What this study adds
Atmospheric pollutants have a well documented effect on mortality and morbidity from cardiorespiratory diseases. This study shows that particulate matter (PM10 and PM2.5) and nitrogen dioxide (NO2) might trigger exacerbations in patients with autoimmune diseases with specific epigenetic mechanisms in their pathogenesis, such as Hashimoto thyroiditis, rheumatic diseases, and multiple sclerosis. Exacerbations are mostly revealed through changes in treatment (increased dose and reduced refill time), especially in young patients. The findings underline the importance of using drug prescription data to study the effects of air pollutants in patients with diseases usually managed out of hospital.
The health effects of atmospheric pollutants on cardiorespiratory diseases are well documented and other diseases are increasingly studied. Autoimmune diseases are possible candidates because important premises suggest that air pollutants might affect both the onset and the exacerbation of these conditions.
Impairment of epigenetic mechanisms has been recognized in basic research as being involved in the pathogenesis of autoimmune diseases, in so far as they might interfere with the differentiation of lymphocytes T, by restricting the regulation of immune-tolerance.1–3
The most important mechanisms inducing autoimmunity are DNA methylation1,2 and those involved in providing short-term cellular memory of gene expression such as post-translational modification of histones1–3 and regulation mediated by microRNAs (miRNAs).3,4
Impairment of epigenetic mechanisms has been also recognized in epidemiological and toxicological studies as the interaction factors between gene expression and environmental exposure, such as cigarette smoking,5 alcohol consumption,6 and air pollution.7,8 The mechanisms most involved in the interaction between air pollution and autoimmune disease are changes in DNA methylation,9 that were observed rapidly after daily exposure7 and were also associated to long-term exposure.8 In addition histone modification, and post-transcriptional noncoding RNA are important to provide short-term memory of an environmental signal.3
Thus, the epigenetic mechanisms most involved in the interaction between environment and autoimmune disease are very similar to the mechanisms identified in the pathogenesis of autoimmunity. On these bases, we studied whether daily increases of particulate matter (PM10, PM2.5) and nitrogen dioxide (NO2) were associated with exacerbations of autoimmune diseases in Rome (Italy) in 2006–2014.
We chose to investigate five autoimmune conditions: Hashimoto thyroiditis, systemic lupus erythematosus (SLE), rheumatoid arthritis (RA), psoriasis, and multiple sclerosis (MS), based on the following criteria: (1) a specific epigenetic mechanism has been identified in their pathogenesis, (2) the disease has a distinct diagnostic code in the International Codes of Diseases-Ninth Revision (ICD-9), and (3) a sufficient number of cases had occurred in Rome (Lazio region, Italy) during the study period.
Study population and health data
A total of 23,898 subjects (all ages and both sexes) who were diagnosed with one of the autoimmune diseases indicated above in 2003–2014 were enrolled in five corresponding cohorts. Patients were identified from hospital discharge and emergency room (ER) registries by means of diagnoses made in any of the hospitals in the Lazio region. The copayment exemption registry was used also to recruit patients with multiple sclerosis. All subjects had to reside in Rome (2,874,529 inhabitants), between 2003 and 2014 and be alive as of 1 January 2006, when the data analysis started. In addition, subjects should not have received another autoimmune diagnosis in the 3 years preceding enrollment. If a new autoimmune diagnosis was given during the follow-up, the subject was enrolled in the corresponding cohort also. The ICD-9 codes to identify autoimmune diseases are reported in the e-Table 1; http://links.lww.com/EE/A18.
We searched the enrolled population for subsequent hospital admissions and ER visits and for any changes in drug prescriptions in 2006–2014. Prescription data for each patient were obtained from the regional pharmacy drug sales registry. More details about all the health data registries we used for the study are reported in the e-Appendix 1; http://links.lww.com/EE/A18.
To be considered as an exacerbation, subsequent hospitalizations and ER visits had to follow the initial diagnosis, occur no sooner than 28 days from the previous one, occur in the city of Rome, and have a principle diagnosis of the same autoimmune disease as first diagnosed, or one of its complications/sequelae or another of the studied autoimmune diseases. The ICD-9 codes of the complications/sequelae of each autoimmune disease are reported in e-Table 2; http://links.lww.com/EE/A18.
To be considered as an exacerbation, a prescription needed to be written for the same patient, to follow the initial diagnosis, to represent an increase in dose, or a shorter time interval between two refills, or a different administration route for a faster/higher internal dose (from cutaneous or nasal or inhalation, to oral, to subcutaneous injection) with respect to the previous one.
