Article Text
PDF
Abstract
Objectives The Seveso accident (1976) caused the contamination with 2,3,7,8-tetrachlorodibenzo-para-dioxin (TCDD) in an area north of Milan, Italy. We report the results of the update of mortality and cancer incidence in the exposed population through 2013.
Methods The study cohort includes subjects living in three contaminated zones with decreasing TCDD soil concentrations (zone A, B and R) and in a surrounding uncontaminated territory (reference). Poisson models stratified/adjusted for gender, age and period were fitted to calculate rate ratios (RRs) and 95% CIs.
Results In zone A in males, we found elevated mortality from circulatory diseases in the first decade after the accident (17 deaths, RR 2.00, 95% CI 1.24 to 3.23). In females, mortality from diabetes mellitus was increased, with a positive trend across zones. Incidence of soft tissue sarcoma was increased in males in zone R in the first decade (6 cases, RR 2.62, 95% CI 1.01 to 6.83). In females in zone B, there was an excess of non-Hodgkin’s lymphoma after 30 years (6 cases, RR 2.87, 95% CI 1.14 to 7.23). Multiple myeloma was increased in the second decade in females in zone B (4 cases, RR 5.09, 95% CI 1.82 to 14.2) and in males in zone R (11 cases, RR 2.15, 95% CI 1.08 to 4.26). In males in zone R, there was a leukaemia excess after 30 years (23 cases, RR 2.02, 95% CI 1.04 to 3.93).
Conclusions Although with different patterns across gender, zone and time, we confirmed previous results of increased cardiovascular diseases, diabetes, soft tissue sarcoma, and lymphatic and haematopoietic cancers.
- Mortality
- Epidemiology
- Public health
- Accidents
- Community Medicine
Data availability statement
Data are available on reasonable request. Data underlying this article will be shared on reasonable request to the corresponding author.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Previous studies of the Seveso population living in the area contaminated with tetrachlorodibenzo-para-dioxin (TCDD) after an industrial accident (1976) showed elevated risks for cardiovascular diseases, diabetes, chronic obstructive pulmonary disease, lymphatic and haematopoietic neoplasms and breast cancer.
WHAT THIS STUDY ADDS
We confirmed lack of elevated all cause mortality and all cancer incidence.
We confirmed elevated mortality from circulatory diseases and diabetes mellitus and elevated incidence of soft tissue sarcomas and lymphatic and haematopoietic cancers. These results are consistent with old and recent literature findings (except for diabetes, for which epidemiological evidence is weak).
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Although patterns of mortality and cancer incidence showed differences across gender and zones and over time, our findings are in line with literature findings and IARC evaluation of TCDD carcinogenicity.
Communication of findings has been and is important for the affected population and for the institutions (the local municipalities and the Regional Government of Lombardy).
Introduction
2,3,7,8-tetrachlorodibenzo-para-dioxin (TCDD) is the most toxic among dioxin-like compounds. It is a persistent organic pollutant,1 with a half-life of 9–15 years on the soil surface and 25–100 years in subsurface soil.2 TCDD is highly lipophilic and accumulates in fat tissues of bio-organisms: the half-life in humans ranges from 5.1 to 11.3 years.3 The International Agency for Research on Cancer (IARC) classified TCDD as a group 1 carcinogen.4 Although epidemiological studies showed positive associations with soft-tissue sarcoma (STS), non-Hodgkin’s lymphoma (NHL) and lung cancer, IARC rated the human epidemiological evidence as sufficient based on results for all cancers combined, sufficient evidence in experimental animals and mechanistic information.4 TCDD binds to the aryl hydrocarbon receptor (AhR) and most of its toxic effects are initiated through this mechanism. There is strong evidence of an AhR-mediated mechanism in humans for TCDD carcinogenesis: the primary mechanism is the promotion of tumour development through modification of cell replication and apoptosis while a secondary mechanism is related to increases of oxidative stress causing DNA damage.4
Human exposures can occur in occupational and environmental settings. Release of TCDD in the environment occurred after industrial accidents in 1953 in Ludwigshafen, Germany,5 6 in 1963 in the Netherlands,7–10 and 1976 in Seveso, Italy. The Seveso accident took place on 10 July 1976, when a chemical cloud was released from a plant producing 2,4,5-trichlorophenol and caused the contamination with more than 34 Kg of TCDD of a densely populated area. The population had been directly exposed to the cloud in the first days: visible effects were not immediate, they appeared a few days later on trees and vegetation, on birds and courtyard animals, and on humans (eye irritation and skin lesions, mainly in children). Consumption of locally produced food was prohibited. The fact that the cloud contained TCDD was communicated by the company owning the chemical plant only on 23 July.1
Preliminary TCDD measurements in soil were made soon after the accident, and in the following months and years systematic campaigns were performed in which TCDD in soil samples was quantified with low-resolution gas-liquid chromatography in combination with MID-mass spectrometry (analytical detection threshold: 0.75 µg/m2).11 Based on thousands of TCDD measurements, the area was divided into three zones with decreasing TCDD levels, named A, B and R (‘respect’) zones (figure 1).1 11 In zone A, residents were evacuated, their houses were destroyed and the top 50 cm of soil removed. The accident had worldwide resonance and a profound impact: three European Directives aimed at controlling major chemical accident hazards were issued in 1982, 1996 and 2012.
