Article Text

PDF

Environment
Urinary 8-hydroxy-2′-deoxyguanosine as a biomarker of oxidative DNA damage induced by ambient pollution in the Normative Aging Study
  1. Cizao Ren1,
  2. Shona Fang2,
  3. Robert O Wright3,4,
  4. Helen Suh1,
  5. Joel Schwartz1
  1. 1Department of Environmental Health, Exposure, Epidemiology and Risk Program, Harvard School of Public Health, Boston, Massachusetts, USA
  2. 2Environmental and Occupational Medicine and Epidemiology Program, Harvard School of Public Health, Boston, Massachusetts, USA
  3. 3Department of Pediatrics, Children's Hospital, Boston, Massachusetts, USA
  4. 4Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, USA
  1. Correspondence to Cizao Ren, Department of Environmental Health, Exposure, Epidemiology and Risk Program, Harvard School of Public Health, Landmark Center, West, Suite 415, PO Box 15677, 401 Park Dr., Boston, MA 02215, USA; rencizao{at}yahoo.com

Abstract

Background Studies show that exposure to air pollution damages human health, but the mechanisms are not fully understood. One suggested pathway is via oxidative stress.

Objectives This study examines associations between exposure to air pollution and oxidative DNA damage, as indicated by urinary 8-hydroxy-2′-deoxyguanosine (8-OHdG) concentrations in ageing participants during 2006–2008.

Methods We fit linear regression models to examine associations between air pollutants and 8-OHdG adjusting for potential confounders.

Results 8-OHdG was significantly associated with ambient particulate matter ≤2.5 μm in aerodynamic diameter (PM2.5), nitrogen dioxide (NO2), maximal 1 h ozone (O3), sulphate (SO42−) and organic carbon (OC), but not with black carbon (BC), carbon monoxide (CO), the number of particles (PN) or elemental carbon (EC). Effects were more apparent with multi-week averages of exposures. Per IQR increases in 21-day averages of PM2.5, PN, BC, EC, OC, CO, SO42−, NO2 and maximal 1 h O3 were associated with 30.8% (95% CI 9.3% to 52.2%), −13.1% (95% CI −41.7% to 15.5%), 3.0% (95% CI −19.8% to 25.8%), 5.3% (95% CI −23.6% to 34.2%), 24.4% (95% CI 1.8% to 47.1%), −2.0% (95% CI −12.4% to 8.3%), 29.8% (95% CI 6.3% to 53.3%), 32.2% (95% CI 7.4% to 56.9%) and 47.7% (95% CI 3.6% to 91.7%) changes in 8-OHdG, respectively.

Conclusions This study suggests that ageing participants experienced an increased risk of developing oxidative DNA injury after exposure to secondary, but not primary, ambient pollutants.

  • 8-Hydroxy-2'-deoxyguanosine
  • air pollution
  • DNA damage
  • oxidative stress
  • biomarker
  • epidemiology
  • public health
View Full Text

Statistics from Altmetric.com

What this paper adds

  • Particulate air pollution is associated with health outcomes but the mechanism remains to be clarified.

  • One of suggested mechanistic pathways is via oxidative stress; however, only a limited number of studies have directly examined associations between air pollution and oxidative DNA damage.

  • Results show that 8-OHdG, a biomarker of DNA damage, was significantly associated with secondary pollutants, but not with primary pollutants among an ageing population.

  • This result delivers an important message that exposure to secondary pollution may further increase risk in chronic human conditions such as ageing, cancer and other degenerative diseases.

Introduction

Growing evidence suggests that reactive oxygen species (ROS) play a vital role in human disease development.1–5 ROS refer to oxidation via multiple chemicals that are oxidising agents and/or are easily converted into radicals such as ozone (O3), peroxynitrite, HOCl, hydrogen peroxide and singlet oxygen.5 ROS have been shown to possess many characteristics of carcinogens and can result in endothelial dysfunction as well as DNA damage.2 ROS can cause DNA structural alteration, including base pair rearrangements, mutations, insertions, deletions and sequence amplification.6 They can also influence cytoplasmic and nuclear signal transduction pathways,7 modulate the activities of genes and proteins responding to oxidative stress and regulate the genes related to cell proliferation, differentiation and apoptosis.5

