This systematic review was conducted to help clarify the effect of lifting at work on pregnancy outcome, by focusing on specific exposure categories. A search in Medline and Embase identified 51 articles reporting association of spontaneous abortion (SA), preterm delivery (PTD) or small-for-gestational-age (SGA) infant with exposure to occupational lifting. A global validity score was assigned to each study and six potential sources of bias were considered in sensitivity analyses. For each exposure–outcome combination, a summary risk estimate (RE) was obtained from all studies and from a subset of studies with high validity score, this latter summary RE was selected as a final result. Statistical heterogeneity was measured with I2 and Q tests and the possibility of a publication bias was also assessed. For each meta-analysis, the strength of evidence was established from explicit criteria. Heavy (or ≥10 kg) loads often (or ≥10x/day) lifted were associated with increased risks of SA (summary RE=1.31, 95% CI 1.17 to 1.47) and PTD (summary RE=1.24, 95% CI 1.07 to 1.43), with good strength of evidence. No association was identified with SGA, nor with lower exposure levels and SA or PTD. These results are reassuring for lower levels of exposure; however, observed associations can guide health professionals’ recommendations aimed at the prevention of SA and PTD for pregnant women who frequently lift (or ≥10x/day) heavy (or ≥10 kg) loads at work.
- occupational health practice
- female reproductive effects and adverse pregnancy outcomes
- physical work
Statistics from Altmetric.com
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.
- occupational health practice
- female reproductive effects and adverse pregnancy outcomes
- physical work
What is already known about this subject?
Several studies suggest that occupational weight lifting may increase the risk of adverse pregnancy outcomes, but evidence from systematic reviews is conflicting.
Definitions of occupational lifting differ substantially among the studies.
What are the new findings?
Results from 51 studies were distributed among five exposure categories and three adverse pregnancy outcomes.
For pregnant workers who lift frequently (or ≥10x/day) heavy (or ≥10 kg) loads, positive associations are measured with preterm delivery and spontaneous abortion.
The strength of evidence established following the assessment of statistical precision, results of sensitivity analyses and consistency indicates a good level of confidence towards these results.
How might this impact on policy or clinical practice in the foreseeable future?
These results can guide health professionals’ recommendations to reduce the risks of spontaneous abortion and premature delivery in pregnant workers who lift loads.
In the Province of Quebec, Canada, 83% of women work during their reproductive age,1 and occupational lifting is quite common for pregnant workers. Depending on the sources consulted and the definitions used, it is estimated that 12%–28% of pregnant women lift loads at work.2 3 As in other parts of the world, we have a programme to protect the health of pregnant workers, and occupational physicians should assess whether the tasks involve dangers.4
Several studies suggest that occupational weight lifting may increase the risk of adverse pregnancy outcomes, but evidence from systematic reviews is conflicting.5–9 For some, lifting loads at work is significantly associated with increased risks of adverse pregnancy outcomes,5 9 while for others, the risks of spontaneous abortion (SA), preterm delivery (PTD) and small-for-gestational-age (SGA) infant are, at most, minimal.6–8 However, almost all mention that the definitions of lifting differ substantially among the studies,6–9 so much so that, for some, the results cannot be combined.6 8 While in some reviews, the results were combined according to broad exposure categories such as physically demanding work including lifting,5 lifting more than 5 kg9 or lifting at least 100 kg (product of weight and frequency) per working day.7
As the evidence is controversial and the exposure categories used in systematic reviews are broad, it is difficult to make effective and specific recommendations to protect pregnant workers lifting loads. The implementation of effective preventive measures requires a better knowledge of the level of lifting able to produce detrimental effects.
This systematic review of observational studies carried out among pregnant workers was conducted to help clarify this issue by focusing on more specific exposure categories for the risk assessment of three major adverse pregnancy outcomes (SA, PTD and SGA).
This systematic review is reported according to Preferred Reporting Items for Systematic Review and Meta-analysis10 and follows a methodology developed for previous meta-analyses.11 12
Literature search and study selection
A search in Medline and Embase databases (online supplementary file 1) was conducted to identify potentially eligible papers published until January 2019. Reviews and guidelines were set aside, while conference abstracts, comments and results already published were excluded. Articles were included for evaluation if they were written in English, French or Spanish and reported the results of observational studies (cohort or case–control) assessing the risks of SA, PTD or SGA by comparing pregnant women who lift weight at work to pregnant workers who do not. Results have to be relative risk (RR) or OR with 95% CIs, or sufficient information to calculate these estimates. Reference lists from selected studies and relevant reviews were checked to find additional articles. Study selection is described in online supplementary file 2.
