Article Text

Download PDFPDF

Original article
Physical workload and risk of long-term sickness absence in the general working population and among blue-collar workers: prospective cohort study with register follow-up
  1. Lars Louis Andersen1,2,
  2. Nils Fallentin1,
  3. Sannie Vester Thorsen1,
  4. Andreas Holtermann1,3
  1. 1National Research Centre for the Working Environment, Copenhagen, Denmark
  2. 2Physical Activity and Human Performance Group, SMI, Department of Health Science and Technology, Aalborg University, Aalborg, Denmark
  3. 3Department of Sports Science and Clinical Biomechanics, Institute of Sports Science and Clinical Biomechanics, Physical Activity and Health in Work Life, University of Southern Denmark, Odense, Denmark
  1. Correspondence to Professor Lars L Andersen, National Research Centre for the Working Environment, Lersø Parkalle 105, Copenhagen DK-2100, Denmark; lla{at}nrcwe.dk

Abstract

Objective To determine the prospective association between physical workload—in terms of specific physical exposures and the number of exposures—and long-term sickness absence (LTSA).

Methods Using cox-regression analyses, we estimated the risk of register-based incident LTSA (at least 3 consecutive weeks) from self-reported exposure to different physical workloads among 11 908 wage earners from the general working population (Danish Work Environment Cohort Study year 2000 and 2005).

Results The incidence of LTSA was 8.9% during two-year follow-up. Spending 25% or more of the total work time with a bent or twisted back (HR 1.59 (95% CI 1.39 to 1.83)), arms above shoulder height (HR 1.35 (95% CI 1.14 to 1.59)), squatting or kneeling (HR 1.30 (95% CI 1.09 to 1.54)), pushing/pulling or lifting/carrying (HR 1.40 (95% CI 1.22 to 1.62)) and standing in the same place for 50% or more of total work time (HR 1.19 (95% CI 1.00 to 1.42), were risk factors for LTSA when adjusted for baseline age, gender, psychosocial work environment, lifestyle, musculoskeletal and mental disorders, and socioeconomic status. HR increased from 1.25 (95% CI 1.04 to 1.51) for one to 1.94 (95% CI 1.56 to 2.41) for four combined physical workloads. Results largely remained stable in subgroup analyses including only blue-collar workers (n=5055). Population attributable risks for LTSA from one or more physical workloads were 26% and 40% in the general working population and among blue-collar workers, respectively.

Conclusions Several of the investigated types of physical workload were risk factors for LTSA when exceeding 25% of the work time. A higher number of combined physical workloads was associated with progressively increased risk. Our study underscores the importance of physical workload as risk factors for LTSA in the general working population as well as among blue-collar workers.

View Full Text

Statistics from Altmetric.com

What this paper adds

  • High physical workloads can cause ill health, but detailed knowledge about exposure time and type of exposure, and the combination of different types of physical workloads in relation to development of ill health, is lacking.

  • Our study shows that exposure to physical workloads—such as bending or twisting the back, working with arms above shoulder height, squatting or kneeling, or pushing/pulling/lifting/carrying—for as little as a quarter of the working time, is a risk factor for long-term sickness absence in the general working population as well as among blue-collar workers.

  • Importantly, a higher number of combined physical workloads was associated with progressively higher risk for long-term sickness absence.

  • While risk factors in the physical working environment are typically evaluated as single risk factors, our study shows that combined exposure to different physical workloads is even more relevant to consider. Work environment professionals and authorities should use this knowledge when evaluating safety and health at workplaces.

Introduction

Years of hard manual labour can take a toll on the body. Although the physical work environment has improved substantially, preventing work-related musculoskeletal disorders and sickness absence remains a challenge in the 21st century.1–4 Work-related musculoskeletal disorders are associated with sickness absence in several occupations2–8 and remain among the leading reasons for sickness absence across Europe.1 ,9 ,10 Kiwimäki et al11 associated five or more spells of sickness absence with increased risk of mortality during follow-up, positioning sickness absence as a global indicator of health. For the individual, being gainfully employed is also important for social identity and well-being.12 Thus, measures to prevent sickness absence are relevant for individuals, workplaces and societies alike.13