An increase in the prescribed dose was defined as the positive difference in each active substance between one prescription and the previous same substance, for the same patient. A reduction in prescription timing was defined as a shorter elapsed time interval between two consecutive refills, with respect to the preceding time interval (requiring a substance to be prescribed at least three separate times, for the same patient). Finally, dose increases as well as prescription time decreases needed to be higher than the median dose increases or lower than the mean interval between prescriptions observed for the substance under study, in all patients with the same disease, in the year preceding each prescription analyzed. Patients enrolled in 2006 were compared with median dose increases and mean interval between prescriptions estimates in all patients with the same disease from the same year (2006) because pharmaceutical data were available only from 2006. Changes in prescriptions were assessed only for medications suggested as treatment of specific autoimmune diseases in clinical guidelines or in the scientific literature and approved in Italy for the diseases studied.
These procedures provided a reliable mean interval of prescriptions of an active substance in all the patients with a specific disease, excluding the influence of occasionally taken drugs, and also independent from prescription habits, adherence to treatment, or other events that are likely in chronic patients, such as directly sold drugs.
Prescriptions were considered wherever they were supplied in the entire region. The World Health Organization-Anatomical Therapeutic Chemical (WHO-ATC) codes of specific medications for each autoimmune disease are reported in e-Table 3; http://links.lww.com/EE/A18.
Hourly PM10, NO2, and O3 concentrations were available from three fixed monitoring stations of the Regional Environmental Protection Agency network, chosen to be representative of the background levels in the whole city, for the entire period 1 January 2006 to 31 December 2014. Starting with hourly data, we assessed daily mean levels of pollutants in the city. Daily completeness of 75% per season was used as the inclusion condition for the monitored data. Missing values from a monitor were imputed using the average measurements from the other monitors on the same day, weighted by the ratio of the yearly average at that monitor to the yearly average at the others.10 Daily levels were expressed in micrograms per cubic meter (μg/m3) for each pollutant, and the exposure metric was defined as a daily increase in levels as high as 10 μg/m3. Daily temperature and humidity readings were provided by the Italian Air Force Meteorological Service. Apparent temperature was calculated from air temperature and dew point temperature for relative humidity.11
Study design and statistical analysis
To evaluate whether daily short-term increases of PM10, PM2.5, and NO2 levels were associated with exacerbations of autoimmune diseases in Rome, we used a time stratified case-crossover design,12 which is suitable to assess the acute effects of short-term environmental exposures by comparing days when the event occurs (cases) with the same weekdays of the month and year of the event as controls.13 Then, exposure on the case day and on case-free (control) days was contrasted to estimate the effect of interest. Results are comparable with those of the time series design, subject to a few conditions.14
We used a conditional logistic regression model to estimate the association. We explored 6 days of exposure before the event, by using cumulative lag intervals of 0–1 days (consistent with an immediate effect) and of 0–5 days (consistent with a prolonged effect).
Time trend is adjusted by design in the stratified case-crossover.14 We adjusted for other time-varying confounders including meteorological factors, influenza epidemics, and population dynamics during summer and vacation periods. In more detail, apparent temperature was adjusted for by including nonlinear terms for both warm and cold temperatures. However, though temperature is known to influence health very rapidly, warm temperatures have their maximum influence after a few days and cold temperatures, over longer periods. As a result, warm temperature was defined as days with lag 0–1 temperatures above the median and was modeled with a penalized spline, whereas cold temperatures were defined as days with lag 1–6 temperature below the median and was modeled with another penalized spline.15 Both influenza and vacationing reduce the number of exposed people, changing population numbers even within a month. A dichotomous variable was used to adjust for influenza days (identified by the weekly national influenza surveillance) and for single-day holidays,15 whereas a three-level variable was used for summer days assuming value “2” in the 2 weeks around mid-August, value “1” from mid-July to end of August and value “zero” elsewhere.16
Results are given as the percentage increase in risk (%IR) of hospital recourse or treatment change (and 95% confidence intervals [CIs]) as derived from conditional odds ratios. Treatment changes are reported for specific active substances (such as levothyroxine, hydroxychloroquine) or specific groups (such as antipsoriatic drugs) or generic groups (such as corticosteroids, nonsteroidal anti-inflammatory drugs [NSAIDs], opioids, and biological drugs).