The following information was extracted from each paper: study design, location and study period, number of subjects and response rate, definition of exposure and of reference group, assessment of exposure, definition and ascertainment of outcome, potentially confounding factors controlled for each outcome and other occupational exposures taken into account. A validity score of up to 17 was assigned to each study, for each outcome, and a validity score over 12/17 was considered high. The scoring system, described in online supplementary file 3, results from an adapted version of the analysis grid of scientific articles developed by the Maternity and Work Scientific Group of the Institut national de santé publique du Québec, a public health expertise and reference centre.
Risks of bias of individual studies were assessed for the following six aspects of validity (the conditions required for the risk of bias to be low are indicated in parentheses): external validity or transferability to the population of a developed country like Canada (study population from a developed country and study period after 1975), participation (response rate ≥80%), timing of exposure assessment (before delivery or prospective), definition of exposure (quantitative definition of weight and frequency), control of personal factors (adequate as explained below) and control of some other occupational exposures (control of some other occupational exposures including standing).
Among the known risk factors for each adverse pregnancy outcome,13 factors with an important confounding potential were selected based on their prevalence, the strength of their association with adverse pregnancy outcome and the plausibility that they could be associated with occupational exposure without being part of the causal pathway between exposure and adverse outcome. Four factors were selected for all outcomes: maternal age, obstetric history, smoking and some chronic maternal diseases (eg, diabetes, high blood pressure). Socioeconomic level and height or weight of the mother before pregnancy were also retained for PTD and SGA. Alcohol or drug consumption was retained for SA and SGA, and maternal febrile disease was added for SA. Control of personal factors was considered adequate if at least four factors were controlled for SA, and if at least five factors were controlled for PTD or SGA, including smoking for SGA.
The study selection procedure, the quality assessment and the extraction of results were carried out independently by the author and a collaborator for a random sample of 10% of the studies, and the differences were resolved by consensus. For study selection, the agreement was 100%. For the quality assessment of nine items, the agreement was 83% and the consensus supported the author’s opinion at 94%. For the extraction of results, the agreement was 92% and the consensus supported the author’s opinion at 100%.
Synthesis of results
To categorise exposure, encountered definitions were first divided according to the weight of loads. On the one hand, heavy loads or loads weighing at least 10 kg per lift and, on the other hand, light loads or less than 10 kg. In some definitions, most loads were under 10 kg (eg, <12 kg) and they were placed with light loads (or <10 kg) to form category 1. In some definitions, if the weight of loads was not mentioned and so could be less or greater than 10 kg (eg, >5 kg), they were placed in category 2. Then, definitions not already classified in categories 1 or 2 were divided according to the lifting frequency. This step only applies to heavy loads because the frequency was generally undocumented for non-heavy loads. When heavy loads are rarely lifted or lifted no more than 10 times a day, they form category 3 and when they are often lifted or lifted at least 10 times a day, they form category 5. Finally, when the frequency of heavy lifting was not mentioned and so could be less or greater than 10 times a day (eg, ≥1 time a day), they were placed in category 4. Thus, any definition that can be classified in more than one category will preferably be classified in the more specific categories 1, 3 or 5.
For each outcome, studies’ risk estimates (RE) (OR or RR, but most often OR) were distributed into those five exposure categories; some studies provided several REs corresponding to various categories.
For each adverse pregnancy outcome and each exposure category, a meta-analysis was performed using Comprehensive Meta-Analysis software (V.3.3.070) to obtain summary RE and 95% CI using a random effect model.14 Six potential sources of bias were considered in sensitivity analyses. The six subsets of studies selected for the sensitivity analyses were studies with population from developed countries and study period after 1975; response rate ≥80%; prospective assessment of exposure; quantitative definition of lifting for weight and frequency; adequate control of personal factors and adequate control of other occupational exposures. In each meta-analysis, a summary RE was obtained from all studies and the final result was obtained from the subset of studies with a validity score greater than 12.