Research from different countries has established several risk factors in the physical work environment, for development of ill health—either in terms of health symptoms or sickness absence.14–22 Among the general working population of Norway, Sterud found physical risk factors in the work environment to account for a quarter of all long-term sickness absence (LTSA), with neck flexion, hand/arm repetition, standing, working with upper body bent forward and awkward lifting, as significant risk factors.18 In the Danish Work Environment Cohort Study conducted in the year 2000, Bang Christensen et al9 estimated that physical workload accounted for 10–30% of all LTSA.9 Lund et al23 found that bending of the back during work was a risk factor for sickness absence both in Denmark and in Sweden. However, the majority of studies on physical workload and ill health have used self-reported symptoms as outcome. Thus, Mayer et al14 found, in their systematic review—including 19 high-quality prospective studies—an association between development of shoulder and/or neck pain with manual material handling, repetitive work, vibration, trunk flexion or rotation and working with hands above shoulder height. In another systematic review of longitudinal studies, da Costa and Vieira16 found that the most commonly reported biomechanical risk factors for musculoskeletal disorders were excessive repetition, awkward postures and heavy lifting. Likewise, Griffith et al24 found, in an individual data meta-analysis, an association between physical workload and low back pain. Overall, the major physical risk factors for ill health based on the existing literature can be summarised into heavy lifting, working with a bent or twisted back or elevated arms, and using repetitive movements and vibration. Because the majority of these studies used self-reported exposure and outcome, there is a need for more studies using objective outcomes such as register-based sickness absence.

While previous studies assessed risks from specific isolated types of physical workload for LTSA, real-life working situations often consist of a multitude of different types of physical workloads. No studies have assessed whether exposure to a higher number of different types of physical workload is associated with progressively higher risk of LTSA. Thus, determining the association between the number of physical workloads in the work environment and development of ill health is vital for reaching consensus for practical guidelines and policy regulations—both in regard to exposure time and number of exposures.

Using data from the Danish Work Environment Cohort Study (DWECS) merged with a register of social transfer payment (DREAM), this study determines the prospective association between physical workload—in terms of specific physical exposures and the number of exposures—and LTSA in the general working population as well as among blue-collar workers.

Methods

Study population

Data on work environment and health in the study population were obtained from the 2000 and 2005 rounds of the Danish Work Environment Cohort Study (DWECS).25 This study consists of a survey assessing work environment and health in the general working population, and the 2000 and 2005 round were analysed together to increase sample size, and, thus, statistical power. In the 2000 survey, the National Research Centre for the Working Environment (NRCWE) collected data by telephone interview (94%) or face-to-face interview in the respondents’ own homes (6%); in the 2005 survey, NRCWE collected data by postal questionnaire (61%), telephone interview (36%) and internet (3%). The questions on physical work environment are specified below. Lund et al22 and Bang Christensen et al9 previously reported some of the data from the 2000 round. The 2000 sample consisted of 11 498 people, of whom 8577 participated (74.6%), 1721 refused to participate (15.0%), 253 were sick, hospitalised, away or disabled (2.2%), 880 were not at home, had moved or were otherwise not possible to reach (7.7%), and 67 had emigrated or died (0.6%). Of the 8577 who participated in the 2000 round, 5632 were wage earners. The 2005 sample consisted of 26 244 people, of whom 15 232 participated (58.0%), 2830 refused to participate (10.8%), 246 were sick, hospitalised, away or disabled (0.9%), 6832 were not at home, had moved or otherwise not possible to reach (26.0%), and 1104 had emigrated or died (4.2%). Of the 15 232 who participated in the 2005 round 11 500 were wage earners. If participants of the cohort had replied in 2000 and 2005, only their reply from 2000 was used. For the present analyses, we included only wage earners free from LTSA. After excluding 1160 individuals who were on LTSA during the baseline year and the year before (ie, 1999 and 2000 for the 2000 round, and 2004 and 2005 for the 2005 round), the combined 2000 (N=4692) and 2005 (N=7216) data set for this study consisted of 11 908 wage earners (unique individuals with one observation each) from the general working population in Denmark. Blue-collar and white-collar workers constituted 5055 and 6853 individuals, respectively. Table 1 shows the baseline characteristics of the study population.

Table 1

Characteristics of the participants (N=11 908)

Ethical approval

The study has been notified to and registered by Datatilsynet (the Danish Data Protection Agency; journal number 2007-54-0059). According to Danish law, questionnaire-based and register-based studies do not need approval by ethical and scientific committees, nor do they need informed consent.26 ,27 All data were de-identified and analysed anonymously.