We carried out a few sensitivity analyses to explore whether alternative choices might influence the results. The first one analyzed hospitalizations for incident autoimmune diseases as exacerbations in 2006–2014, in addition to those in already identified patients. It aimed to verify whether incident cases were as liable to be associated with air pollution as were the exacerbations in patients already diagnosed. The second analysis used individual controls as a reference group for changes in prescription medication, in other words, mean changes in the same patient, in the 2 years previous. It aimed to determine if population controls moved our estimates of exacerbations toward the null hypothesis.
All analyses were performed using STATA software (V.13.0, 2013, Texas) and R software (V.3.1.3; R Core Team, 2015. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL: http://www.R-project.org/).
Table 1 illustrates the characteristics of the subjects in the five cohorts. Autoimmune thyroiditis, rheumatoid arthritis (RA), and psoriasis were the most common diagnoses in the autoimmune cohorts (Table 1). Females outnumbered males for all diseases studied except psoriasis. The youngest patients were those affected by MS, thyroiditis, and SLE; the oldest were those affected by RA and psoriasis (Table 1).
Most exacerbations were revealed through changes in prescriptions rather than hospitalizations in all patient groups (Table 2), yet there were important differences between them. Autoimmune thyroiditis showed many more changes in prescriptions than hospitalizations with a ratio of 31:1; RA showed a ratio of 10:1, suggesting that exacerbations can be controlled by changing out-of-hospital treatment, as true for thyroiditis. The ratio between prescriptions and hospitalizations was lower for psoriasis and SLE (ratio = 5:1:1 and 3:1, respectively), suggesting that serious exacerbations of these diseases that need hospital admission might occur in some patients. Finally, there was almost the same number of prescriptions for MS as there were hospitalizations. It is worth noting here that the percentage of patients without any further contact with monitored health services after the following first diagnosis was elevated for RA (37%) and SLE (32%), lower for MS (20%), psoriasis (19%), and especially for thyroiditis (3%).
Daily mean levels of PM10, PM2.5, and NO2 were 31.7, 18.7, and 56.0 μg/m3 in the entire period (Table 3), well below the standards required by the European Union and Italian law.17 However, interquartile ranges (IQRs) revealed a daily variability greater than 10 μg/m3 (the increase in levels that is assumed capable of producing health effects) and the maximum concentrations of both PM10 and PM2.5 reached values as high as 160 and 72.3 μg/m3, respectively. Table 3 reports also mean values and distributions of temperature, relative humidity, and their combination (apparent temperature) that we analyzed as potential confounders.
Figure 1 shows the pollutant effects as both hospitalizations and drug prescription changes for each autoimmune disease. The complete results are shown in the e-Table 4; http://links.lww.com/EE/A18.
Exacerbations associated with all pollutants were revealed in thyroiditis patients (Figure 1A) as changes in levothyroxine prescriptions, both in dose increases (the highest IR was 3.42%; 95% CI = 2.15, 4.71, for NO2 increase) and in time interval decreases (the highest IR was 2.42 [1.21, 3.64], for PM2.5 increase). All changes occurred the same day or the day after exposure (lag 0–1). No increase in hospitalizations occurred.
Exacerbations due to all pollutants were revealed in SLE patients (Figure 1B) by dose increase of corticosteroids (the highest IR was 3.55%; 95% CI = 0.70, 6.49) and changes involved essentially oral corticosteroids; but there was evidence of a larger risk of changing medicine form from oral to injection (8.57%; 95% CI = 0.06, 17.81). Also, shorter intervals between prescriptions of hydroxychloroquine indicated an exacerbation of SLE (IR = 9.73%; 95% CI = 4.38, 15.35). All these changes occurred for increases in PM2.5, at lag 0–1, except hydrossicloroquine whose changes lasted up to 6 days. Dose and prescription time changes were observed also for NO2 increases at lag 0–1.
Exacerbations from all pollutants were revealed in RA patients (Figure 1C) by oral corticosteroid prescription changes, both in dose increases (the highest IR is 4.93%; 95% CI = 2.11, 7.83) and in shorter time intervals between prescriptions (the highest IR is 1.59%; 95% CI = 0.35, 2.84). Even changes in NSAIDs prescriptions indicated RA exacerbations, both in dose increases (IR = 3.39%; 95% CI = 1.14, 5.69) and in shorter time intervals (IR = 2.08%; 95% CI = 0.81, 3.36). All these above important changes occurred for rising PM2.5 levels, at lag 0–1.