Statistical heterogeneity was measured with I2 test and p-value from Q test.15 16 Comprehensive Meta-Analysis software (V.3.3.070) was also used to assess the possibility of publication bias by generating funnel plots and performing Trim and Fill analysis.17 18
Strength of evidence
For each exposure–outcome combination, the strength of evidence indicates the level of confidence in the final result, whether or not a positive association was observed. The strength of evidence depends on the assessment of three aspects relating to the final result: statistical precision, sensitivity analyses and consistency.
In the case of positive association (final result ≥1.05), statistical precision is based on the lower limit of 95% CI with the following interpretation: >1.00: good (statistically significant, p-value<0.05); 0.91–1.00: moderate; ≤0.90: low. In the case of no association (final result <1.05), statistical precision is based on the upper limit of 95% CI with the following interpretation: <1.05: good; 1.05–1.14: moderate; ≥1.15: low. If the upper limit of 95% CI is less than 1.05, it is reasonable to think that a true positive association is unlikely.
Assessment of sensitivity analyses is based on the number of results from sensitivity analyses that are similar to the final result from high validity studies with the following interpretation: 6: good; 3–5: moderate; 0–2: low.
Assessment of consistency is based on the results of heterogeneity tests (I2 and p-value from Q test) performed from high validity studies if they are several and, if not, from all studies. Consistency is considered good when I2 <30% and p-value≥0.10, and considered low when I2 >50% or p-value<0.10.
The strength of evidence is attributed as follows: good if the assessment of the three aspects is good; moderate if the assessment of two aspects is good but moderate for the third; weak if the assessment of one aspect is low; inconclusive if the assessment of more than one aspect is low.
The literature search (online supplementary file 1) identified 2156 records after excluding duplicates. The flowchart that summarises the study selection with reasons for exclusions and references of excluded studies is available in online supplementary file 2. A total of 51 reports (42 studies) were selected from 117 full-text articles assessed for eligibility.
Main characteristics of the 51 included reports are listed in online supplementary file 4 with complete reference list (online supplementary file 5). For each outcome, the results were analysed according to five categories of exposure: (1) light lifting or <12 kg, (2) lifting unspecified weight or ≥5 kg, (3) heavy lifting or ≥10 kg, ≤10 times per day or rarely, (4) heavy lifting or ≥10 kg at unspecified frequency or ≥1 time per day, (5) heavy lifting or ≥10 kg, ≥10 times per day or often.
For SA, PTD and SGA, respectively, the main methodological features, the validity score and the REs obtained by categories of exposure are summarised in table 1 and in online supplementary files 6 and 7.
The relation between occupational lifting and SA was examined in 22 studies, 20 of those are included in the meta-analysis. Their 37 REs were divided into five exposure categories. Results of the five meta-analyses performed for SA, one for each exposure category, are summarised in table 2. The final results from high validity studies show a statistically significant positive association for the fifth exposure category and non-significant slight associations for categories 2, 3 and 4.
For exposure category 5, ‘heavy lifting or ≥10 kg, ≥10 times per day or often’ and SA, a summary RE of 1.28 (95% CI 1.00 to 1.64) results from the seven RE19–24 of this combination (Forest plot in online supplementary file 8), and the final result from the two studies with high validity score is 1.31 (95% CI 1.17 to 1.47) (table 2). Here, two studies report SA RE for different periods of pregnancy. Juhl’s21 results for early and late SA (table 1) are similar to the final result, while the results of McDonald22 are slightly higher for SA at 10–15 and 16–27 weeks of pregnancy (table 1, footnote ‘†††’). According to p-value and 95% CI, final result is statistically significant and heterogeneity tests (Q=0.61, p=0.43, I2=0%) reveal good consistency. Moreover, the results of six sensitivity analyses are similar (range 1.20–1.36) to the final result from higher quality studies (table 2). And although funnel plot indicates the possibility of a slight publication bias, the correction by the trim and fill method only slightly affects the result (online supplementary file 9). These analyses indicate a positive association with good strength of evidence (table 3).
For exposure category 1, RE from the only study having a high validity score is 0.97 (95% CI 0.53 to 1.78) showing a low statistical precision. The results of five sensitivity analyses are similar to the RE of this study (table 2). Heterogeneity tests, from all studies, reveal good consistency. Therefore, analyses indicate weak evidence of no association (table 3).
