Article Text


Exposure to metal welding fume particles and risk for cardiovascular disease in Denmark: a prospective cohort study
  1. Else Ibfelt1,
  2. Jens Peter Bonde2,
  3. Johnni Hansen1
  1. 1Institute of Cancer Epidemiology, Danish Cancer Society, Copenhagen, Denmark
  2. 2Department of Occupational and Environmental Medicine, Copenhagen University Hospital, Bispebjerg, Denmark
  1. Correspondence to Else Ibfelt, Institute of Cancer Epidemiology, Danish Cancer Society, 49 Strandboulevarden, Copenhagen DK-2100, Denmark; else{at}


Objectives To study welding fume particles in relation to cardiovascular diseases.

Methods In 1986, 10 059 male metal workers in 75 welding companies were sent a questionnaire about their welding experience and lifestyle (83.3% response rate). Of these, 5866 were available for analysis and had ever welded at baseline. Information on exposure to welding fumes after 1986 was obtained by individual linkage to the National Pension Fund. Lifelong exposure to welding fume particles was estimated from a job–exposure matrix based on more than 1000 welding-specific measures of fume particles. Hospital contacts for cardiovascular disease were obtained from the Danish National Patient Registry by individual linkage. The nine disease outcomes considered were acute myocardial infarct (AMI), angina pectoris, other acute ischaemic heart diseases, chronic ischaemic heart disease (CHD), cardiac arrythmias, cardiac arrest, heart failure, cerebral infarct, arterial embolism and thrombosis. The cohort was followed up from baseline until the end of 2006.

Results When the incidence of each of the nine cardiovascular outcomes among welders was compared with 5-year age- and calendar year-specific male national rates, the number of observed cases significantly exceeded that expected for AMI (standardised incidence ratio, 95% CI) (1.12, 1.01 to 1.24), angina pectoris (1.11, 1.01 to 1.22), CHD (1.17, 1.05 to 1.31) and cerebral infarct (1.24, 1.06 to 1.44). Internal comparisons of the cohort with adjustment for tobacco smoking, alcohol and hypertension medicines showed a significantly increasing hazard rate ratio for CHD and non-significant increases for AMI, angina pectoris and cerebral infarct with increasing exposure to particles.

Conclusions This study supports the hypothesis that exposure to welding processed particles increases the risk for cardiovascular disease.

  • Occupational exposure
  • ultrafine particles
  • cardiovascular diseases
  • epidemiology
  • particulates
  • welding

Statistics from

What this paper adds

  • Few studies have investigated the relationship between welding and the risk for cardiovascular diseases; most found an increased risk, although some did not find this association.

  • Generally, previous studies lacked detailed exposure measurements, combined many cardiovascular diseases together, and rarely adjusted for confounding.

  • This study examines lifelong welding particulate exposure in relation to risk for nine cardiovascular outcomes adjusted for confounding by smoking, alcohol and medicine use.

  • Our study suggests that welders are at increased risk for ischaemic and cerebrovascular diseases.


Several studies conducted in Europe and the USA during the past two decades have shown an association between the concentration of ultrafine particles in ambient air and risk for cardiovascular disease (CVD).1 Welders are exposed to much higher concentrations of particles than the general public2 3; investigation of CVD in this population, therefore, contributes to elucidating associations between ambient air pollution and cardiovascular risk.

Metal welding consists of several processes that may result in substantial exposure to particles, fumes and gases. It is estimated that more than 1 million workers worldwide perform welding as a part of their job.4 Particles in general are a source of local irritation, as their physical, chemical or biological properties can initiate an inflammatory response. The inflammation may be acute and easily healed, or it may become chronic.5 Studies of exposure to welding fumes have shown that the particles are in the ultrafine size range of 0.01–0.10 μm,6 and approximately 50%–75% are of submicron size, with a diameter of 0.4–0.8 μm.7 As smaller particles penetrate further into the airways, gases and other active species attached to their surfaces are carried deeper into the lung.8 The hypothesis has been proposed that inhaled particles create pulmonary inflammation and a subsequent systemic inflammatory response. This will increase the probability of atherosclerosis, followed by other cardiovascular events.9 The compounds of the welding emission formed during the processes are also suspected of increasing CVD, due to the formation of, for example, ozone and carbon monoxide.