Psoriasis sufferers (Figure 1D) had somewhat different patterns. Exacerbations appeared as shorter time intervals between prescriptions for corticosteroids with increased risk for both PM2.5 (2.14%; 95% CI = 0.37, 3.96) and NO2 (1.75%; 95% CI = 0.6, 2.91) or for antipsoriatic drugs with NO2 (1.82%; 95% CI = 0.67, 2.9). Treatment changes are appreciable also as a dose increase of immune-suppressant drugs (IR = 4.04%; 95% CI = 0.21, 8.01) with PM2.5. Exacerbations occurred again immediately (lag 0–1). It is worth noting the statistically significant change of corticosteroids toward less effective forms (−10.47%; 95% CI = −19.6, −0.27) which followed PM10 increases, in a longer 0–5 lag.
MS patients (Figure 1E) experienced a great risk of changing corticosteroid form (IR = 11.08%; 95% CI = 0.33, 22.98) from oral to subcutaneous, for PM2.5, at lag 0–1. This change was likely to concern corticosteroids usually assumed per os (IR = −4.41%; 95% CI = −8.26, −0.41), over a longer time of 0–5 lag. No hospitalization increases occurred.
Including incident cases among exacerbations (e-Table 5; http://links.lww.com/EE/A18) did not show important differences from the main results, apart from reducing the uncertainty of estimates when there were many incident cases, such as in thyroiditis (76% additional cases). Using individual data as the reference to assess changes in drug prescriptions (e-Table 6; http://links.lww.com/EE/A18) showed lower estimates in thyroiditis, SLE, and MS, and higher in RA and psoriasis, but did not change the direction of the associations.
We found that air pollution increased the frequency of treatment changes in patients with autoimmune diseases as an indication of exacerbation. PM2.5 was the pollutant most likely to exacerbate the conditions studied, with immediate medication changes.
We selected five autoimmune diseases using the knowledge of the specific epigenetic mechanisms for each disease. In detail, an epigenetic modulation of histone methylation in the thyroglobulin promoter has been reported in autoimmune thyroid diseases.18 DNA hypomethylation, histone hypoacetylation and hyperacetylation, and decreased and/or enhanced expression of miRNAs have been identified in SLE.19 More than one disorder of epigenetic regulation is involved in RA, causing an upregulation of matrix metalloproteinase gene expression that plays a crucial role in cartilage destruction.20 DNA methylation is involved in psoriasis, and differentially methylated regions have been identified in patients with and without skin lesions.21 Multiple sclerosis is a different matter because epigenetic mechanisms have been identified initially as changes in DNA methylation and histones modifications22 based on studies of epigenetic modifying drugs.23 Recently, it has been reported that mitochondrial dysfunction could occur in MS due to air pollutants.24
In contrast, we could not include diseases like type I diabetes, autoimmune hepatitis, and celiac disease because no univocal code was available in the ICD-9. We also excluded Parkinson’s disease because the nature of the hypothesized gene/environment interaction remains unclear.25 Finally, we excluded diseases like Sjogren syndrome and Felty syndrome due to their very low frequency in our data.
Relying on hospital data only to identify autoimmune patients could indicate an incomplete cohort, especially when a diagnosis is frequent in young outpatients, such as for thyroiditis, SLE, and psoriasis (e-Figure 1; http://links.lww.com/EE/A18). Conversely, the high percentage of patients who had no more contact with health services—following identification—suggests a possible misclassification for RA and SLE, but because we analyzed people with exacerbations, which confirm the disease, a selection of severe cases is more likely than misclassification. Both these selections of hospitalized and confirmed diseases would have restricted our cohorts restrict but not nullify the meaning of our results.
We assumed that exacerbations were more reliable to reveal the relationship between air pollutants and autoimmune diseases because confirming diseases by their exacerbations increases the validity of the diagnosis,26 and exacerbations might be diagnosed quicker than an incident disease. Although the results of the sensitivity analysis including incident cases do not show essential change with respect to the main results, a misclassification of incident diagnoses cannot however be excluded. In addition, a recent study27 showed that airborne particles were able to trigger MS relapses.
We assumed also that using pharmaceutical data offers more appropriate metrics than hospitalizations to study the relationship between air pollution and autoimmune diseases given that these occur mainly in young people and do not require frequent hospital care. A concern exists with thyroiditis because thyroxin supplement is in most cases a treatment for the chronic phase, not the acute inflammatory phase. On the other hand, the results we observed for levothyroxine are very clear for both dose and prescription time, were confirmed even in sensitivity analysis with individual controls, and are not explained by chance or bias. Therefore, exacerbations seem have a role in increasing thyroxin supplements.