For categories 2, 3 and 4, final results show slight associations with a low statistical precision. The assessment of sensitivity analyses is good for categories 2 and 4, and moderate for category 3. Heterogeneity tests reveal that consistency is good for category 2, but low for categories 3 and 4 (table 2). Results indicate weak evidence of positive association for category 2 and are inconclusive for categories 3 and 4 (table 3).
In some reports, the effect of lifting was analysed according to the total burden lifted per day21 25 26 and those results are not included in the meta-analysis. Data used by two of them21 26 come from the Danish National Birth Cohort. In one report, a job-exposure matrix was used to estimate exposure and no exposure–response relationship was found between lifting and SA risk,26 while results of the other report show significant trends for increased SA risk with total burden lifted per day, starting from a daily load of 101 kg for early SA (<13 weeks) and from a daily load of 201 kg for late SA (13–21 weeks).21 Otherwise, an OR of 1.9 was obtained for a daily lifting score of at least 30 points, which could be, for example, 20 kg raised six times a day or 5 kg raised 30 times.25
The relation between occupational lifting and PTD was examined in 28 studies, 27 of which are included in the meta-analysis. In almost all, PTD is defined as a live birth before 37 weeks of gestation, while one study defines PTD as a live birth before 36 weeks of gestation27 and another as live birth from 22 to 36 weeks of gestation.28 Results of the five meta-analyses performed for PTD, one for each exposure category, are summarised in table 4. The final result from high validity studies shows a statistically significant positive association for the fifth exposure category.
For exposure category 5, ‘heavy lifting or ≥10 kg, ≥10 times per day or often’ and PTD, a summary RE of 1.25 (95% CI 1.10 to 1.43) results from the 13 studies19 29–40 of this combination (Forest plot in online supplementary file 10) and the final result from the seven studies with a high validity score is 1.24 (95% CI 1.07 to 1.43). These results are statistically significant, heterogeneity tests for the final result (Q=0.91, p=0.99, I2=0%) reveal good consistency and results of the six sensitivity analyses are similar (range 1.21–1.25) to the final result from higher quality studies (table 4). Funnel plot and trim and fill analysis indicate that a publication bias is unlikely (online supplementary file 11). These analyses indicate a positive association with good strength of evidence (table 3).
For exposure categories 1, 2 and 3, neither the summary REs from all studies, nor the final results show a positive association. The upper limit of 95% CI signals low statistical precision for categories 1 and 2, but a moderate statistical precision for category 3. For categories 1 and 3, the results of the six sensitivity analyses are similar to the final result, but the assessment of sensitivity analyses is moderate for category 2. Heterogeneity tests reveal good consistency for the three categories (table 4). Therefore, analyses indicate no association and the strength of evidence is weak for categories 1 and 2, and moderate for category 3 (table 3).
For category 4, a final result of 1.08 (95% CI 0.90 to 1.29) is obtained from studies with a high validity score. The lower limit of 95% CI (0.90) signals a low statistical precision, the assessment of sensitivity analyses is moderate and heterogeneity tests reveal moderate consistency (table 4). Therefore, the strength of evidence is weak (table 3).
In some reports, the effect of lifting was analysed according to the total burden lifted per day26 37 and those results are not included in the meta-analysis. Data used by both reports come from the Danish National Birth Cohort. Results show significant trends for increased PTD risk with total burden lifted per day, starting from a daily load around 101 kg.26 37
The relation between occupational lifting and SGA was examined in 15 studies. Results of the five meta-analyses performed for SGA are summarised in table 5. The final results, from high validity studies, show no positive association with the risk of SGA for any exposure category.
Except for category 3, the statistical precision is low. The assessment of sensitivity analyses is low for categories 3 and 4. Heterogeneity tests reveal that consistency is good for all categories except for category 2 (table 5). Therefore, analyses indicate no association and the strength of evidence is weak for exposure categories 1, 3 and 5, while results are inconclusive for categories 2 and 4 (table 3).
This systematic review considered the relationship of lifting at work with three adverse pregnancy outcomes (SA, PTD and SGA). Given the variety of definitions of lifting encountered in studies, definitions were grouped into five categories: (1) light lifting or <12 kg, (2) lifting unspecified weight or ≥5 kg, (3) heavy lifting or ≥10 kg, ≤10 times per day or rarely, (4) heavy lifting or ≥10 kg at unspecified frequency or ≥1 time per day and (5) heavy lifting or ≥10 kg,≥10 times per day or often.