Studies of ambient air pollution, in which the level of exposure to ultrafine particles is usually much lower than that experienced during welding (at least 10–100-fold, depending on the welding method and the place and time of air measurements2 10 11), have shown acute effects on heart rate, blood pressure and blood coagulation and effects of more prolonged (chronic) exposure to air pollution particles on the progression of atherosclerosis.1 12

Some of the few available studies on welding have shown increased morbidity and mortality due to CVD,13–17 whereas others have shown no effect.18 19 Previous studies of welders and CVD lacked detailed exposure measurements (eg, duration, mass or number of particles inhaled), combined all diagnoses of cardiovascular or ischaemic heart disease, or recorded only mortality from CVD, and few studies included adjustment for confounding by, for example, tobacco smoking. In the study reported here, we estimated lifelong exposure to welding fume particles, analysed dose–response relationships for different diagnoses of cardiovascular diseases and attempted to adjust for individual lifestyle risk factors.


Study population

Potential study participants were selected from a cohort of 10 059 male production workers that was established in 1986 with the aim of studying associations between metal welding and lung cancer. The various steps of the construction of the cohort have been described in detail elsewhere20 and are only summarised here. In 1982, employees of 79 companies in Denmark with metal welding activities were identified in a survey. After exclusion of shipyards, because of possible exposure to asbestos, 75 companies remained, employing about 60% of all Danish stainless-steel welders; five of the larger companies had substantial numbers of mild-steel welders. All men born before 1964 and employed by one of these companies in the period April 1964–December 1984 for a minimum of 1 year were then identified from the computerised files of the nationwide Danish Pension Fund.20 This pension scheme, with compulsory membership for all wage earners aged 16–66 years, contains the person's name, the dates of the start and end of each employment at company level (a unique company identification number is assigned for tax purposes) and the unique personal identification number assigned by the Central Population Registry to all residents of Denmark.21 These two registers are regarded as both accurate and complete.21 In order to identify workers potentially exposed to welding, all the companies were visited in 1986, and employees in departments with welding activities were identified from independent information in company records, foremen, long-term workers and other sources. Information on emigration, disappearance and death for each worker identified, was retrieved from the files of the Central Population Registry, which was established in 1968. In autumn 1986, all the workers (or spouses or long-term colleagues of deceased or emigrated workers) were sent questionnaires eliciting information on lifetime exposure to welding and other exposures such as tobacco smoking, use of medications and alcohol consumption. Non-respondents were contacted up to three times. Responses were received from 8376 workers (83.3%). Of these, 6162 (74.6%) had ever welded at baseline according to the questionnaire. We did not use men who had never welded as an internal comparison group, as this group was small and heterogeneous. We excluded three welders with inconsistent personal identification numbers, eight welders who immigrated to Greenland and 285 welders who had died before 1 January 1987, leaving 5866 men for the analyses.

Disease experience

Information on the occurrence of CVD in this population was retrieved by linkage to the Danish National Patient Registry by personal identification numbers. The welders were followed up from the start of 1987 until the end of 2006. This computerised register, with high validity,22 stores information on all persons hospitalised in Denmark since 1 January 1977 and includes dates of admission and discharge, codes for the primary diagnosis and up to 19 codes for additional diagnoses. Information on outpatients was added on 1 January 1994. Until the end of 1993, diagnoses were coded according to an extended Danish version of the International Classification of Diseases, Revision 8 (ICD-8), and after that date according to ICD-10. We defined the first occurrence of CVD as the date of first hospitalisation with CVD as a main diagnosis after 1 January 1987; if a similar diagnosis had been made earlier as a secondary diagnosis, however, we used this to define the time of diagnosis. We excluded 236 welders' hospitalisations with CVD diagnoses (main or secondary) that had occurred between 1977 and the start of follow-up.