The Seveso, Italy, area, including the territory of 11 towns. The map indicates the three contaminated zones A (black), B (dark grey) and R (light grey), and the surrounding non-contaminated reference zone (white). The table reports soil and lipid-adjusted 2,3,7,8-tetrachlorodibenzo-para-dioxin (TCDD) levels. *Schecter1; †Needham et al 27; ‡Landi et al 28.
In the early postaccident period, the only ascertained health effect was chloracne, occurring mainly in children.12 Subsequent studies reported the association of TCDD with lymphocyte alterations,13 thyroid-stimulating hormone (TSH) levels in children of exposed mothers,14 dental abnormalities15 and reduction of the male/female ratio at birth,16 IgG levels17 and sperm quality.18 The Seveso Women’s Health Study (SWHS), a cohort study launched in 1996 including 981 women with blood samples collected near the time of the accident, reported positive associations between TCDD blood levels and incidence of all cancer, breast cancer and cardiovascular diseases.19 Workers of the chemical plant (about 325 subjects) were included in an IARC multicentre mortality study of workers exposed to phenoxy herbicides, chlorophenols and dioxins (36 cohorts in 12 countries, almost 22 000 workers), which found elevated mortality from STS, NHL, all cancers, lung cancer and ischaemic heart disease (IHD).20 21
The whole population living in TCDD polluted zones and a reference population from the surrounding non-contaminated territory were included in a large cohort study. In the most polluted zones, we recorded elevated cardiovascular mortality in the first years after the accident and increases in diabetes and chronic obstructive pulmonary disease (COPD).22–24 In the same zones, the most consistent finding was the elevated mortality from and incidence of lymphatic and haematopoietic (LH) neoplasms in both genders.25 26 Elevated breast cancer incidence in the most polluted zone was observed 15–19 years after the accident.26 We report here the mortality and cancer incidence regarding the period 1976–2013 among people living in the area at the time of the accident.
Methods
Methods for cohort identification, exposure definition, follow-up and cause of death or cancer diagnosis ascertainment (which varied over time) were previously described and are briefly summarised here.23 26
The Seveso cohort
Subjects living in zones A, B and R at the time of the accident represent the exposed group of the cohort. They had been directly exposed to the toxic cloud and may have consumed locally produced food in the days and weeks after the accident, before preventive measures had been undertaken. The three zones included parts of the territory of six municipalities (Barlassina, Bovisio Masciago, Cesano Maderno, Desio, Meda and Seveso) (figure 1). Residents in the unaffected territory of these six municipalities and those living in five surrounding non-contaminated towns (Lentate sul Seveso, Muggiò, Nova Milanese, Seregno and Varedo) were taken as reference. The overall study cohort thus included residents in 11 municipalities.