8-Hydroxy-2′-deoxyguanosine (8-OHdG) is a common biomarker of DNA lesion that is induced by the reaction of hydroxyl radicals with 2′-deoxyguanosine at the C-8 position in DNA.8 8-OHdG is released into the circulation after DNA repair by the DNA base excision repair pathway and then is excreted into urine.8 9 Studies show that the urinary concentration of 8-OHdG is not influenced directly by either diet or cell turnover. Therefore, it is a good biomarker for ROS or oxidative stress.9

A large literature has shown some human diseases are related to exposures to air pollutants such as PM2.5 and O3.10–15 Many studies have shown that oxidative stress is one important pathway through which particulate matter and O3 execute their effects on cardiovascular and other diseases.16–19 Exposure to concentrated air particles can result in oxidative stress in both the lung and the heart.18 20 A limited number of studies have examined associations between 8-OHdG and exposures to indoor and ambient pollution or smoking, but they were usually conducted among a small number of children or occupationally exposed employees.1 21 22 The present study was designed to investigate whether exposure to a rich set of ambient pollutants was associated with urinary 8-OHdG among a larger cohort of elderly individuals from the Normative Aging Study (NAS), and whether the pattern of the associations was informative about which types of pollutants produced these effects. The pollutants involved in this study included ambient particulate matter ≤2.5 μm in aerodynamic diameter (PM2.5), nitrogen dioxide (NO2), O3, carbon monoxide (CO), sulphate (SO42−), black carbon (BC), organic carbon (OC), number of particles (PN) and elemental carbon (EC).

Methods

Study population

The NAS, a longitudinal study of ageing, was established by the Veterans Administration (VA) in 1961 when 2280 men from the greater Boston area who were free of known chronic medical conditions were enrolled.23 Participants underwent detailed examinations every 3–5 years, including routine physical examination, laboratory tests, collection of medical history, social status information, and administration of questionnaires on smoking history, food intake and other factors that may influence health. Between January 2006 and December 2008, all 320 participants who appeared for examination were evaluated for urinary 8-OHdG and other covariates. The present investigation was approved by the Institutional Review Boards of all participating institutions.

Plasma analysis of B vitamins, creatinine and 8-OHdG

Fasting plasma samples were drawn at the VA field site and stored at −80°C. Urinary samples were collected on the same morning of the visit and stored at −80°C until analysis. Folate, vitamin B6 and B12 in fasting plasma were measured at the USDA Human Nutrition Research Center on Aging at Tufts University. Folate and vitamin B12 were assessed by radioassay using a commercially available kit from Bio-Rad (Hercules, California, USA) and vitamin B6 (as pyridoxal-5-phosphate) by an enzymatic method using tyrosine decarboxylase. Further details are described elsewhere.24 25 Blood and urinary creatinine was measured with urine 8-OHdG using spectrophotometric assay. The method has been described elsewhere in detail.26

A competitive ELISA (Genox, Baltimore, Maryland, USA) was used to determine urinary 8-OHdG.27 28 In brief, 50 μl of urine samples, quality control (QC) and standards were added to micro-titre plates pre-coated with 8-OHdG protein conjugate. After washing out thrice with 250 μl wash solution, an enzyme-labelled secondary antibody (100 μl) was added to plates for incubation for 1 h at 37°C. After washing as above, 100 μl of the chromatic substrate, (3,3′,5,5′)-tetramethylbenzidene, was added to the plates and then allowed to react for 15 min at room temperature. The intensity of colour produced for each sample was measured at an optical density of 490 nm in standards, QC and urine samples. The concentration of 8-OHdG was calculated based on the colour intensity. A pooled urine sample from several healthy adults was used as the quality control samples. For each standard 96-well microplate, six to nine QC samples were randomly placed along with the unknown samples. The measured QC values were averaged and compared with the established QC value. For each unknown sample, either a duplicate or triplicate measurement was performed. The average, SD and coefficient of variation (CV) (%) were calculated and any sample with CV (%) equal to or greater than 20% was re-tested.