The longer lag shown by hospitalizations (compared with the more immediate medication changes) suggests that only the most serious episodes, possibly not resolved by initial out-patient treatment, were hospitalized. On the other hand, when exacerbations included both medication changes and hospitalizations, interpreting changes in medication becomes difficult because hospitalization is a competing risk in our analysis. There may be competing risks even in assessing different groups of drugs.
Based on this assumption, we interpret the seemingly contradictory results observed in psoriasis and MS. We assumed that the statistically significant changes in corticosteroids toward less effective forms to treat psoriasis could be a consequence of contemporary treatment changes, rather than a direct effect of pollutants. In other words, in a multitherapy framework, a dose increase of immuno-suppressant drugs or the introduction of antipsoriatic drugs could have influenced changing corticosteroids from oral to dermatological. Similarly, the changes in corticosteroids from oral to subcutaneous administration in MS patients could be the first effect of pollutants, although reducing usual oral corticosteroids could be a consequence of the previous treatment change—a clinical protocol introduced in 2011, at the regional level, suggests treating MS relapses with systemic corticosteroids.
These hypotheses need to be tested further in analyses that consider the multidrug treatment of these diseases.
The highest risks of exacerbation are due to PM2.5, which is consistent with the role that this pollutant plays in interfering with epigenetic mechanisms.7,8 The role of both PM2.5 and NO2 is notable on autoimmune thyroiditis, RA exacerbations, and psoriasis, suggesting an involvement of traffic as the source of effective pollution. Though annual average pollutant concentrations decreased in Rome over the study period (e-Figure 1; http://links.lww.com/EE/A18), we observed that related mortality increased from 2001 to 2014,28 suggesting constant toxicity of the pollution mixture over time.
Other papers have studied the relationship between air pollutants and autoimmune diseases but they considered one disease at a time. Bernatsky et al29 found a relationship between PM2.5 and specific aspects, such as anti-DNA antibodies and cellular casts in SLE patients; Hart et al30 assessed long-term air pollutants in RA, but did not find increased risk; and Angelici et al31 found an association between PM10 and hospitalizations for incident or relapses of MS. To our knowledge, this is the first paper that addresses the issue of autoimmune diseases with a systematic approach, studying them by themselves, justified by the epigenetic mechanisms of their pathogenesis. It is also the first paper to study medication changes as possible exacerbations of air pollutants.
Our results show that short-term exposure to air pollutions might induce exacerbations in patients affected by autoimmune diseases.
Exacerbations are mostly revealed through changes in treatment, especially in young patients and in diseases usually managed out of hospital. Although our results need to be further studied to validate the indicators of exacerbation based on the treatment changes, we underline the importance of using prescription data to study the health effects of air pollution.
Conflicts of interest statement
The authors declare that they have no conflicts of interest with regard to the content of this report.
Interested researchers may send requests to access the anonymized and deidentified data and computing code write Matteo Renzi (firstname.lastname@example.org).
We thank Margaret Becker for revising the English, Mirko De Martino for his help with the assessment of drug changes, and Giovanna Cappai for her help with the validation estimates of patients’ residence from health registries.