This systematic review has the merit of establishing the strength of evidence from explicit criteria. For each exposure–outcome combination, whether there is a positive association or not, the strength of evidence makes it possible to estimate the degree of confidence that one can have towards the final result.
The results of this systematic review show that exposure category 5 (heavy lifting or ≥10 kg, ≥10 times per day or often) is associated with increased risk of SA and PTD with a good strength of evidence, but not with SGA. There is also weak evidence of associations between exposure category 2 (lifting unspecified weight or ≥5 kg) and SA, and between exposure category 4 (heavy lifting or ≥10 kg, at unspecified frequency or ≥1 time per day) and PTD. For all other exposure–outcome combinations, there is either no association or the data are inconclusive.
The inclusion of definitions of lifting where loads could reach 12 kg in category 1 does not seem to have affected the SREs, the results from the two studies concerned19 41 being comparable to the other REs of category 1 (table 1 and online supplementary files 6 and 7). As in categories 2 and 4, some workers can lift heavy loads frequently. It is plausible that the two weak evidence associations (SA with category 2 and PTD with category 4) could be explained by the overlap of categories 2 and 4 by category 5. The lack of high-quality studies especially in some SA and SGA exposure–outcome combinations may have limited the ability to conclude. Thus, for three of the four combinations with only one high quality study, data were inconclusive.
Physically demanding work like heavy lifting can affect pregnancy outcome through several mechanisms. Reduced uterine blood flow19 42–44 can affect fetal growth and was suggested as a potential cause of SA42 43 such as increased intra-abdominal pressure.42 43 45 Increased catecholamines5 8 37 44 and intra-abdominal pressure19 42 43 46–48 may increase uterine contractility5 8 19 37 44 46 47 and risk of PTD,5 8 37 especially in the last trimester.47 48
A weakness of this systematic review is that only 10% of studies were selected and evaluated in duplicate. However, the agreement was very high. Moreover, in the case of disagreement, the consensus supported the author’s opinion, to 94% for the quality assessment and to 100% for the extraction of results.
As the studies used in this review are observational, different methodological weaknesses may have biassed the results. For that reason, the summary REs from higher validity studies are chosen as final results and, for each exposure–outcome combination, sensitivity analyses were conducted to estimate the impact of the six types of methodological weakness described below.
(1) As women’s working conditions are likely to be more difficult in developing countries, a lack of external validity could limit the extrapolation of the results, but here most studies were conducted in developed countries after 1975. (2) Low participation of subjects (<80% or unknown) may have led to selection bias. (3) Studies having assessed exposure retrospectively may have been subject to recall bias. Indeed, if workers with an adverse pregnancy outcome overestimated their exposure, an inflationary bias could occur. (4) Although the heterogeneity of exposure definitions has been partially taken into account by grouping definitions into five categories, misclassification of exposure is more likely when the definition is qualitative. (5) A confounding bias is possible when the control of potentially confounding personal factors is not adequate. (6) Lifting at work is often associated with other occupational constraints (eg, standing) possibly associated with adverse pregnancy outcomes and, if studies did not adjust for other occupational exposures, inflationary confusion bias could occur.
For each methodological weakness described above, a sensitivity analysis was carried out with the studies free from this methodological weakness, and the result obtained was compared with the final result from higher validity studies. Even though all the higher validity studies are not free from a particular methodological weakness, the similarity of their summary RE to the result of the sensitivity analysis indicates that this methodological weakness does not seem to have biased their result.
Another source of bias could arise from variability in definitions of adverse pregnancy outcome. Comparison of the results obtained for early SA (<13 and <10 weeks) with those obtained for later SA (13–21, 10–15 and 16–27 weeks) does not allow any trend to be observed. In most studies, SGA was defined as birth weight <10th centile of weight for gestational age and gender, but some studies used different definitions that also included a measure of birth weight taking into account the duration of pregnancy,19 27 39 49 50 while the definition was not specified in one study.34 If these six studies were excluded, the summary REs from higher validity studies would be unchanged for exposure categories 1, 3 and 4, and the meta-analysis could not be done for exposure category 2 because there would be only one study in this category. And, for exposure category 5 ‘heavy lifting or ≥10 kg, ≥10 times per day or often’, the summary RE would have been 1.06 (95% CI 0.89 to 1.25) with a weak strength of evidence instead of 1.03 (95% CI 0.87 to 1.21). These studies, however, are kept in the analyses because their definitions correspond to the concept of intrauterine growth retardation.