Air pollution and weather data

Continuous hourly PM2.5, PN, NO2, O3, CO, SO42−, BC, OC and EC were measured at a stationary monitoring site 1 mile from the examination location. The pollutants were divided into two groups, the primary and the secondary pollutants. The primary pollutants refer to those that are directly emitted into the atmosphere. The secondary pollutants refer to those produced after the primary pollutants react or interact with other components in the atmosphere. Some pollutants may be both primary and secondary, that is, they are both emitted directly and formed from other primary pollutants. In this study, CO, EC and BC are the primary pollutants and NO2, O3 and SO42− are the secondary pollutants. PM2.5 and OC are mixtures of primary and secondary pollutants, which we categorised into the secondary group. PN is the number of particles, highly related to the secondary pollutant PM2.5. BC was measured using an aethalometer (Magee Scientific, Berkeley, California, USA) and PM2.5 was measured using a Tapered Element Oscillating Microbalance (model 1400A; Rupprecht & Patashnick, East Greenbush, New York, USA), operated at 50°C with two 4 l per minute PM2.5 impactors before the inlet. CO was measured using a gas filter correlation carbon monoxide analyser by comparing infrared energy absorption (Model 300E; Teledyne Advanced Pollution Instrumentation, San Diego, California, USA). O3 was measured using a UV absorption ozone analyser (Model 400E; Teledyne Advanced Pollution Instrumentation). NO2 was measured using the proven chemiluminescence detection principle, coupled with state-of-the-art microprocessor technology to provide sensitivity and stability (Model 200E; Teledyne Advanced Pollution Instrumentation). EC and OC were measured with a Rupprecht & Patashnick Ambient Carbon Particulate Monitor (Model 5400).29 SO42− was measured based on XRF determination of the sulphur on the particle filters. Daily averages of their concentrations were used for this study during 2006–2008. Data for OC and EC were only available between January 2007 and December 2008. Gaseous air pollutant data were provided by the Massachusetts Department of Environmental Protection. The moving averages of daily PM2.5, NO2, SO42−, CO, BC, EC, OC and maximal 1 h O3 up to 4 weeks before the visit were used as the exposure indices. To adjust for outdoor weather, we used apparent temperature as an index, defined as a person's perceived air temperature, given the humidity.30

Statistical analyses

Statistical analyses were performed with R v 2.7.2. We fitted linear regression models to separately examine the association between a single pollutant and urinary 8-OHdG using different moving averages of the exposure. We calculated the Pearson correlations between pollutants. Because we felt that longer averaging times were probably more appropriate for a marker of oxidative damage, we considered moving averages of exposure up to 28 days. Because primary analyses showed a skewed distribution of 8-OHdG, we used the log transformation of 8-OHdG concentrations to improve the normality of residuals and to stabilise the variance. We identified a priori the following variables as important determinants of 8-OHdG, based on our previous NAS studies and other studies, because they might confound the associations between air pollution and 8-OHdG: age, body mass index (BMI), smoking status (never, former, current), pack-years of cigarettes smoked, alcohol consumption (≥2 drinks/day; yes/no), use of statin medication (yes/no), season, plasma folate, vitamin B6 and vitamin B12.19 31 We adjusted for age, BMI, pack-years of cigarettes smoked, plasma folate, vitamin B6 and vitamin B12 as continuous variables. We adjusted for smoking status, alcohol consumption, use of statin medication and season as categorical variables. Because of the potential non-linear relationship between temperature and 8-OHdG, we also adjusted for 3-day moving average of apparent temperature using both linear and quadratic terms. In addition, because the concentration of 8-OHdG was related to kidney function, we adjusted for creatinine clearance rate using the Cockcroft-Gault formula ((140−age(year))×weight(kg))/(72×serum creatinine(mg/dl))).32 We also adjusted for chronic disease status (cardiovascular disease or chronic respiratory diseases) as a dummy variable. For comparisons with day moving average effects of pollution, we examined the accumulative lag effects of each pollutant up to 4 weeks using unconstrained distributed lag methods.33

Results

Tables 1 and 2 present the study population characteristics and average concentrations of pollutants. The study population consisted of 320 men, 309 (97.5%) of whom were non-Hispanic white subjects. Their ages ranged from 63 to 96 years, with a mean±SD age of 76.7±6.1 at the time of their visit. On average, the 8-OHdG concentration was 20.8±12.3 ng/ml, with the log-transformation 2.81±0.78 log ng/ml. Overall, 68.8% of the participants ever smoked, 29.1% never smoked and only 2.2% still smoked at the time of their visit. The means of daily concentrations of pollutants are shown in table 2.