1. Meda F, Folci M, Baccarelli A, Selmi C. The epigenetics of autoimmunity.Cell Mol Immunol201183226–236
2. Zhao M, Wang Z, Yung S, Lu Q. Epigenetic dynamics in immunity and autoimmunity.Int J Biochem Cell Biol20156765–74
3. Leoni C, Vincenzetti L, Emming S, Monticelli S. Epigenetics of T lymphocytes in health and disease.Swiss Med Wkly201523145–w14191
4. Pauley KM, Cha S, Chan EK. MicroRNA in autoimmunity and autoimmune diseases
5. Breton CV, Byun HM, Wenten M, Pan F, Yang A, Gilliland FD. Prenatal tobacco smoke exposure affects global and gene-specific DNA methylation.Am J Respir Crit Care Med2009180462–467
6. Zhang H, Zhu Z, Meadows GG. Chronic alcohol consumption decreases the percentage and number of NK cells in the peripheral lymph nodes and exacerbates B16BL6 melanoma metastasis into the draining lymph nodes.Cell Immunol20112662172–179
7. Baccarelli A, Wright RO, Bollati V, et al. Rapid DNA methylation changes after exposure to traffic particles.Am J Respir Crit Care Med20091797572–578
8. Madrigano J, Baccarelli A, Mittleman MA, et al. Prolonged exposure to particulate pollution, genes associated with glutathione pathways, and DNA methylation in a cohort of older men.Environ Health Perspect20111197977–982
9. Hashimoto S, Ogoshi K, Sasaki A, et al. Coordinated changes in DNA methylation in antigen-specific memory CD4 T cells.J Immunol201319084076–4091
10. Stafoggia M, Colais P, Serinelli M; Gruppo collaborativo EpiAirMethods of statistical analysis to evaluate the short term effects of air pollution
in the EpiAir Project.Epidemiol Prev2009336 suppl 153–63. Italian
11. Steadman RG. The assessment of sultriness. Part I: a temperature-humidity index based on human physiology and clothing science.J Appl Meteorol197918861–873
12. Maclure M. The case-crossover design: a method for studying transient effects on the risk of acute events.Am J Epidemiol19911332144–153
13. Levy D, Lumley T, Sheppard L, Kaufman J, Checkoway H. Referent selection in case-crossover analyses of acute health effects of air pollution
14. Lu Y, Zeger SL. On the equivalence of case-crossover and time series methods in environmental epidemiology.Biostatistics200782337–344
15. Stafoggia M, Forastiere F, Faustini A, et al. Susceptibility factors to ozone-related mortality – a population-based case-crossover analysis.Am J Respir Crit Care Med20101823376–384
16. Belleudi V, Faustini A, Stafoggia M, et al. Impact of fine and ultrafine particles on emergency hospital admissions for cardiac and respiratory diseases.Epidemiology201021414–423
17. Directive 2008/50/EC of the European parliament and of the Council of the European Union on ambient air quality and cleaner air for Europe, adopted by Italian legislative decree 155/2010.
18. Tomer Y. Mechanisms of autoimmune thyroid diseases: from genetics to epigenetics.Annu Rev Pathol20149147–156
19. Wu H, Zhao M, Chang C, Lu Q. The real culprit in systemic lupus erythematosus: abnormal epigenetic regulation.Int J Mol Sci201516511013–11033
20. Araki Y, Mimura T. Matrix metalloproteinase gene activation resulting from disordred epigenetic mechanisms in rheumatoid arthritis.Int J Mol Sci2017185E905
21. Zhang P, Zhao M, Liang G, et al. Whole-genome DNA methylation in skin lesions from patients with psoriasis vulgaris.J Autoimmun20134117–24
22. Khorshid Ahmad T, Acosta C, Cortes C, Lakowski TM, Gangadaran S, Namaka M. Transcriptional regulation of brain-derived neurotrophic factor (BDNF) by methyl CpG binding protein 2 (MeCP2): a novel mechanism for re-myelination and/or myelin repair involved in the treatment of multiple sclerosis (MS).Mol Neurobiol20165321092–1107
23. Gangisetty O, Murugan S. Epigenetic modifications in neurological diseases: natural products as epigenetic modulators a treatment strategy.Adv Neurobiol2016121–25
24. Esmaeil Mousavi S, Heydarpour P, Reis J, Amiri M, Sahraian MA. Multiple sclerosis and airpollution exposure: mechanisms toward brain autoimmunity.Med Hypotheses201710023–30
25. Feng Y, Jankovic J, Wu YC. Epigenetic mechanisms in Parkinson’s disease.J Neurol Sci20153491–23–9
26. Faustini A, Canova C, Cascini S, et al. The reliability of hospital and pharmaceutical data to assess prevalent cases of chronic obstructive pulmonary disease.COPD201292184–196
27. Roux J, Bard D, Le Pabic E, et al. Air pollution
by particulate matter PM10 may trigger multiple sclerosis relapses.Environ Res2017156404–410
28. Renzi M, Stafoggia M, Faustini A, et al. Analysis of temporal variability in the short-term effects of ambient air pollutants on nonaccidental mortality in Rome, Italy (1998–2014).EHP20171256067019
29. Bernatsky S, Fournier M, Pineau CA, Clarke AE, Vinet E, Smargiassi A. Associations between ambient fine particulate levels and disease activity in patients with systemic lupus erythematosus (SLE).Environ Health Perspect2011119145–49
30. Hart JE, Källberg H, Laden F, et al. Ambient air pollution
exposures and risk of rheumatoid arthritis.Arthritis Care Res (Hoboken)20136571190–1196
31. Angelici L, Piola M, Cavalleri T, et al. Effects of particulate matter exposure on multiple sclerosis hospital admission in Lombardy region, Italy.Environ Res201614568–73