We considered nine main disease outcomes from ICD-8 and ICD-10: acute myocardial infarct (AMI) (410, I21), angina pectoris (413, I20), other acute ischaemic heart diseases (411, I24), chronic ischaemic heart disease (CHD) (412, I25), cardiac arrythmias (427, I48 and I49), cardiac arrest (none, I46), heart failure (none, I50), cerebral infarct (433 and 434, I63) and arterial embolism and thrombosis (444, I74).

Assessment of exposure to welding fumes

The questionnaire elicited information on the welding material used (stainless or mild steel), welding process (manual metal arc, tungsten inert gas, metal inert gas, metal active gas, spot or gas welding), first year of welding, number of years welding in various decades, use of exhaust ventilation and welding in confined spaces. In an attempt to cover any exposure to welding fumes after the baseline in 1986 and thereby obtain more complete information about lifelong welding, we included data from the National Pension Fund register on duration of employment in welding companies after 1986.

To estimate total accumulated exposure to welding fume particles before baseline, we used an exposure matrix based on more than 1000 measurements of ambient air particles in the workplace between 1971 and 1985 made by the Danish Welding Institute and the National Institute of Occupational Health. Air samples were collected on filters placed in the breathing zone behind welding helmets. Filters used were pore size 0.8 μm, so particles smaller than this did not contribute to the mass measurements. Earlier exposures were estimated by extrapolation, on the assumption of a declining trend in exposure in all welding processes during 1971–1985. The database provided geometric mean values (mg/m3) for exposure to particulates in each welding process.23

By linking the exposure matrix to data from the questionnaires on decade, type of steel, welding process, frequency, exhaust ventilation and welding in confined spaces, we were able to compute the total of number of years that each man had spent in welding and the particulate concentration in each decade and for all decades. This gave a summary measure of total exposure to welding fumes (mg/m3×years) before baseline. To obtain a measure of total lifelong exposure before and after baseline, we calculated an average exposure for 1980–1986 (the last interval of the questionnaire data) and multiplied it by number of years in the welding industry after baseline. Total exposure was calculated from data for 3499 welders for whom consistent information on welding was available from the questionnaire, and especially details on years, processes, etc, for the period 1980–1986.

Statistical analysis

The risk for CVD was calculated from 1 January 1987, and follow-up ended on the date of hospitalisation for CVD, death, emigration from Denmark or 31 December 2006, whichever came first.

In the external analysis, we compared the incidence of CVD among welders with 5-year age- and calendar year-specific national rates among men, calculated from the CVD diagnosis in the files of the Danish National Patient Registry. The expected number of CVD diagnoses was calculated by applying the national gender-, age- and calendar time-specific rates to the appropriate person-years under observation in each exposure group. Standardised incidence ratios (SIRs) were calculated as the observed incidence divided by that expected with a modified version of the Person-Years (PYRS) software program.24 Significance and 95% CIs were determined on the basis of the assumption that the observed number of diagnoses of heart disease in any category was Poisson distributed.

In the internal analyses, we compared the rates of CVD among welders by number of years of welding with a lag time of 1 year and by estimated concentration of particles. We adjusted for the effects of tobacco smoking (never, former or current smoker), alcohol consumption (0–3.5, 4–20.5 or ≥21 drinks per week), regular use of hypertension or ‘heart’ medicine within 1 year prior to baseline (yes or no) and calendar time. Tests for interaction between covariates were performed for one pair at a time using the Wald test statistics. No statistical significant interactions were observed between exposure and covariates. We performed sensitivity analyses after excluding welders who had started welding before 1960 in order to minimise selection bias due to the healthy worker effect. Cox regression analysis was used to estimate hazard rate ratios (HRRs) for heart disease, a measure of RR, with the Breslow method for ties. Age was used as the underlying time scale to ensure that estimates were based on comparisons of individuals of the same age. Estimated lifelong exposure to welding particles (mg/m3×years) was included as a time-dependent variable, as duration of welding included employment time both before and after the study baseline. Two-sided 95% CIs were calculated for the HRRs with the Wald test of the Cox regression parameter. The proportional hazard assumption for the analysis was tested by Kaplan–Meier plots and the Schoenfeld test of proportional hazard assumption. Cox regression analyses were performed with Stata statistical software v 9.2.25