Table 1

Descriptive statistics of the demographic and health variables of participants (n=320)

Table 2

Daily averages of air pollution and temperature (n=320)

The Pearson correlation coefficients between all pollutants on the day of the participants' visit show that PM2.5 was highly correlated with SO42−, BC and OC, and moderately correlated with O3 and EC (table 3). O3 was only moderately or poorly correlated with most of the other pollutants. EC and OC tended to be highly correlated with the other pollutants.

Table 3

Pearson correlation coefficients of air pollutants on the days of participant visits

We fit models to separately estimate associations between a single pollutant and 8-OHdG using different day moving averages up to 4 weeks. Table 4 shows the per cent changes in 8-OHdG per interquatile range (IQR; calculated separately for each averaging period) increase in each pollutant for the current day, 7-, 14- and 21-day moving averages. Results for 28-day averages were similar to, but usually less significant than, the 21-day results, and are omitted to keep the table size manageable. They are shown in figure 1. PM2.5, maximal 1 h O3, SO42−, NO2 and OC were significantly associated with 8-OHdG at various moving averages, but there were no significant associations between 8-OHdG and CO, EC, PN or BC. Across the IQR in 3-week moving averages of daily concentrations of PM2.5, PN, BC, EC, OC, CO, SO42−, NO2 and maximal 1 h O3, urinary 8-OHdG increased by 30.8% (95% CI 9.3% to 52.2%), −13.1% (95% CI −41.7% to 15.5%), 3.0% (95% CI −19.8% to 25.8%), 5.3% (95% CI −23.6% to 34.2%), 24.4% (95% CI 1.8% to 47.1%), −2.0% (95% CI −12.4% to 8.3%), 29.8% (95% CI 6.3% to 53.3%), 32.2% (95% CI 7.4% to 56.9%) and 47.7% (95% CI 3.6% to 91.7%), respectively. Overall, significant associations were found for more than 10-day moving averages of the secondary pollutants except for OC where a significant effect only appeared after 16-day moving averages. There were no significant effects for the primary pollutants. 3-Week moving averages showed stronger associations for the secondary pollutants in spite of the variation across pollutants (figure 1).

Table 4

Adjusted per cent change (95% CI) in 8-OHdG associated with IQR increases in the moving averages of each pollutant in different period (days)

Figure 1

Adjusted per cent changes in urinary 8-hydroxy-2′-deoxyguanosine for ambient particulate matter ≤2.5 μm in aerodynamic diameter (PM2.5), black carbon (BC), elemental carbon (EC), organic carbon (OC), carbon monoxide (CO), nitrogen dioxide (NO2), sulphate (SO42−), maximal 1 h ozone (O3) and particle number (PN) at various moving averages. Solid lines refer to estimates and dashed lines represent 95% CIs.

The trends of accumulative lag effects of PM2.5, OC, SO42− and maximal 1 h O3 showed significant associations at different days (figure 2). Accumulative lag effects of PM2.5 and OC slowly escalated with fluctuation within the exploration period. The accumulative effect of SO42− increased and fluctuated. The trends of NO2 and 1 h maximal O3 escalated and then declined. There was a declining trend for CO, but there was no significant accumulative effect. There were no accumulative effects for BC, PN or EC. These findings are consistent with effects of moving averages of pollutants.

Figure 2

Adjusted per cent changes in urinary 8-hydroxy-2′-deoxyguanosine for ambient particulate matter ≤2.5 μm in aerodynamic diameter (PM2.5), black carbon (BC), sulphate (SO42−), nitrogen dioxide (NO2), element carbon (EC), organic carbon (OC), maximal 1 h ozone (O3) and particle number (PN) at cumulative lag effects on different days. Solid lines refer to estimates and dashed lines represent 95% CIs.

As a sensitivity analysis, we also adjusted for different day moving averages of apparent temperature up to 3 weeks including its linear and quadratic terms as above. Overall, the estimates were slightly attenuated when using only the current day of apparent temperature, but were basically unchanged using the 2- or more-day moving averages. In addition, after adjusting for urinary creatinine, results varied a little but the trend was the same.