Physical workload

The questions asked regarding exposure to physical workload28 were, ‘does your work cause you to…’(1) stand in the same place?, (2) work with the back bent strongly forward without support from the hands and arms?, (3) twist or bend the back several times per hour?, (4) have the hands lifted to or above shoulder height (2000 round), have the arms lifted to or above shoulder height (2005 round), (5) perform the same arm movements several times a minute? (eg, package work, mounting, machine feeding, carving), (6) squat or kneel when you work?, (7) lift or carry?, and (8) push or pull? Response options were (1) almost all the time, (2) approximately 3/4 of the time, (3) approximately 1/2 of the time, (4) approximately 1/4 of the time, (5) seldom/very little, and (6) never. For subsequent analyses, the extreme response options were collapsed, in other words, 1 and 2 were defined as 75–100% of the time, option 3 as 50% of the time, option 4 as 25% of the time and option 5 and 6 as 0–12.5% of the time.18

In additional analyses, we dichotomised the variables concerning the different types of physical workload to obtain more statistical power. Further, for types of physical workload that were highly correlated, the highest value was used, that is, ‘back strongly bent’ and ‘frequent twisting or turning of the back’ (spearman's r= 0.60) as well as ‘pushing/pulling’ and ‘lifting/carrying’ (spearman's r=0.61). Cut points for the dichotomisation were based on the lowest exposure time that reached statistical significance, that is, according to table 2, 25% or more of the work time was default except for ‘standing in the same place’, where 50% or more of the time was used.

Table 2

Percentages of work time spent in eight different types of physical workload and risk of long-term sickness absence among the total cohort

Outcome variable

Data on sickness absence was obtained from a register of social transfer payment (DREAM),29 and linked to DWECS via the unique personal identification number that is given to all Danish citizens at birth. The DREAM register contains weekly information on sickness absence compensation, employment, education, disability pension, etc, for all citizens in Denmark. Sickness absence compensation is given to the employer, who can apply for a refund from the state for employees after 3 weeks of sickness absence. LTSA was therefore defined as having registered sickness absence for a period at least three consecutive weeks in the follow-up period from 1 January 2001 to 31 December 2002 for the 2000 round, and from 1 January 2006 to 31 December 2007 for the 2005 round (ie, 2 calendar-years).

Control variables

Control variables at baseline were gender, age, year of questionnaire reply (2000, 2005), four dimensions of psychosocial work environment30 (described below), smoking status (‘Never’, ‘Former’ and ‘Current’), body mass index (BMI), leisure-time physical activity (described below), back pain and neck-shoulder pain31 ,32 (described below), mental health (from the SF-36 questionnaire33) and socioeconomic status34 (described below). Age, psychosocial work environment, BMI, back pain, neck pain, mental health and socioeconomic status, were entered into the analyses with linear effects on LTSA.

The four psychosocial work environment dimensions included (1) influence at work, (2) support from colleagues, (3) support from superiors and (4) emotional demands. Each dimension included a number of questions from the Copenhagen Psychosocial Questionnaire (COPSOQ).30 Leisure-time physical activity was assessed by the question ‘How much time have you spent on each of the following leisure-time activities during the last year (including commuting to and from work)?’, with the following three subquestions: (1) ‘Walking, biking or other low-intensity exercise, where you do not get short of breath and do not begin to sweat (eg, Sunday walks or low-intensity gardening)?’, (2) ‘Exercise training, heavy gardening, or higher intensity walking/biking, where you sweat and get short of breath?’ and (3) ‘Strenuous exercise training or competitive sports?’. The response options for each subquestion were ‘>4 h/week’, ‘2–4 h/week’, ‘<2 h/week’ or ‘do not perform this activity’. For subsequent analyses (1) ‘low physical activity’ was defined as performing <4 h of low-intensity physical activity per week, and not performing moderate-intensity and high-intensity activities at all, (2) ‘moderate physical activity’ was defined as performing more than 4 h of low-intensity physical activity per week or moderate activity for <4 h/week or high-intensity activity for <4 h/week and (3) high physical activity was defined as performing moderate or high activity for >4 h/week, or a combination of moderate and high activity for 2–4 h/week. Intensity of musculoskeletal pain in the low back, neck, right shoulder and left shoulder, respectively, was assessed by a modified version of the PRIM questionnaire,31 based on a 10-point pain score.32 The following question was asked ‘On a scale of 0–9 with 0 being no discomfort at all and 9 being the worst possible pain, state your average degree of discomfort in your (body region) in the past 3 months’. Subsequently, pain in the neck and shoulders was collapsed due to the high intercorrelation of pain between these regions.35 Thus, for the neck, and right and left shoulder, only the highest of the three values was used. For socioeconomic status, respondents were classified into five groups according to employment grade, job title and education.34 Blue-collar workers comprised skilled, semiskilled or unskilled workers (socioeconomic status 4 and 5). White-collar workers included professionals, managers, academics, people with 3–4 years of vocational education and other salaried workers (socioeconomic status 1, 2 and 3).