Of the 5866 welders, 1126 died before the end of follow-up, 51 emigrated from Denmark and 4689 were still alive at the end of follow-up. Selected descriptive characteristics of the welders are shown in table 1.

Table 1

Characteristics of 5866 welders at baseline, Denmark, 1987

Table 2 gives the results of the external analysis. For four of the nine CVD outcomes, the number of observed cases significantly exceeded the number expected. We observed an increasing risk with higher levels of particulate exposures for angina pectoris, CHD and cerebral infarct, but the increase was significant only for cerebral infarct in the group exposed to >100 mg/m3×years. Increased risks for AMI and CHD were also seen for a group of welders for whom inadequate data were available for calculation of cumulative exposure.

Table 2

Standardised incidence ratios (SIRs) and 95% CIs for selected cardiovascular diseases among 5866 male Danish welders followed up from 1986 to 2006 and compared with the general male population

In the internal analyses, we found no association between total years of welding (duration) and risk for AMI, angina pectoris, CHD or cerebral infarct, before or after adjustment for calendar year, tobacco smoking, alcohol consumption and use of hypertension or ‘heart’ medicine (data not shown).

Table 3 gives the estimates for cumulative exposure to particulates after adjustment for potential confounders. A significantly higher RR for CHD was found for welders exposed to 10–50 mg/m3×years (HRR 2.51, 95% CI 1.15 to 5.49) and 50–100 mg/m3×years (2.79, 1.29 to 6.04) than those exposed to 0–10 mg/m3×years (reference), while the risk for the group with the highest exposure, >100 mg/m3×years, was not significantly increased (1.70, 0.78 to 3.72). Similar tendencies for an exposure–response relationship were found for AMI, angina pectoris and cerebral infarct, although these were not significant. The unadjusted results (data not shown) were similar to those found after adjustment.

Table 3

Adjusted HRRs and 95% CIs for cardiovascular disease among 3499 welders according to accumulated exposure to particles

When analysing exposure generated from data before baseline only, the results were similar to those generated from lifelong exposure shown in table 3, but with lower RR estimates. For the three exposure categories 10–50, 50–100 and >100, the HRRs for AMI were 1.04 (0.70 to 1.53), 1.31 (0.89 to 1.93) and 1.01 (0.68 to 1.52), for angina pectoris 1.04 (0.72 to 1.49), 1.27 (0.88 to 1.81) and 1.09 (0.75 to 1.58), for CHD 1.27 (0.82 to 1.96), 1.48 (0.96 to 2.28) and 0.95 (0.61 to 1.50) and for cerebral infarction 1.19 (0.65 to 2.17), 1.05 (0.57 to 1.95) and 1.39 (0.76 to 2.52).

After exclusion of welders who had started welding before 1960, the estimates for the group with the highest exposure (>100 mg/m3×years) were increased, with HRRs of 1.24 (0.66 to 2.36) for AMI, 1.28 (0.68 to 2.40) for angina pectoris, 1.87 (0.72 to 4.85) for CHD and 2.14 (0.77 to 5.92) for cerebral infarct. The tests for trend, however, remained non-significant (data not shown).