Discussion

Results show that exposure to the secondary pollutants or mixtures of the primary and secondary pollutants (PM2.5, NO2, OC, SO42− and maximal 1 h O3) was significantly associated with oxidative stress biomarker urinary 8-OHdG for more than 10-day moving averages, but no significant associations were observed for the primary pollutants (CO, BC and EC) in the NAS population (table 4 and figures 1 and 2). Significant associations between the secondary pollutants and 8-OHdG were found for both moving day averages and accumulative lag effects of these pollutants. Overall, the 21-day moving averages of the secondary pollutants showed strong effects.

The use of moving averages of air pollution provides statistical stability in estimating longer term effects, but by constraining the estimates to effectively be the same for each lag in the average, may miss part of the pattern. Unconstrained distributed lag models, in contrast, allow the pattern of association to vary by lag, but because of the multicollinearity of the different lags, are quite unstable. The sum of the effects over all lags is more stable, and provides a reasonable check on the moving average estimates to make sure they are not biased by the use of simple moving averages.33 This study used both approaches to estimate associations between an array of air pollutants and 8-OHdG and we found consistent estimates for both methods (figures 1 and 2). As expected, there were more fluctuations for accumulative distributed lag effects, as compared to separately modelled lagged daily averages. We estimated the effect of each pollutant up to 4 weeks (daily moving average or accumulative lag effect) because we think this period would be long enough for the development of acute oxidative or inflammatory responses in humans exposed to pollutants, and we stopped at that point because the effect sizes were falling, and longer averages were beginning to conflict with control for season.

ROS in living cells are continuously generated as the consequence of metabolic reactions. For example, mitochondria (oxidative phosphorylation), leucocytes (oxidative burst), peroxisomes (degradation of fatty acids) and the cytochrome p450 system (mixed function oxidation system) all release ROS. Under normal physiological status, the generation and deletion of ROS are balanced, a situation known as homeostasis. If the generation of endogenous or exogenous oxidants increases or that of antioxidants decreases, global oxidative stress occurs.34 Inhaled pollutants containing oxidants or oxidant-related components can cause physiological oxidative stress. Oxidative stress can cause impaired physiological function, resulting in DNA damage, ageing and disease.35 36 Oxidative stress can cause the oxidisation of 2-deoxyguanosine in DNA to 8-OHdG. Damaged DNA is repaired by the DNA excision repair system and 8-OHdG is released into the circulation.1 31 Urinary 8-OHdG can be used as a biomarker of oxidative stress because it is not influenced directly by diet or cell turnover. Urinary 8-OHdG can be measured non-invasively.5 37

In urban areas, primary pollutants such as NO, CO, BC or EC are mainly produced by petrol or diesel fuel combustion. NO and CO are not oxidants. The ability of BC or EC to produce ROS is not fully known. In contrast, secondary pollutants are formed mostly by atmospheric photochemical processes. Photochemical reactions can result in the formation of strong oxidative molecular, such as O3 and hydrogen peroxide, as well as organic radicals and bi-radicals. Therefore, it is anticipated that when concentrations of important secondary pollutants such O3, SO42− and nitrate among others, are high, concentrations of many gaseous and particulate oxidant species are elevated.38 39 For example, NO2 is mainly produced from NO via oxidative atmospheric chemical reactions. O3 is produced by complicated photochemical procedures via chemical reactions of oxygen with nitrogen oxide, organic peroxy radicals, hydrocarbons, etc, under sunlight; it is a strong ROS.38 39 PM2.5 is a mixture of fine aerosol particles, of which SO42− and OC account for the largest fraction in Eastern Massachusetts. Heavy metals, especially Fe and Cd, are also important components of PM2.5, and they are strong ROS.38–40

A limited number of studies have reported that exposures to ambient and indoor pollution or smoking are associated with 8-OHdG.1 20 41 Chuang et al investigated associations between pollutants and oxidative stress among 76 students aged 18–25 years, and found that increases in blood 8-OHdG were significantly associated with the averages of SO42−, O3 and NOx on various days (1–3 day).21 Kim et al reported that urinary 8-OHdG was significantly associated with fine particle exposure among 20 boilermakers working at a power plant during an overhaul of oil-fired boilers.31 Calderón-Garcidueñas et al found that primary school students exposed to higher air pollution had two- to threefold increased 8-OHdG in the nasal respiratory epithelium compared to those who were exposed to lower levels of air pollution.1 Similar associations have also been found in other populations who were exposed to other pollution sources (such as cigarette smoke), office employees and restaurant workers.22 41 42 Our study found evidence that 8-OHdG was significantly associated with exposures over 10-day averages of secondary pollutants, but not primary pollutants among 320 ageing participants.