Statistical analyses

Using the PHREG procedure on SAS V.9.3, the Cox proportional hazard model36 was used for modelling the probability of incidence of LTSA during the 2-year follow-up period, with physical workloads as explanatory variables. The data on LTSA correspond to survival times, which in most cases are censored as the cohort is only followed for 2 years. When individuals had an onset of LTSA during the 2-year follow-up, the survival times were non-censored and referred to as event times. The estimation method was maximum likelihood. Results are reported as HRs with 95% CIs.

The first set of analyses concerned associations between percentage of work time of each physical workload separately and LTSA. Model 1 was adjusted for age, gender and year of questionnaire reply (2000, 2005). Model 2 was the same as model 1, but additionally included psychosocial work environment (influence at work, emotional demands, support from colleagues and support from superiors). Model 3 was the same as model 2, but additionally included lifestyle factors (smoking, leisure physical activity, BMI). Model 4 was the same as model 3, but additionally included back pain, neck-shoulder pain and mental health. Model 5 was the same as model 4, but additionally included socioeconomic group.

Using the same five models as described above, the second set of analyses used the dichotomised physical exposure variable. Subgroup analyses were also performed among blue-collar workers only (n=5055 socioeconomic status 4 and 5).

Using the same five models as described above, the third set of analyses concerned association between number of physical workloads and LTSA. For each factor, the cut point from the second set of analyses was used, that is, 25% or more of the work time, except for ‘standing in the same place’, where 50% of the time was used. The number of physical workloads reaching this cut point was then added for a total of 0 to 6. Because few individuals reached 5 or 6 physical workloads, the upper limit was set to ≥4. Subgroup analyses were also performed among blue-collar workers only. Additionally, population attributable risks (PAR) were calculated for the general working population as well as for blue-collar workers, based on the HRs and proportions exposed (Pe) from model 5 of the third set of analyses where PAR=Pe(HRe−1)/(1+Pe(HRe−1)). Pe included all the numbers of exposures above 0.

Results

In the study population—free from LTSA during the baseline year and the year before—the incidence of LTSA during 2-year follow-up was 8.9% in the general working population and 8.5% among blue-collar workers.

Table 2 shows prospective associations between percentage of work time being exposed to eight different types of physical workloads and risk of LTSA among the total cohort. For six of eight physical workloads—back strongly bent, frequent twisting or turning the back, arms above shoulder height, squatting or kneeling, pushing or pulling, and lifting or carrying—spending more than 25% of the work time was a risk factor for LTSA, but the results revealed no consistent pattern of more harm from a relatively higher than 25% time of exposure during work.

Table 3 shows the results from the analyses where the physical exposure variables are dichotomised. Spending 25% or more of the total work time with bent or twisted back (HR 1.59 (95% CI 1.39 to 1.83)), arms above shoulder height (HR 1.35 (95% CI 1.14 to 1.59)), squatting or kneeling (HR 1.30 (95% CI 1.09 to 1.54)), pushing/pulling or lifting/carrying (HR 1.40 (95% CI 1.22 to 1.62)) and standing in the same place for 50% or more of total work time (HR 1.19 (95% CI 1.00 to 1.42), were associated with LTSA in the fully adjusted model. Repetitive arm movements did not reach statistical significance (HR 1.15 (95% CI 0.99 to 1.34)). Table 4 shows that these results largely remained in subgroup analyses among blue-collar workers only, although some associations were attenuated in the final model adjusting for musculoskeletal pain as well as mental health. By contrast, repetitive arm movements became statistically significant (HR 1.28 (95% CI 1.03 to 1.58)).

