Article Text

Download PDFPDF

Original article
Job-exposure matrix for the assessment of alkylphenolic compounds
  1. Mayte Martín-Bustamante1,
  2. Anna Oliete-Canela1,
  3. Marta Diéguez-Rodríguez1,
  4. Yolanda Benavente2,3,
  5. Delphine Casabonne2,3,
  6. Juan Alguacil4,
  7. Manolis Kogevinas3,5,6,7,
  8. Silvia de Sanjosé2,3,8,
  9. Laura Costas2,3,8
  1. 1Institut de Seguretat i Salut Laboral, Departament de Treball, Afers Socials i Famílies, Generalitat de Catalunya, Barcelona, Spain
  2. 2Unit of Infections and Cancer, Cancer Epidemiology Research Programme, IDIBELL, Catalan Institute of Oncology, Barcelona, Spain
  3. 3CIBER Epidemiologia y Salud Pública (CIBERESP), Madrid, Spain
  4. 4Department of Environmental Biology and Public Health, Huelva University, Huelva, Spain
  5. 5Centre for Research in Environmental Epidemiology (CREAL), Barcelona, Spain
  6. 6Universitat Pompeu Fabra (UPF), Barcelona, Spain
  7. 7National School of Public Health, Athens, Greece
  8. 8Department of Medicine, University of Barcelona, Barcelona, Spain
  1. Correspondence to Dr Laura Costas, Unit of Infections and Cancer, Cancer Epidemiology Research Programme, Institut Català d'Oncologia, Av. Gran Via de l'Hospitalet 199-203, L'Hospitalet de Llobregat, Barcelona 08908, Spain; lcostas{at}


Objectives Our aim was to develop a job-exposure matrix (JEM) to assess occupational exposure to alkylphenolic compounds in epidemiological research, considering changes in their use over time, and including exposure probabilities in the assessments.

Methods We consulted multiple sources of information, and performed interviews with 9 key people from industry and academia. 3 hygienists coded frequency (minority or majority of workers involved) and intensity of exposure (including dispersive processes, with shaking, or aerosol generation, or otherwise) to alkylphenolic compounds for all the 390 International Standard Classification of Occupations (ISCO)-88 job titles by period of time. Intensity and frequency of exposure were combined in a single score as follows: unlikely=0, occasionally+low intensity=1, occasionally+high intensity=2, frequent+low intensity=2, and frequent+high intensity=3.

Results We identified 54 (13.8%) of the 390 ISCO-88 job titles with potential exposure to alkylphenolic compounds. In 6 of jobs deemed as exposed, exposure depended on the economic sector of the occupation. Nonylphenol ethoxylates were the compounds most commonly involved (30 job titles, 55.6% of the exposed). Variations in alkylphenolic compounds use varied greatly over time; while they are still used in the plastic and rubber industry, in domestic cleaning agents their use began to decline before 1995.

Conclusions We built a JEM to assess exposure to alkylphenolic compounds, taking into account changes in use over time, different types of alkylphenolic compounds and different scenarios of exposure, which can be a valuable tool for exposure assessment in epidemiological research on the health effects of these chemicals.

  • alkylphenols
  • nonylphenol
  • octylphenol
  • ethoxylates
  • job-exposure matrix

Statistics from

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.

What this paper adds

  • Alkylphenolic compounds are ubiquitous environmental pollutants with endocrine disrupting properties used in a wide range of industrial applications.

  • Epidemiological studies are hampered by limitations on the available tests, which are not sensitive enough and too susceptible to contamination, and previous JEMs did not consider relevant changes in use over time.

  • A new JEM was developed considering time windows of exposure and different types of alkylphenolic compounds.

  • The present JEM could be a valuable tool for exposure assessment in epidemiological research on the health effects of these compounds.


Alkylphenols are organic chemicals obtained through alkylation of phenols, produced generally for the manufacture of alkylphenolic ethoxylates. Alkylphenolic ethoxylates are composed of a polyethoxy chain and an alkyl radical connected to a phenolic ring, and they are mainly used as non-ionic surfactants, as well as in a wide range of applications. There are numerous types of alkylphenolic compounds, and nonylphenol and octylphenol and their ethoxylates are the most common ones.1 Butylphenol, decylphenol and dodecylphenol are other alkylphenols less frequently used in the industry. Uses of alkylphenolic compounds are summarised in table 1. Occupational exposure can take place during their production or with exposure to detergents, specialty paints, pesticides, cosmetics and hair dyes among others.