In our cohort of 5866 welders, we found significantly higher risks for the four most frequent of nine cardiovascular outcomes (AMI, angina pectoris, CHD and cerebral infarct) than in the general population. The risks for AMI and CHD were increased in a group of welders for whom data for calculation of cumulative particulate exposure were inadequate. It is possible that this is a group of vulnerable welders who were unable to fill in the questionnaire or did not work between 1980 and 1986 for various reasons.

When comparing the welders internally, we also found an approximate dose–response relationship between cumulated exposure to particulates and each of the four diagnoses, although the trend was not significant. The group with the highest exposure had non-significantly lower risks for AMI, angina pectoris and CHD than the groups with intermediate exposure when compared with the reference. It is possible that the group with the highest exposure were selected ‘healthy workers’. When we excluded welders who had started work before 1960 in order to reduce this possible selection bias, the risk for the group with the highest exposure increased but was still not significantly higher than that of the group with the lowest exposure. This may indicate that there is a threshold of exposure to particles, a ‘healthy worker’ survivor effect or misclassification of exposure.

One limitation is that particles less than 0.8 μm in size were not included in the job-exposure matrix. Another is that the effect of specific compounds involved in the welding processes is not represented in the job–exposure matrix algorithms used. In particular, carbon monoxide and ozone generated from shielding gases used in various welding processes are toxic to the respiratory system, organs and blood. Carbon monoxide may have an acute effect because of the formation of carboxyhaemoglobin and have long-term effects via a possible atherogenic mechanism.26 Ozone has been shown to affect levels of the cytokine interleukin-6 in the bronchoalveolar lavage, which might increase concentrations of plasma fibrinogen and thereby the risk for ischaemic heart disease.16 Even though the questionnaire provided information on specific welding processes and materials, it was not possible to construct ‘clean’ groups by type of welding (eg, manual metal arc or metal active gas) or by exposure to specific compounds, since the majority of the welders worked with different methods over the years.

The major strengths of our study are the large cohort of relatively young welders, detailed, prospectively obtained information about welding processes and the long follow-up of nearly 30 years. In Denmark, all residents have equal, free admission to hospital, and, through the Danish National Patient Registry, we were able to obtain information on all relevant diseases that required hospitalisation. One limitation of the study is that we did not have information on men with cardiovascular disorders who were not admitted to hospital; for instance, angina pectoris is sometimes treated by general practitioners. As it is unlikely that welders are admitted more or less often to hospital for a particular disease than the general population, we consider that this limitation does not reduce the validity of the study. Furthermore, we considered disease outcomes that are usually treated in hospital.

Information on exposure to welding fumes before baseline was obtained from self-reports by welders up to 40 years later. Therefore, the details of welding processes, time spent welding and working conditions might not be totally accurate for the oldest group. Calculation of cumulative exposure to particulates with the job–exposure matrix and responses to a questionnaire is not as precise as individual measurements of inhaled particles for each welder. Therefore, some differential misclassification of exposure might have occurred, resulting in underestimation of the risk for CVD among the most heavily exposed men.27 28

The addition of information on estimated duration of welding after baseline to the data from the questionnaire strengthens the association of exposure during welding and CVD. The compulsory routine registration of employment in the Danish Pension Fund register ensures that there is no selection of individuals or missing or incorrect reporting of employment. One limitation in the calculation of cumulative exposure after baseline was the lack of information on whether the work specifically included welding and whether the welders continued to have the same hours and processes reported in 1986. Also, we assumed that there was no change in exposure after 1980–1986. The semi-quantitative estimate of exposure to particulates is therefore an approximate measure of the true lifelong exposure. We consider, however, that it is a more informative, complete measure than measures based on total number of years welding (used in other studies) or cumulative exposure calculated only from data before baseline.