Some studies from other fields also support these results. For example, using a florescent marker of ROS, Evelson et al. reported that exposure of animals to concentrated air particles induced oxidative stress, and conversely moving them from room air to filtered air reduced ROS in multiple organs in vivo.43 Air pollution exposure has also been associated with thiobarbituric acid reactive substances, an ROS.44 Sørensen et al reported that personal exposure to PM2.5 was associated with 7-hydro-8-oxo-2′-deoxyguanosine, a DNA damage biomarker.45 On the other hand, unlike other studies, for example that of Delfino et al,46 we did not find associations with the primary pollutants, but rather with the secondary pollutants. Whether this is due to population differences, pollution differences between southern California and Boston, averaging time, the more precise exposure measures used in Delfino et al46 or chance remains to be determined.

One limitation is that we used air pollution concentrations from a single monitoring site as surrogates for recent personal pollution exposure, which may lead to exposure misclassification. Although we were not able to quantify the impact of exposure error on our results, previous studies suggest that exposure error will be greater for primary pollutants such as BC than for secondary pollutants such as PM2.5 and O3.47 This differential exposure error likely results from spatial variability in outdoor concentrations, which have been shown to be greater for primary pollutants. For example, outdoor spatial variability in concentrations of the secondary pollutants O3, SO42− and PM2.5 is limited within urban areas.48 49 In contrast, outdoor CO and BC concentrations have been shown to vary by as much as a factor of 3.50 51 This may overstate the relative differences in the impacts of measurement error on the effect estimates, however. To see this, consider the following. If Xijt is the exposure of the ith subject, in the jth neighbourhood, at time t, then we have the following identity52:Xijt=Ut+(XjtUt)+(XijtXjt)where Ut is the measurement at the central monitor, and Xjt is the average exposure in neighbourhood j, on day t.

While traffic pollutants have much more spatial variation, it occurs on a very fine scale, with noticeable changes between an address on a busy street and one around the corner on a side street. That is, much of the spatial variation in traffic pollution will be seen within neighbourhoods, between subjects. This is the third term of the equation above. And it is Berkson error, which does not bias downward the regression coefficient. Secondary pollutants vary much more slowly spatially. Therefore, a larger fraction of their spatial measurement error is captured in the second term. Hence, while spatial variation overall is larger for traffic pollutants, much of that is on a fine enough scale to be Berkson, and not downwardly bias effect estimates. The second term, which includes classical error, does produce bias, but the relative difference in spatial measurement error on the neighbourhood scale between traffic and non-traffic pollutants is much lower than the overall difference, and hence we believe focusing on the overall spatial variability overstates the potential for greater bias in the coefficients from traffic pollution. Nevertheless, greater downward bias is still likely for traffic pollutants. Despite this greater measurement error, most of the previous reports from this cohort have found a stronger association with primary pollutants. For example in Mordukhovich et al,53 we reported an association between BC and blood pressure, and did not observe an association with the secondary pollutants. In Madrigano et al,54 we reported BC was associated with increases in levels of vascular cellular adhesion molecules. While there are substantial spatial gradients in primary pollutants, the analyses in this study were based on temporary variation or day-to-day fluctuations in pollution concentrations in Boston, which were primarily driven by meteorology. Hence, while our analysis has missed the additional gradient concentrations of primary particles that occurs over space, it captures the temporal gradient.

Conclusion

This study found that in the NAS population exposure to secondary pollutants was significantly associated with 8-OHdG (PM2.5, NO2, OC, SO42− and maximal 1 h O3), but exposure to primary pollutants (CO, EC and BC) was not. These effects were more apparent with multi-week averages of exposures.

References

View Abstract

Footnotes

  • Funding This work was supported by National Institute of Environmental Health Sciences grants ES014663, ES 15172 and ES-00002, by US Environmental Protection Agency grant R832416 and USDA Contract 58-1950-7-707. The Normative Aging Study is supported by the Cooperative Studies Program/Epidemiology Research and Information Center of the US Department of Veterans Affairs, and is a component of the Massachusetts Veterans Epidemiology Research and Information Center. It is partially supported by Harvard-NIOSH ERC Pilot (T42 OH008416).

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Harvard School of Public Health.

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

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.