Table 1

Types of alkylphenolic compounds and their industrial uses

Owing to the toxicity and bioaccumulation of alkylphenolic compounds in marine organisms, the European Union (EU) limited the marketing of nonylphenol and nonylphenol ethoxylates in 2003.2 Indeed, long-chain alkylphenolic ethoxylates can degrade to metabolites like short-chain alkylphenolic ethoxylates and alkylphenols, which can bioaccumulate in soils, plants and animals.3 Alkylphenols are ubiquitous, and they have been detected in rivers and bottled water,4–6 as well as in human fluids or tissues, such as urine,7 blood,8 placenta,9 breast milk10 and adipose tissue.11 These compounds have the capacity to interact with the endocrine system, especially in aquatic organisms and among humans. Alkylphenols and short-chain alkylphenolic ethoxylates are considered endocrine disruptors in animals and humans, mainly because of their oestrogenic effects mediated by oestrogen receptors.12–15 In 2012, the WHO expressed concern about the potential effects of exposure to endocrine disruptors, such as female reproductive disorders and hormonal cancers, and considered a research priority to improve the assessment of exposure to these chemicals.16

Given the great variations of use of alkylphenols over time in industry, the study of the potential effects of alkylphenolic compounds on chronic diseases requires development of retrospective exposure assessment tools. The half-lives of these compounds in blood are likely to be short,17 ,18 and their detection most likely reflect recent exposures. Available tests measuring alkylphenolic compounds in sera or plasma are generally not sensitive enough and often too susceptible to contamination.19

Job-exposure matrices (JEM) are a widely used tool for retrospective exposure assessments in occupational studies, and they have the benefit of being more efficient than expert assessment.20 Although JEM estimates may be affected by misclassification of exposure,21 they perform better than self-reported occupational exposures in large population-based studies.22 JEMs are based on job titles, which suffer less from differential recall bias than self-reported chemical exposures.

We aimed to develop a comprehensive JEM for estimating lifetime exposure to alkylphenolic compounds considering relevant changes in their use over time. The JEM could then be used as a tool for exposure assessment in epidemiological research on the health effects of these compounds regardless of the latency period of the disease under study.


We originally developed this JEM to apply it to the MCC-Spain study,23 which included more than 10 000 participants in Spain. Population controls and cases of breast, prostate, stomach, colon cancer and chronic lymphocytic leukaemia were enrolled and they were asked for their occupational histories, among others. Lifetime occupational history was assessed for all jobs held for at least 1 year, and in total 26 149 job registries were registered in this study. Occupational histories in this study spanned a period of time from 1931 to 2014. Each occupation was coded with the national codes of occupation Clasificación Nacional de Ocupaciones (CNO)-94, which is the Spanish adaptation of the International Standard Classification of Occupations (ISCO)-88.

To identify potential occupational uses of alkylphenolic compounds, different sources of information were consulted, including PubMed, official webpages from different governments, including safety and health at work agencies from local governments, doctoral dissertation databases, databases on importation and usage from the Spanish Tax Agency, newspaper libraries, web pages from international organisations (the WHO, Occupational Safety and Health Administration Europe). The main key words for search were alkylphenol, nonylphenol, octylphenol, butylphenol, dodecylphenol, surfactant, tensioactive, occupation, job-exposure matrix, endocrine disruptor, resin, formaldehyde and epoxy resin.

Three industrial hygienists independently coded exposure to alkylphenolic compounds in a first step for each of the CNO-94 occupational codes, as follows: (1) ever exposure, (2) never exposure and (3) insufficient information. In 29.8% of occasions, the three hygienists did not agree on the ever/never classification, or at least one of them considered that there was insufficient information to classify exposure. This first classification helps the hygienists in order to obtain further information to fill these knowledge gaps. For instance, there were disagreements or insufficient information in regard to specific uses in the livestock and pharmaceutical industry, the industrial processes in certain sectors—involving closed methods or not—changes in work practices over time, as well as specific questions about the dates that use changed due to potential industry voluntary restrictions. This specific information was searched on specialised libraries (Chemist Faculty of Barcelona University, Science library from Autonomous University of Barcelona, Polytechnic University of Catalonia—Terrassa Campus, Spanish National Institute for Safety and Health at Work, and Statistical Institute of Catalonia), newspaper libraries, doctoral dissertation databases, and databases on importation and usage from the Spanish Tax Agency.