True risk can be underestimated when the morbidity rates of welders are compared with those of the total population (given no confounding from other exposures), because the general population includes sick and disabled people who are unable to work.16 In our internal analysis, we were able to lessen this possible bias and potential confounding by including tobacco smoking, alcohol consumption and use of hypertension or ‘heart’ medicines. These exposures had, however, only a marginal influence on the results. Data on these lifestyle factors were only collected at one time point and we had no information about changes over time. Smoking exposure is likely to have decreased since 1986, and if so our results of welding effect may be underestimated. Alcohol consumption is likely to have stayed the same over the years for this group of people. The number of welders using hypertension or ‘heart’ medicine is likely to increase as the cohort grows older, but is only relevant for a smaller group and can be an intermediate step before experiencing a CVD diagnosis. We had no information on other risk factors for CVD, such as obesity, high intake of fatty foods, low vegetable intake and low physical activity.29 As we found an approximate dose–response relationship between exposure to welding particles and CVD, the increased risk is unlikely to be due only to other, unknown confounders.

Few other studies have investigated the risk for CVD in welders, and half of them addressed only mortality from such causes. One Danish cross-sectional study of AMI and three cohort studies of mortality due to ischaemic heart disease found positive associations with welding, the results being: OR 2.1 (95% CI 1.05 to 4.30) only for welders with blood type O, unadjusted17; standardised mortality ratio (SMR) 1.06 (1.02 to 1.11) in the 1970 census and SMR 1.35 (1.10 to 1.64) in the 1990 census, unadjusted16; SMR 1.51 (1.00 to 2.18) and SMR 1.79 (p<0.05) for welders who had worked ≥20 years, adjusted for smoking14; and SMR 1.30 (1.04 to 1.56), unadjusted.15 A cohort study of 236 workers showed increased risks adjusted for smoking for angina pectoris and chest pain, with prevalence ratios of 2.61 (1.2 to 5.7) and 2.27 (1.5 to 4.9), respectively, and a non-significant effect for AMI (prevalence ratio 1.37, 0.7 to 2.8).13 The risk estimate for AMI observed in our study is within the range of the prevalence ratio from this study. In a cohort study of ischaemic heart disease in construction workers, no effect was found (RR 1.01, 0.95 to 1.08) in 831 workers exposed to metal fumes.19 In a study of pooled data for 11 092 welders in 135 companies in nine countries, no significant effect was found (SMR 94, 86 to 103), but the inclusion criteria for different countries differed, and all diseases of the circulatory system combined (ICD-8: 390–458) were considered.18

Our finding of an increased risk for ischaemic heart disease in welders is supported by most of these studies, and the elevated risks are close to our estimates. The two studies that did not find increased risks were unable to analyse dose–response relationships, addressed all ischaemic heart disease or circulatory diagnoses or did not include adjustment for smoking. In addition, other studies found acute effects on intermediate cardiovascular outcomes after exposure to welding fumes: cardiac autonomic control,30 aortic augmentation31 and acute systematic inflammation,32 which further support our results. Studies on ambient air pollution have also shown associations with cardiac arrhythmias mostly in patients with underlying cardiac disease,33 however, we did not find such association in this study.

In conclusion, this study supports the hypothesis that exposure during welding increases the risk for cardiovascular and cerebrovascular disease. As the particles in welding fumes are likely to be of ultrafine or nano sizes, like particles from traffic and combustion pollution, we consider that our results are relevant in the discussion of the health hazards of environmental pollution from particulates in general. However, very small particles are usually measured in numbers, a quantitative measurement we did not have for our cohort of welders. Further studies on numbers of ultrafine and nano sized welding particles and diseases of the lungs and cardiovascular system will be needed to answer questions about the health hazards of these particles in general. Also questions about specific welding compounds and CVD risk could be addressed in future studies analysing air samples or questionnaires aimed at welders working with specific welding processes only.


We are indebted to Klaus S Hansen for providing access to the data on the cohort of welders, to Andrea Meersohn for help with data processing and to Kirsten Frederiksen for statistical assistance.


View Abstract


  • Funding This study was supported by a grant from the Danish Working Environment Research Fund. The funding source had no role in the design or analysis of the study or in the decision to submit the manuscript for publication.

  • Competing interests None.

  • 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.