Exposure definition
Assignment of individuals to zones was based on the official residence on the day of accident, provided by the Vital Statics Offices (VSO) of the 11 municipalities. The main advantage of the official residence is its availability for every cohort member. The zone-based classification was in good agreement with blood TCDD measurements (figure 1).27 28 Living in the area after the accident appeared not to entail additional exposure: none of zone B residents’ serum dioxin levels increased over time and no detectable serum TCDD levels were found in a small sample of people who entered the area after the accident.27 For this reason, we restricted the analysis to people who were residents there at the time of the accident.
Follow-up
For the period 1976–2006, the VSO of the 11 municipalities regularly provided population updates (residents and deaths, migration outside the area). For those who emigrated outside the study area (either within the Lombardy Region or to other Italian regions), we performed a postal follow-up through the VSOs of thousands of municipalities throughout Italy. For the period 2007–2013, for all members of the cohort (including residents in the 11 municipalities) vital status was ascertained through record linkage with two databases of the Lombardy Region (one containing all residents and one containing deaths). For those not linked or emigrated outside region, we performed a postal follow-up as above. These latter activities were started in mid-2014.
Causes of deaths
In the first years of follow-up, death certificates were provided by VSO and coding of the underlying cause of death was performed by trained personnel following international rules of the International Classification of Diseases (ICD). Since 1996 death codes were retrieved by record linkage with databases of the National Central Statistics Institute (ISTAT), of Lombardy region, or of local Lombardy Region Local Health Units, or (for those who emigrated outside the region) by postal contact with VSOs and Local Health Units in other Italian regions.
Cancer cases
Due to lack of cancer registries, identification of cancer cases was restricted to people living in Lombardy using different procedures. For the period 1977–1996, the cohort dataset had been linked with anonymous hospital admission databases of the region using gender, date of birth and residence. On verification of demographics, clinical information was collected from paper medical records and data stored in a dedicated database. Then, date of first primary cancer diagnosis, site and histology were defined and coded for each individual according to the ICD in use at the time of cancer occurrence and to the ICD for Oncology (ICD-O-3).26
For the period 1997–2006, the same procedures were followed, but (because of the increasing number of cancer cases over time and the resulting work burden) data collection was restricted to all subject in the contaminated zones and to a random sample of 40 000 residents in the non-ABR reference zone.
For the period 2007–2013, for all subjects, we obtained data by record linkage with the ATS Brianza Cancer Registry, which covers a population of approximately 1.25 million people, including all those living in the 11 municipalities of the Seveso area.
Statistical analysis
We included in analyses only people who were residents in the area at the time of the accident. For mortality analyses (non-neoplastic diseases), we computed person-years of observation from 10 July 1976 to the earliest of death, loss to follow-up or end of study (31 December 2013). For cancer incidence analysis, we started counting person-years from 1 January 1977 to the earliest of first cancer occurrence (excluding non-melanoma skin cancers), death, loss to follow-up or 31 December 2013. Only the first cancer occurrence was analysed. We included in analyses death certificate only (DCO) cancer cases. The reason for this choice was that we had no information on cancer occurrence for people who emigrated outside Lombardy in the period 1977–2006 and for those who emigrated outside the area covered by the local cancer registry in the period 2007–2013. In cancer incidence analyses, non-sampled subjects in the reference zone in 1997–2006 were censored on 31 December 1996.
We calculated cause-specific mortality and cancer incidence rate ratios (RR) and 95% CI using Poisson regression models adjusted for period (5-year categories except 2007–2013) and age (<1, 1–4 years, then 5 year categories until 84 and 85+ years). For selected diseases, we performed age-adjusted analyses of time since first exposure (‘latency’, 0–9, 10–19, 20–29, 30+ years). All analyses were adjusted/stratified by gender. We performed secondary analyses on selected subgroups of the cohort including 180 persons with chloracne and 2418 residents in a quarter named ‘Polo’ in the town of Meda, possibly with higher exposure than the surrounding zone R, as witnessed by the frequency of chloracne in children (19 cases, 2.5%), which was higher compared with that in zone B (8 cases, 0.5%) and R (63 cases, 0.7%).12 29 Data management, person-year calculation and statistical analyses were performed with Stata V.18 (StataCorp. 2023).