Furthermore, nine key people from five industries in Spain using alkylphenolic compounds in sectors such as textile, detergent, paints, plastic additives and cosmetic industries, and from universities and other agencies were contacted and personally interviewed. They were asked for usage of alkylphenolic compounds over time in the specific areas where they had worked. All relevant information on the industrial uses of alkylphenolic compounds obtained in each interview and literature review was gathered and summarised and this reference material served as a tool for the expert assessment. An example to illustrate the flow of information and consequent decisions is the following case from the textile industry. The Spanish Ministry of Environment24 published a document specifying that the use of alkylphenol ethoxylates was still common in the textile industry in 2004 (after the European regulation). This was further confirmed by the consulted key expert in this sector. This information, along with the decline in importation of alkylphenolic compounds over time observed using other sources, such as the Spanish Tax Agency, helped the hygienists assign a specific key time point in this sector. The interviews also helped understand specific changes in work practices over time. For example, closed-system methods were encouraged from 2000 onwards in the phenolic resins industry in order to decrease exposure to formaldehyde. Consequently, workers' exposure to alkylphenolic compounds also decreased in this specific industry and period of time.

A temporal axis was created taking into account the calendar years that alkylphenolic compounds were used for each of the occupations. The exposure probability scores were assigned by means of consensus discussions by the three hygienists. Two external hygienists, who previously reviewed all reference material from the literature review and interviews, were also invited to the discussion meetings to help the three main hygienists assign the scores. The intensity of exposure for each occupation and each period of time was coded as high or low, and the frequency as occasionally or frequent. Owing to the low volatility of alkylphenolic compounds, in general we considered that the intensity of exposure was ‘low’ if it involved a non-dispersive process, a generally closed production process, with manipulation of alkylphenolic compounds in a solid status, at ambient temperature, or other matrices with difficult migrations. We considered that the intensity of exposure was ‘high’ in dispersive processes, with shaking or stirring, aerosols generation, at high temperatures, or in a liquid status. Frequency of exposure was coded as ‘occasional’ if only a minority of workers is usually involved in tasks with alkylphenolic compounds, in very specific tasks and a short time of exposure compared with the annual amount of work, and ‘frequent’ otherwise. We combined intensity and frequency of exposure in a single score as follows: unlikely=0, occasionally+low intensity=1, occasionally+high intensity=2, frequent+low intensity=2, and frequent+high intensity=3. We identified those job titles in which exposure depended on the economic sector. We used the Statistical Classification of Economic Activities in the European Community (NACE) to list these economic sectors.

We considered two routes of exposure (dermal and inhalation) with no distinction between them. The main types of alkylphenolic compounds used were identified (octylphenol, nonylphenol and nonylphenol ethoxylates) for some occupations, while others involved a mixture of alkylphenols or alkylphenolic ethoxylates. We grouped the job codes under the following scenarios of exposure: (a) manufacture and use of plastic and rubber products, (b) use of industrial tensioactives, (c) manufacture and use of paints and lubricants, (d) use of domestic tensioactives, (e) use of cosmetic and hair products, and personal hygiene products and (f) use of pesticides. The JEM was finally translated to ISCO-88, which is more commonly used in other countries than Spain.


Several key time points in alkylphenolic compounds use were identified in order to define time windows of exposure. These time points were related to the first estimated use of alkylphenolic compounds, the internationalisation of producing companies, the first concerns on their effects, the voluntary agreements on reduction or elimination of alkylphenolic compounds in domestic and industrial detergents, and the EU regulations on nonylphenol (table 2). However, use greatly varied over time by scenario; for example, decreases in alkylphenolic compounds use on domestic detergents occurred in 1995 voluntarily in Spain in industries after a European agreement,25 while alkylphenolic compounds are still used in the production of plastics and rubber.

