influenced the players’ adaptive response to training. For instance, players in the HIIT₁₀₀ model tended to accumulate a higher (with a 73% probability) TL derived from these other training contents (i.e., not involving HIIT sessions) than the HIIT₈₆ model (Figure 5).

The explained variances derived from regression models (TL and HIIT type inserted as covariates) for the changes in maximal (VO₂peak, PS_FIET, PS_TREADMILL) and submaximal (VT₂ and VT₁) aerobic performance measures were considered low (25–28%) and moderate (45–59%), respectively. Contrary to what was observed in the anaerobic and jump performance measures, the dose-response relationship for aerobic performance outcomes did not show a distinct pattern between HIIT models (i.e., regression slopes in contrary directions). Of note, positive and negative/null relationships were observed between accumulated TL and changes in maximal (VO₂peak and PS_TREADMILL) and submaximal (VT₂ and VT₁) aerobic indices, respectively. Several studies have demonstrated a linear dose-response relationship between RPE-based TL and changes in aerobic performance indicators in team sports athletes (Oliveira et al., 2013; De Freitas et al., 2015; Clemente et al., 2019; Ellis et al., 2020; Nakamura et al., 2020). Interestingly, a recent study conducted by Ellis et al. (2020) also found a negative weak association between TL and changes at fixed lactate thresholds of 2 and 4 mmol/L (Bayesian r = −0.17 and −0.16, respectively) after 6 weeks of pre-season in a sample of male junior soccer players. Furthermore, in agreement with our data, a small positive relationship was verified between TL and PS_TREADMILL changes (Bayesian r = 0.37) in the previously cited study (Ellis et al., 2020). It should be highlighted that the present study included 2 different shuttle-run HIIT types, so we chose to analyze the dose-response data with an interaction between TL and HIIT type. Thus, our Bayesian R² values consider both the relationship of the response to the TL and the HIIT model. Therefore, comparisons with the R² from other studies should be performed with caution to avoid misinterpretations, since they only used the TL as a covariate.

An interesting finding from our study to be highlighted was that similar improvements in the resting HRV were noticed after both HIIT models (HIIT₈₆ and HIIT₁₀₀) outlined here using PS_FIET as the reference speed to calibrate running distance. Although improved resting HRV after a period of futsal training has been previously documented in the literature (Oliveira et al., 2013; De Freitas et al., 2015; Nakamura et al., 2020), our data reinforce the effectiveness of shuttle run HIIT training models to induce positive adaptations in the cardiac autonomic function of young futsal players. This assumption was previously confirmed by Buchheit et al. (2008) who showed that HIIT models are preferable compared to repeated all-out sprint training methods, since its effects on cardiac autonomic function were significantly more pronounced in young handball players. The resting HRV, a non-invasive assessment of cardiac autonomic modulation, has also been constantly related to aerobic fitness indices and high-intensity running performance changes in team sport athletes (De Freitas et al., 2015; Esco et al., 2016; Nakamura et al., 2020). The present study showed a positive relationship between changes in resting HRV (i.e., rMSSD) and VO₂peak after the training period, suggesting that players in both HIIT models with greater increases in resting rMSSD demonstrated the largest increments in VO₂peak. Of interest, changes in resting HRV and HIIT types accounted for 72% of the inter-individual variance in VO₂peak changes in our sample. Another two studies performed with futsal players also showed that an enhanced vagal modulation (inferred by an increase in resting rMSSD value) was largely positively correlated (r = 0.62 and 0.64) with improvements in Yo-Yo IR1 performance (De Freitas et al., 2015; Nakamura et al., 2020). These explained variances (38–40%) are, at least in part, like those observed in our study for the ΔPS_FIET (39%) and ΔPS_TREADMILL (47%) (Figure 7). Collectively, the current and previous results indicate that resting rMSSD may be a simple and sensitive indicator to monitor changes in physical fitness during training.

One of the strengths of this study was to show how different HIIT models associated with other training components can influence the adaptive responses of players from the same futsal team. In this scenario, where few traditional HIIT sessions are planned by the team technical staff due to the matches schedule and the importance given to technical-tactical and strength/injury prevention sessions, the selection of training stimuli in HIIT sessions is key to maximizing subsequent performance adaptations. From a practical perspective, strength and conditioning specialists should consider spending more time in less intense shuttle run HIIT sessions with more COD than in more intense and shorter sessions with fewer directional changes.

The main limitation of the present study was the small sample size (n = 11). This can be justified by the low number of players who are part of a futsal squad (12–15 players). It would be extremely difficult to monitor more than one futsal team at one time. However, Bayesian analysis is better suited for making inferences on small sample sizes, as the MCMC methods used to produce posterior distributions do not depend on asymptotic behavior in the same way that traditional frequentist methods do (Kadane, 2015). Additionally, Bayesian inferences are more intuitive by posterior probability distributions of parameters (Kruschke, 2018). Nevertheless, a challenge in Bayesian statistics is the necessity to impose a prior knowledge in the parameters (i.e., prior distributions). In this way, we used non-informative default priors of the brms package (Bürkner, 2017). Therefore, the priors had little influence on the results (Gelman et al., 2008). Another point related to the analysis is that in team sports, such as futsal, there is no ROPE directly linked to performance. Thus, we used a ROPE of 20% between subjects SD [i.e., Cohen’s d standardized mean differences transformed to percent units (Hopkins et al., 2009; Kruschke, 2018)] to infer about the substantiality of our results. Also, we acknowledge that the use of a single day HRV records at pre and post-training period is in disagreement with recent statements suggesting a minimum of 3 randomly selected HRV measurements per week (Plews et al., 2014; Nakamura et al., 2020). Another limitation of the present study is that no objective external load measure was used. A recent prior study suggested that decelerations, number of sprints, and distance covered should be considered to better discriminate the physical load of elite futsal players (Ribeiro et al., 2020). Future studies should consider using larger samples and adding more objective measures to run multiple linear regressions to explain the variations in determinant fitness variables.

This study showed that those players who underwent 8 shuttle run HIIT sessions at 86% PS_FIET had superior gains in aerobic, RSA, and neuromuscular performance measures than those who trained at 100% PS_FIET during a typical 10-week training period. In addition, the variance explained by the TL along with the HIIT type was clearly larger for the changes in RSA performance outcomes than that observed for aerobic and neuromuscular performance changes. Finally, monitoring resting HRV could be a suitable tool to track changes in VO₂peak, since temporal alterations in HRV are strongly related to VO₂peak changes.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by this study was approved by the local research ethics committee (n° 93777318.0.0000.0121). Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

FC participated in the design of the study, data collection, data organization, and drafted the manuscript. FB participated in the data organization, and performed the statistical analysis, interpretation, and discussion of results. LF and LB contributed with support of materials and data collection. AT participated in the design of the study and interpretation and discussion of results. RHN contributed to the design of the study and interpretation of results. LG contributed to the design of the study, interpretation and discussion of results, and coordination of project. All authors contributed to the writing and approved the final version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.