Results
The cohort included 218 682 subjects (111 832 females and 106 850 males) living in the area on 10 July 1976 (table 1). Age was slightly lower in zone B. Most of the residents in zone A lived in Seveso. The follow-up was over 98% complete in each zone. The proportion of missing or non-informative causes of death was 1.1% (2397), with small differences across zones. Cancers with histological or cytological confirmation were the majority; DCO cancer diagnoses were about 10%.
Characteristics of subjects of the Seveso, Italy, cohort, by zone of residence at the time of the accident, 1976–2013
Mortality
In the whole study period, all-cause and all non-cancer mortality in zones A-B were not elevated while in zone R a minimal elevation was found, especially in females (table 2). In females, mortality from diabetes was higher than the reference in all the three polluted zones, with a positive trend from the less to the most polluted zone; slightly elevated mortality from circulatory disease, chronic IHD and cerebrovascular disease was observed in zone R while increased risk from hypertension was observed in zone A. In males, chronic IHD and other heart disease mortality were higher in zone A; other heart diseases mortality was slightly higher also in zone R. Mortality from COPD was elevated in zone B (females) and R (males). Finally, slightly elevated mortality from digestive diseases was observed in zone R (both genders).
Results of Poisson regression analyses of mortality in the Seveso cohort, Italy, 1976–2013: number of deaths (n), rate ratio (RR) adjusted/stratified for gender, age and period, and 95% CIs for the polluted zones (A, B, R) compared with the reference zone
Cancer incidence
Overall cancer incidence did not increase in both genders in the three zones (except for a slight elevation in females in zone A) (table 3). Rectal cancer incidence was increased in zone B (both genders). The excess of other digestive cancers in zone A was based on only two cases in females. Ovarian cancer incidence was increased in zone R. Incidence of LH cancers was increased in females in zone A (based on very few cases across various LH cancer types), and in zone B, to which contributed Hodgkin’s disease (HD, few cases), NHL, multiple myeloma and myeloid leukaemia.
Results of Poisson regression analyses of cancer incidence in the Seveso cohort, Italy, 1976–2013: number of cancer cases (n), rate ratio (RR) adjusted/stratified for gender, age and period, and 95% CIs for the polluted zones (A, B, R) compared with the reference zone
Analyses of time since the accident
In both genders, we observed 41% elevated mortality in zone A in the first 10 years since the accident (table 4). Mortality from diabetes mellitus showed a different pattern across zones and latency. There was a 59% excess mortality from circulatory diseases in zone A within the first decade. COPD mortality was elevated in zone A in the first 20 years. STS was increased in zone R in the first decade. The incidence of LH tissue cancers increased by 60% in zone B in the first two decades, driven by multiple myeloma.
Results of Poisson regression analyses of mortality and cancer incidence in the Seveso cohort, Italy, 1976–2013, for selected causes, by time since the accident (latency): number of deaths or cancer cases (n), rate ratio (RR) adjusted for gender and age and 95% CIs for the polluted zones (A, B, R) compared with the reference zone
Analyses of time since the accident in females (online supplemental table 1) showed a more than doubled mortality from diabetes mellitus in zone B in the second decade and (based on few deaths) after 30 years, while in zone R, we found about a 50% excess 10–37 years after the accident. Breast cancer incidence was moderately in excess in zone A 10–19 years after the accident, but the estimate was imprecise. LH tissue cancers were in excess in zone B in the first two decades and more than doubled in zone A in the second and third decades (based on very few cases). We observed an excess of NHL after 30 years in zone B and R and of multiple myeloma (based on a few cases) in the second decade in zones A–B.
Supplemental material
In males (online supplemental table 2), in zone A we observed elevated mortality in the first 10 years after the accident, especially from circulatory diseases and few deaths from COPD. STS was increased in zone R in the first decade. In zone R, multiple myeloma was increased in the second decade and leukaemia after 30 years. In zone B, leukaemia incidence was increased in the second and third decade.
Secondary analyses
Among 180 chloracne cases, on average quite young at the time of the accident, we recorded four deaths and four cancer cases (online supplemental table 3). We observed RRs below unity for all-cause mortality and for all cancer incidence (online supplemental table 4).