Table 2

Key chronological events identified in relation to the use of alkylphenolic compounds in Europe

Out of 390 job titles, we identified 54 potentially exposed to alkylphenolic compounds (13.8%; see online supplementary table S1). In six of these exposed titles, exposure was dependent on the economic sector of the occupation. Those economic sectors with potential exposure to alkylphenolic compounds among these six job titles are specified in online supplementary table S2. These are related to manufacture of textiles, leather, paper, chemicals (including dyes and pigments, plastics, synthetic rubber, pesticides, paints, varnishes, printing ink and mastics, soap and detergents, cleaning and polishing preparations, and perfumes), and pharmaceutical preparations. For five job titles (1.3%), there was not enough information to assign a level of exposure, and they were treated as unknown exposure level.

Supplementary table

Job titles classified as exposed to alkylphenolic compounds.

Supplementary table

Sectors of economic activities to be considered in the assignment of exposure in selected job titles.

Of the 54 job titles classified as exposed, 30 were classified as mainly exposed to nonylphenol ethoxylates (55.6% of the exposed), 12 as exposed to a mixture of alkylphenolic compounds (22.2%), 10 as exposed to alkylphenolic ethoxylates (18.5%), and 2 to octylphenol or octylphenol+nonylphenol (3.7%; table 3). Manufacture and use of paints and lubricants was the scenario with more job titles involved (20 job titles), followed by use of industrial tensioactives and use of pesticides (10 job titles each), manufacture and use of plastic and rubber products (6 job titles), and domestic tensioactives (5 job titles), while cosmetics was the scenario with less job titles involved (3 job titles).

Table 3

Compounds and scenarios involved in the JEM

The structure of the JEM is shown in table 4. Briefly, one axis contains all 390 ISCO-88 titles, and the other contains the exposure scores for all the time periods, the scenarios and compounds involved, as well as a column to identify those titles with exposure dependent on the economic sector. ‘Fibre-preparing-, spinning- and winding-machine operators’ is a job title classified under the industrial tensioactives scenario, and coded as frequent and high intensity of exposure from 1960 to 1998. From 1999, alkylphenolic compounds use in this specific occupation became less frequent and probable, and from 2008 it is coded as unlikely. The job title ‘Rubber-products machine operators’ from the plastic and rubber scenario is coded as unlikely before 1979, frequent and high intensity of exposure from 1979 to 2000, while the exposure is coded as probable or frequent from 2001 onwards. ‘Computer equipment operators’ is an example of a job title for which exposure to alkylphenolic compounds is classified as unlikely. In the job title ‘Chemical engineering technicians’, exposure is considered depending on the industry sector of economic activities involved (see online supplementary table S2 to see the sectors to be considered in the assignment of exposure for these selected job titles).

Table 4

Examples of entries in the job-exposure matrix


We constructed a new JEM for estimating occupational exposure to alkylphenolic compounds in epidemiological studies. We considered changes in their use over time, included exposure probabilities in the assessments, and refined some of the scores of specific job titles by using an industrial–occupational coding, which altogether allow the assignment of more accurate exposure estimates. Alkylphenolic compounds are ubiquitous compounds commonly detected in food, drinking water,4–6 ,8 as well as in human fluids or tissues.7 ,8 ,10 ,11 Alkylphenols, such as nonylphenol, mainly derive from the environmental degradation of alkylphenolic ethoxylates. Alkylphenolic ethoxylates are used as detergents, cleaners, emulsifiers and a variety of other applications such as plastics and pesticides. They were regulated in 2003 in the EU due to their large toxicity on aquatic organisms due to endocrine disruption. The effects of these compounds on human health are mostly unknown. Literature on the subject is scarce and limitations exist due to difficulties in measuring alkylphenolic compounds in epidemiological studies. Occupational studies represent an opportunity to explore the role of these compounds.

Fifty-four (13.8%) job titles out of a total number of 390 listed in ISCO-88 were considered to be exposed to alkylphenolic compounds in the present JEM. Windows of exposure, which reflect changes in production and use over time, varied greatly depending on the considered scenario of exposure. For instance, alkylphenolic compounds are still used in the plastic and rubber industry, as well as in special paints, nowadays. In contrast, job titles under the domestic tensioactives scenario were coded as unlikely from 2004 onwards, due to voluntary restrictions in its use from the industry and the subsequent EU regulation on nonylphenol in 2003.2 Nonylphenol ethoxylates were the compounds most often linked to an occupation in the JEM; however, only two occupations (both related to the plastic and rubber industry) scored as exposed to octylphenol or nonylphenol. This low number of titles may imply a low prevalence of exposure, and thus a power issue, when evaluating exposure to octylphenol/nonylphenol. Twelve other job titles involving alkylphenols actually involved a mixture of alkylphenols and alkylphenolic ethoxylates, and therefore we did not assign a more specific category of compound. These 12 occupations were in general related to the manufacture of alkylphenolic compounds and to the use of plastic and rubber products.