Table 3

Associations between exposure to different types of physical workload and long-term sickness absence in the total cohort

Table 4

Associations between exposure to different types of physical workload and long-term sickness absence in the subgroup of blue-collar workers only (socioeconomic status 4 and 5)

In addition to the analyses presented in table 3, we performed a sensitivity analysis for model 5, also including individuals who were on LTSA at baseline and the year before (not shown in the table). This analysis resulted in HRs quite similar to those in the main analysis; spending 25% or more of the total work time with bent or twisted back (HR 1.50 (95% CI 1.34 to 1.69)), arms above shoulder height (HR 1.38 (95% CI 1.20 to 1.57)), squatting or kneeling (HR 1.33 (95% CI 1.16 to 1.53)), pushing/pulling or lifting/carrying (HR 1.47 (95% CI 1.31 to 1.66)), repetitive arm movements (HR 1.19 (95% CI 1.05 to 1.34)), and standing in the same place for 50% or more of total work time (HR 1.21 (95% CI 1.04 to 1.40), were associated with LTSA in the fully adjusted model.

Table 5 shows associations between the summed number of physical workloads among the total cohort and risk of LTSA. In general, these analyses showed increased risk with exposure to a higher number of physical workloads with HRs increasing from 1.25 (95% CI 1.04 to 1.51) for one to 1.94 (95% CI 1.56 to 2.41) for four combined physical workloads. Table 6 shows that these results largely remained stable in subgroup analyses including blue-collar workers only.

Table 5

Summed number of physical workloads among the total cohort from table 3 and risk of long-term sickness absence

Table 6

Subgroup analyses of blue-collar workers only (socioeconomic status 4 and 5)

In addition to the analyses presented in table 5, we performed a sensitivity analysis for model 5 also including individuals who were on LTSA at baseline and the year before (not shown in the table). This analysis resulted in quite similar HRs as the main analysis; exposure to 1 (HR 1.27 (95% CI 1.08 to 1.49)), 2 (HR 1.58 (95% CI 1.33 to 1.87)), 3 (HR 1.69 (95% CI 1.40 to 2.03)) and 4 or more (HR 2.06 (95% CI 1.72 to 2. 47)) physical workloads was associated with LTSA in the fully adjusted model.

On the basis of model 5 of tables 5 and 6, the PRA for LTSA of having one or more physical workloads was 26% in the general population and 40% among blue-collar workers.

Discussion

The main findings of our study are that (1) several of the investigated types of physical workload posed risk for LTSA when exceeding 25% of the work time, (2) exposure to a higher number of physical workloads was associated with higher risk and that (3) the findings among the general working population largely remained stable for blue-collar workers only.

Before discussing the results, some limitations and strengths need to be addressed. First, Danish registries hold no information on causes of sickness absence, which limits any causal inference between physical workload and sickness absence caused by a specific type of illness or disease. Second, the type and exposure time of the physical workloads were self-reported, hence being prone to less accuracy—and potential bias.37 Thus, because questionnaire-based information of physical workload depends on workers’ understanding, interpretation and memory, the precision is less than for direct measurements.38 This may cause larger CIs of the risk estimates, and therefore increase the probability for committing a type II error in our study. However, questionnaire-based information on a variety of physical behaviours is documented to be systematically biased by factors such as musculoskeletal disorders, and socioeconomic and demographical variables.39 ,40 In our study, this may lead to an increased probability of committing a type I error. However, because we adjusted for several potential self-reporting bias factors at baseline, we reduced the risk of imposing a significant effect on the results of our study. Nevertheless, the results should be interpreted within these limitations. Our study also has strengths. First, combining the 2000 and 2005 rounds increased the statistical power of the study. Although changes in exposure may occur between the different rounds, this will only affect prevalence rates and not affect risk estimates. Nevertheless, we controlled for each round by including the year of questionnaire reply as a confounder in the statistical analyses. Second, using information on sickness absence from the DREAM register eliminates recall and reporting bias, as the register does not build on the individual workers self-report. The employer can apply for a compensation of employee sickness absence costs from the state after 3 weeks of sickness absence. Because the employer has a strong economic incentive to report sickness absence, the validity of the sickness absence data is high.41 Thus, any change in registration of the outcome is unlikely to have occurred between 2000 and 2005. Third, using a representative sample of the general working population increases the generalisability. Fourth, because previous LTSA is a strong predictor of future LTSA and to avoid cross-sectional associations, we included only workers free from LTSA during the baseline year and the year before. We also performed sensitivity analyses including individuals who were on LTSA at baseline and the year before. These analyses showed results broadly similar to those of the main analyses. Fifth, the main results largely remained among blue-collar workers only, showing that the results were not due to socioeconomic confounding.