Among 2724 residents in the ‘Polo’ quarter (1334 females, 1390 males), we observed no increased non-cancer mortality (online supplemental table 5). Mortality from circulatory diseases (chronic IHD in particular) was elevated in females while other heart diseases were elevated in males. COPD mortality was elevated in both genders.
Among Polo residents, overall cancer incidence was not elevated (online supplemental table 6). In males, we found an increased incidence of cancer of the oesophagus and of sarcomas (any site, based on few cases). Increased incidence of multiple myeloma was observed, and the excess concerned both genders. This excess was concentrated in the period 20–29 years since the accident (6 cases, RR 4.60, 95% CI 1.90 to 11.1) and was observed for both females (3 cases, RR 4.46, 95% CI 1.29 to 15.4) and males (3 cases, RR 4.74, 95% CI 1.33 to 16.9).
Discussion
Over the whole follow-up period, in the polluted zones, we found no increased all-cause mortality and all cancer incidence. In males we observed elevated non-cancer mortality in the most polluted zone A in the first decade after the accident, mainly due to circulatory diseases and a few cases of COPD. In females, mortality from diabetes mellitus was increased in the polluted area, with a positive trend from the less to the most polluted zone. There was a breast cancer excess in zone A in the second decade after the accident. Incidence of soft tissue sarcoma was increased in males in zone R in the first decade. In females, LH tissue cancers were in excess in zone B in the first two decades and in zone A in the second and third decades. In females in zone B, there was an excess of NHL 30+ years after the accident. Multiple myeloma was increased in the second decade in females in zone B and in males in zone R. There was an isolated leukaemia excess in males in zone R 30 years after the accident.
We did not find elevated mortality and cancer incidence risks among chloracne cases. Conversely, multiple myeloma was elevated in both genders in the Polo quarter, an area with possibly higher TCDD exposure within zone R.
These results are in general in agreement with previous mortality (1976–2001)22 23 30 and cancer incidence (1977–1996)25 26 studies in this cohort because most associations were observed in the first decades after the accident. In this updated follow-up, the overall LH cancer increase was less evident while NHL in females (zone B and R) and leukaemia (zone R) emerged 20–37 years after the accident. Moreover, we observed high multiple myeloma incidence in the Polo quarter 20–29 years after the accident.
Strengths and limitations
This study has several strengths. First, it allowed to have a comprehensive picture of mortality and cancer incidence for 37 years after the accident in the absence (until recently) of a cancer registry. Second, a reference area very close to the polluted zones has been used with similar public, free-access and good quality healthcare services, lowering the possibility of important confounding effects from environmental determinants and lifestyle habits and of misclassification of effects from differential ascertainment of causes of death or diagnostic performance. Third, the whole population (exposed and unexposed components) has been followed up as a unique cohort using the same procedures (although varying over time) to determine vital status, causes of death and cancer diagnoses. Fourth, vital status and cause of death ascertainment were nearly complete, so important selection biases are unlikely.
The main limitation is the exposure definition, which was ecological, being based on residence in areas defined according to soil TCDD concentrations. Although studies showed clear gradients of serum TCDD across zones,27 28 31 some degree of misclassification is unavoidable because of heterogeneity of exposure to TCDD within zones. Additionally, it is possible that a fraction of the population was not in the area in the days after the accident (although it was probably small, given that the typical holiday month is August). Second, differently from mortality, cancer diagnoses could only be ascertained for people living in Lombardy (1977–2006) and for those living in the area covered by the cancer registry (2007–2013); for these reasons, we included DCO cancer cases in analysis to minimise potential selection biases. In this way, we introduced some misclassification bias, which, at least on average, was towards the null and probably small, given the relatively small DCO proportion (about 10%).
Biological plausibility
Different mechanisms for the action of TCDD on glucose metabolism have been hypothesised in in vitro and epidemiological studies. Animal and human studies showed that TCDD can adversely affect the cardiovascular system through various mechanisms. Moreover, a potential contributor to the circulatory disease excess in the early postaccident period is the heavy psychosocial impact of the accident on people living in zone A. Regarding respiratory effects, possible mechanisms include direct toxicity of TCDD on bronchiolar and alveolar tissue, interference with the immune system and oxidative stress. Regarding the increased LH cancer incidence, positive associations were found between TCDD and lymphocytes t(14;18) translocations in lymphocytes and impaired B lymphopoiesis. More details and references on biological plausibility can be found in the online supplemental material.