For simplicity, we grouped job titles by scenarios, which had similar windows of exposure. However, for more accuracy, the assignment was done independently for each of the job titles, therefore allowing for certain variations on the scores within the scenario. The inclusion of scenarios allows stratified analyses as sensitivity analyses to account for other chemicals that can occur concomitantly to alkylphenolic compounds. Scenarios may help interpret results in regard to the potential overlap with exposure to other agents. For example, when evaluating associations with alkylphenolic compounds in epidemiological studies using this JEM, positive associations observed in the scenario ‘Use of pesticides’ but not in the rest of the scenarios may imply that associations could be driven by pesticides but not by alkylphenolic compounds.

There is a scarcity of epidemiological studies evaluating the effect of these compounds in human health. To the best of our knowledge, alkylphenolic compounds have been linked to certain maternal exposures leading to health outcomes in their offspring,26–29 congenital heart defects,30 ,31 male infertility32 and certain types of neoplasms33–36 in epidemiological studies among humans. Studies have usually evaluated exposure to alkylphenols in urine samples26–28 ,32 ,35–37 or through occupational studies using JEMs.29–31 ,33 ,34 The evaluation of alkylphenolic compounds in epidemiological studies is hampered by limitations on the available laboratory tests to assess levels on blood serum or plasma samples, which are generally collected in epidemiological studies. Determination by gas chromatography coupled with mass spectrometric detection is the most commonly applied analytical technique to detect these compounds in biological samples. However, a step to increase volatility (termed derivatization) is usually needed to improve sensitivity, but this step increases the risk of contamination from the environmental background and it is time-consuming. Other techniques, such as hybrid solid phase extraction, are still being developed to test alkylphenol levels in difficult biological matrices and they might be future candidates to help reduce these drawbacks.19 Biopsies of adipose tissue are more reliable matrices to evaluate alkylphenols due to the affinity of these compounds for the lipid fraction.11 Measurements in adipose tissue could better reflect past exposures, although biopsies are less suitable for epidemiological studies due to the invasiveness of this procedure. However, their toxicokinetic behaviour among humans is not completely understood. Müller et al17 assessed the kinetics of nonylphenol after oral and intravenous exposures among two healthy volunteers. The authors observed that the half-life of nonylphenol in blood was short (2–3 hours) after oral or intravenous application, in accordance with studies among rats,38 ,39 and that only 11.5% of the oral dose was recovered in the urine and faeces during the study. This could mean that the remainder may be exhaled or taken up by the lipid compartment, but the literature is too scarce to draw conclusions on the toxicokinetics of nonylphenol as well as of other alkylphenolic compounds.40

We identified two previous matrices evaluating occupational exposure to alkylphenolic compounds. In a publication from the National Institute for Occupational Safety and Health (NIOSH)41 in 1981–1983, the estimated numbers of employees potentially exposed to different agents in their survey were provided by occupation. The JEM for endocrine disruptors by Van Tongeren et al and updated by Brouwers et al21 ,42 provided the probability of exposure to alkylphenols and alkylphenolic ethoxylates in two categories, ‘unlikely’ versus ‘possible’ exposure, as none of the job titles scored as ‘probable’ exposure. Previous JEMs were aimed at assessing recent exposures, and therefore none of them considered changes in use over time, while this has dramatically fluctuated over the years due to the start of commercialisation, the concerns on their impact and consequent regulations.