Bending, twisting and turning the back for more than a quarter of the work time was the strongest risk factor for LTSA, and was in the fully adjusted model associated with 59% increased risk in the general working population and 65% among blue-collar workers. Next, pushing/pulling or lifting/carrying for more than a quarter of the working time was also a risk factor for LTSA, and was in the fully adjusted model associated with 40% increased risk in the general working population and 38% among blue-collar workers. Likewise, Sterud found, in a representative sample of the general working population in Norway, the strongest risk of LTSA from awkward lifting, working with the back bending forward and manual lifting.18 Our results confirm that physical exposure on the back—that is, bending, twisting and turning—is a risk factor for LTSA.

Working with arms above shoulder height for more than a quarter of the work time was a risk factor for LTSA, and was in the fully adjusted model associated with 35% increased risk in the general working population and 23% non-significant increased risk among blue-collar workers. This is somewhat in line with the systematic review of high-quality prospective studies by Mayer et al,14 who found an association between development of ill health—in terms of self-reported neck-shoulder pain—and working with hands above shoulder height. In the present study, squatting and kneeling were also risk factors for LTSA, which is in line with the findings of Sterud.18 Mayer et al14 found, in their systematic review, that repetitive work was a risk factor for shoulder and neck pain. In the present study, repetitive arm movements were a significant risk factor for LTSA among blue-collar workers, but not among the general population. Speculatively, repetitive movements among blue-collar workers may be associated with higher forces than those found among the remainder of the general working population.

Because a typical working day, for many workers, consists of a multitude of tasks, we also counted exposure to the number of physical workloads and their prospective association with LTSA. We are not aware of any previous studies that have provided such an analysis. The analyses showed increased risk from exposure to a higher number of physical workloads, with 25% increased risk for one versus no physical workload, to 94% increased risk for four or more physical workloads. These results largely remained in subgroup analyses including only blue-collar workers. The mechanism behind this finding may be a detrimental effect on health from accumulated physical strain on the body from several types of physical workloads. In other words, exposure to several physical workloads likely reflects the total amount of exposure. Our results show that exposure to a combination of physical workloads is more important than exposure to a single specific physical work demand. Working environment professionals, stakeholders and authorities ought to take this into consideration when assessing the physical working environment.

In the fully adjusted models, PRA of physical workload—none versus 1 or more—for LTSA were 26% and 40% in the general working population and among blue-collar workers, respectively. Although the risk estimates were broadly similar between the general working population and blue-collar workers, a larger proportion of blue-collar workers are exposed to physical workloads and consequently the attributable fraction is higher. This emphasises the importance of reducing physical workloads, especially at blue-collar workplaces. Interestingly, our result in the general working population is strikingly similar to those of Sterud who reported an attributable risk of 25% for similar types of physical workload for LTSA. Together, these findings underscore that a large proportion of LTSA in the general population—and especially among blue-collar workers—could theoretically be reduced by decreasing exposure time to these physical workloads. This may, for example, be achieved by job rotation including tasks that are not physically demanding, or by use of technical devices where appropriate.

In conclusion, exposure to several of the investigated physical workloads posed significant risk for LTSA when exceeding 25% of the work time. Exposure to higher number of physical workloads was associated with higher risk. Our study underscores the importance of physical workloads as risk factors for LTSA in the general working population as well as among blue-collar workers.

Acknowledgments

The authors would like to express their gratitude to Elsa Bach and Ebbe Villadsen for valuable help during the project, and access to the DWECS and DREAM data.

References

View Abstract

Footnotes

  • Contributors LLA designed the study and performed the analyses. NF, SVT and AH provided feedback on the study design and analyses. LLA drafted the manuscript, and all the authors provided critical feedback and approved the final version.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement All raw data of this study are available on request to the corresponding author.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.