Comparison with published research
Old studies found weak evidence regarding TCCD and diabetes while results regarding cardiovascular diseases have been more consistent (increased IHD mortality with positive exposure-response associations). In a non-systematic search, we found several studies and reviews on the effects of TCCD published after (or not included in) the last IARC evaluation. Studies of occupational cohorts in the USA, the Netherlands, Germany and New Zealand found positive associations between exposure to TCDD and various diseases, including circulatory diseases, all cancers, NHL, STS, leukaemia, multiple myeloma and breast cancer. Studies among Korean Veterans found elevated risks (mortality, incidence or prevalence) of all cancers, lung cancer, NHL, circulatory and respiratory diseases and diabetes. The SWHS reported positive associations between TCDD blood levels and incidence of all cancer, breast cancer and cardiovascular diseases. A meta-analysis found a dose–response relationship between serum TCDD and diabetes only at low TCDD levels and no association at high levels. The association of TCCD exposure with haematological cancer risk has been evaluated in two reviews that reached opposite conclusions. A meta-analysis found that TCDD was associated with all cancer incidence and mortality and with NHL mortality. A detailed discussion of these studies (with references) and how our study compares with them can be found in the online supplemental material.
Conclusions
The Seveso cohort study represented a landmark in Italy because cancer incidence could be assessed in a large population in an area not covered by a cancer registry until 2006, by means of record linkage and manual search of clinical records in hospital archives. Several papers on results in the whole and in the young population have been published over the years in various journals. Communication of findings has been important for the affected population and for the institutions (the local municipalities and the Regional Government of Lombardy). Although with different patterns across gender, zone and time, some results (increased cardiovascular diseases, soft tissue sarcomas and LH cancers) are consistent with previous researches (in this and other populations) and with IARC evaluation of TCDD carcinogenicity.
Data availability statement
Data are available on reasonable request. Data underlying this article will be shared on reasonable request to the corresponding author.
Ethics statements
Patient consent for publication
Ethics approval
This study was established by law N.2 (17 January 1977) of the Lombardy Region.
Acknowledgments
We dedicate this work to Pier Alberto Bertazzi (1945–2021), who played a major role in conceiving, designing, implementing and conducting the Seveso cohort study. We wish to acknowledge the effort of the personnel of the following institutions, without whom this work would not have been feasible: the Mayors and the Vital Statistics Offices of the towns of Barlassina, Bovisio Masciago, Cesano Maderno, Desio, Lentate sul Seveso, Meda, Muggiò, Nova Milanese, Seregno, Seveso and Varedo for their continuing cooperation; the Health Directorate, Lombardy Region for assistance in record-linkage procedures; the Epidemiology Offices of the Local Health Units, Lombardy Region for providing causes of death; Enrico Radice for the supervision and realisation of the different tasks related to follow-up and cause of death ascertainment. Preliminary results of this study were presented in an invited lecture at the 28th Conference of the International Society for Environmental Epidemiology (ISEE), Rome 2016 (Bertazzi PA. Industrial disasters. The Seveso, Italy accident and its 40-year legacy).
References
Supplementary materials
Supplementary Data
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.
Footnotes
Contributors ACP conceived and designed the study. DC is responsible for the overall content as guarantor, participated in data collection and performed statistical analyses. DC and ACP wrote the first draft of the manuscript. LCd'O and MR contributed to collection of information on cancer incidence and mortality. All authors contributed to revisions of the manuscript.
Funding This work was originally supported by the Regional Government of Lombardy, within the frame of the Environmental Epidemiology Program VIII-002306—2006-2009. Additional funding was received by the National Ministry of University and Research, FIRST Program and the Ministry of Health, Applied Research Program.
Map disclaimer The depiction of boundaries on the map(s) in this article does not imply the expression of any opinion whatsoever on the part of BMJ (or any member of its group) concerning the legal status of any country, territory, jurisdiction or area or of its authorities. The map(s) are provided without any warranty of any kind, either express or implied.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.