The NIOSH study was based on the Classification by the Bureau of the Census 1980, and 148 job titles of the 503 (29.4%) were classified as having one or more employees potentially exposed to alkylphenolic compounds. This prevalence of use seem to be higher than that observed in our study, maybe due to the higher sensitivity rather than specificity of the NIOSH method, although comparisons with our JEM are difficult to assess because of the different methodologies used. In the JEM by Brouwers et al, which was based on the UK Standard Occupational Classification 2000 (SOC2000) coding system, 19 (5.4%) job titles out of a total of 353 were classified as exposed, while 13 job titles (3.7%) were classified as unknown. The lower prevalence of exposure in this JEM compared with our JEM is probably due to the fact that it assessed a short period of time (1996–2006) when use of alkylphenolic compounds was lower. For instance, some of the job titles involving cleaning activities were classified as unexposed by Brouwers et al,21 but as exposed in the previous version of this JEM by Van Tongeren et al.42 A specific difference with the JEM by Brouwers et al, which also contributes to our higher prevalence of use, is that we considered that exposure to alkylphenolic compounds was likely among job titles involving lubricants and cutting oils. We considered that exposure was likely in these occupations, such as ‘Motor vehicle mechanics and fitters’, because nonylphenol ethoxylates are used in metal degreasing and cutting oils may contain nonylphenol as an emulsifier.43 ,44 We have applied the present JEM using data from 4098 controls participating in the MCC-Spain study, and observed that 25.1% of this population has been ever exposed to alkylphenolic compounds. This prevalence is higher than that observed in the list of 390 ISCO codes (13.8%), because some exposed job titles (for instance, cleaners) are over-represented in the general population, while some non-exposed titles (such as legislators) are less represented. As well, a same person can have several job titles over time, increasing the probability to be classified as ever exposed.

We included three key features that have been shown to improve the performance of a JEM in order to obtain more accurate exposure estimates, as demonstrated by Dosemeci et al.45 First, we accounted for differences in exposures over time considering periods of predominant use; second, we included both intensity and frequency of exposures in the assessments (instead of a dichotomous exposure score—yes/no); and third, we refined some of the scores of specific job titles by considering the industry sectors of economic activities involved in specific occupations, allowing a better assessment of the exposure than the job title alone in these particular occupations. The main limitation in the construction of this JEM was the limited availability of data, which on occasions was scarce, especially for periods referring to a long time ago. Therefore, we were not able to assign approximate prevalences of exposed workers for each job title by scenario and by time period.

We originally developed this JEM to be applied to the Spanish population, and therefore we mainly considered relevant changes in use over time in Spain. However, it can be applicable to other populations, especially from European countries, modifying specific key time points that may vary by country. In particular, the second key time point in our JEM (1959, Spanish Plan of Stabilization) should not be considered in other countries because these compounds were not available in Spain until the 1960s, when they started to be imported from other countries. Therefore, in other countries, exposure can be considered from the first identified use, typically in the 1940s (eg, 1944 in the UK46). Also, specific variations in alkylphenolic compounds use or limitations by country should be considered. For example, an European agreement was signed in 199225 to limit use of alkylphenolic ethoxylates in cleaning products, but the year of limitation may slightly vary in each specific country. In summary, we developed a JEM to assess retrospective exposure to alkylphenolic compounds considering relevant changes in their use over time that can be a valuable tool for exposure assessment in epidemiological research on the health effects of these endocrine disrupting chemicals.


The authors thank Emilia Molinero-Ruiz, Santos Hernández, Nereida Gras and Ferran Calduch for their invaluable contributions and helpful comments, as well as the interviewed key people from industry and academia for providing essential information for the execution of this work.



  • Twitter Follow Anna Oliete-Canela at @anaolietecanela, Marta Diéguez-Rodríguez at @marta_dieguez, Manolis Kogevinas at @kogevinasm, Sílvia de Sanjosé at @silviadesanjose and Laura Costas at @laurix_c

  • Contributors MM-B, AO-C and MD-R performed the interviews with the key people, and assigned the exposure scores. All the coauthors contributed to the planning of the study. MM-B and LC drafted the article and are the guarantors.

  • Funding The work conducted by LC was supported by grants from the Spanish Ministry of Economy and Competitiveness—Carlos III Institute of Health (Río Hortega CM13/00232 and M-AES MV15/00025). This work was partially supported by public grants from the Spanish Ministry of Economy and Competitiveness—Carlos III Institute of Health (PI11/01810, PI14/01219), the Catalan Government (2014SGR756) and the European Regional Development Fund-ERDF.

  • Competing interests None declared.

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

  • Data sharing statement Copies of the JEM are available as a Stata database and Excel spreadsheet from the corresponding author at