It would have been interesting to analyse the specific effect of the third trimester exposure on PTD and SGA. Unfortunately, very few studies have measured third trimester exposure39 50 51 and if exposure is measured earlier, it is uncertain whether the workers were still exposed in the third trimester or not. Furthermore, some reported that the most highly exposed workers more often benefited from improvements in their working conditions, or more often stopped working before the third trimester.2 19 34 35 38 40 51–53 When such changes in exposure are not measured or accounted for, the association may be underestimated because the actual exposure is lower than the alleged exposure. However, the extent of this error is difficult to assess because information on exposure changes was rarely available in reports and was rarely taken into account in the analysis.2 51 53
Other recent systematic reviews have estimated the effect of lifting at work on pregnancy. In the review of Bonzini et al, 6 updated by Palmer et al, 8 definitions of exposure were considered too heterogeneous to be combined. However, Palmer et al report that, among the larger and higher quality studies, the median values of RR were 1.02 for PTD and 1.08 for SGA.8 To estimate the association between SA and lifting, Bonde et al combined results of studies that provided RE for pregnant women lifting at least 100 kg (product of weight and frequency) per working day. Pooled OR was 1.02 (95% CI 0.73 to 1.44) for higher methodological quality studies.7 Finally, van Beukering et al pooled PTD risks of pregnant working women who lift more than 5 kg. The summary ORs are 1.29 (95% CI 1.05 to 1.57) for all studies, and 1.24 (95% CI 0.96 to 1.61) for studies that met their criteria for low risk of bias.9
Despite differences in exposure classification and inclusion of studies published up to 2019 rather than up to 2011–2012, our findings for exposure category 5 and PTD are similar to those of van Beukering et al, and our findings for exposure categories 1–4 and PTD, are similar to those of Palmer et al. On the other hand, Palmer’s conclusion differs from our result for category 5 and PTD. This difference could be explained by the inclusion of several REs corresponding to categories 1–4 in the Palmer review and by the inclusion of recent high quality studies in our review, such as Runge.37 Bonde’s conclusion differs from our result for category 5 and SA. This difference could be explained by the inclusion of more recent studies in our review, such as Juhl.21 And, our results differ slightly from Palmer’s conclusion for SGA. Some disparities in the distribution of studies according to whether their quality is high or not may also contribute to the differences observed, especially for SGA here.
The results of this systematic review highlight positive associations between ‘heavy lifting or ≥10 kg, ≥10 times per day or often’ and risks of SA and PTD. Summary REs from high validity studies are respectively 1.31 (95% CI 1.17 to 1.47) and 1.24 (95% CI 1.07 to 1.43) for SA and PTD. In both cases, results are statistically significant, consistency is good and the results of the six sensitivity analyses are similar to the final result. Therefore, the strength of evidence is good, indicating a good level of confidence in both associations. On the other hand, the results are reassuring for lower exposure levels and for SGA.
The observation of increased risks of SA and PTD in the presence of ‘heavy lifting or ≥10 kg, ≥10 times per day or often’ indicates that preventive measures must aim at the avoidance of this level of exposure early enough in pregnancy to effectively prevent SA and PTD.
The author would like to thank Stéphane Caron, M.D., (INSPQ) for his participation in the selection procedure and quality assessment and extraction of results for a random sample of 10% of the studies and for his support and advice. The author would also like to thank the other members of the Maternity and Work Scientific Group (Lise Goulet, M.D., Ph.D., Mylène Trottier, M.D., M.Sc., Marie-Pascale Sassine, M.Sc.), INSPQ, for their support and advice.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Contributors AC developed the methodology for the systematic review. AC selected studies, extracted data and assessed the methodological quality of studies. AC conducted the meta-analysis and wrote the manuscript.
Funding The Institut national de santé publique du Québec (INSPQ), a public health expertise and reference centre, has made possible the realisation of this work.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.