Technical report 122

Titanium dioxide

The human pulmonary studies of titanium dioxide are largely limited to case reports of one or more highly exposed individuals that detail the location of large amounts of titanium dioxide in the tissues. Interpretation of these studies is complicated by co-exposures to other compounds (e.g. cigarette smoke and silica) and a lack of information regarding the estimated delivered pulmonary doses. Therefore, clearance kinetics following acute and chronic exposure to titanium dioxide is poorly characterised in humans relative to animals (IARC, 2010).

Three epidemiological cohort studies and one population-based case–control study from North America and Western Europe were available for evaluation of titanium oxide exposure (IARC, 2010).

The largest of the cohort studies was among white male production workers in the titanium dioxide industry in six European countries. The study indicated a slightly increased risk for lung cancer compared with the general population. However, there was no evidence of an exposure–response relationship within the cohort. No increase in the mortality rates for kidney cancer was found when the cohort was compared with the general population, but there was a suggestion of an exposure–response relationship in internal analyses. The other cohort studies, both of which were conducted in the USA, did not report an increased risk for lung cancer or cancer at any other site; no results for kidney cancer were reported, presumably because there were few cases (IARC, 2010).

One population-based case–control study conducted in Montréal did not indicate an increased risk for lung or kidney cancer (IARC, 2010).

In summary, the studies do not suggest an association between occupational exposure to titanium dioxide as it occurred in recent decades in Western Europe and North America and risk for cancer (IARC, 2010).

All the studies had methodological limitations; misclassification of exposure could not be ruled out. None of the studies was designed to assess the impact of particle size (fine or ultrafine) or the potential effect of the coating compounds on the risk for lung cancer (IARC, 2010).

Respiratory effects that have been observed among groups of titanium dioxide exposed workers include a decline in lung function, pleural disease with plaques and pleural thickening, and mild fibrotic changes. However, the workers in these studies were also exposed to asbestos and/or silica (IARC, 2010).

No data were available on the genotoxic effects in titanium dioxide-exposed humans. Many data on deposition, retention and clearance of titanium dioxide in experimental animals are available for the inhalation route. Titanium dioxide inhalation studies showed differences—both for normalised pulmonary burden (deposited mass per dry lung, mass per body weight) and clearance kinetics—among rodent species including rats of different size, age and strain. Clearance of titanium dioxide is also affected by pre-exposure to gaseous pollutants or co-exposure to cytotoxic aerosols. Differences in dose rate or clearance kinetics and the appearance of focal areas of high particle burden have been implicated in the higher toxic and inflammatory lung responses to intratracheally instilled versus inhaled titanium dioxide particles. Experimental studies with titanium dioxide have demonstrated that rodents experience dose-dependent impairment of alveolar macrophage-mediated clearance. Ultrafine primary particles of titanium dioxide are cleared more slowly than their fine counterparts (IARC, 2010).

Titanium dioxide causes varying degrees of inflammation and associated pulmonary effects including lung epithelial cell injury, cholesterol granulomas and fibrosis. Rodents experience stronger pulmonary effects after exposure to ultrafine titanium dioxide particles compared with fine particles on a mass basis. These differences are related to lung burden in terms of particle surface area, and are considered to result from impaired phagocytosis and sequestration of ultrafine particles into the interstitium.

Fine titanium dioxide particles show minimal cytotoxicity and inflammatory/profibrotic mediator release from primary human alveolar macrophages in vitro compared with other particles. Ultrafine titanium dioxide particles inhibit phagocytosis of alveolar macrophages in vitro at mass dose concentrations at which this effect does not occur with fine titanium dioxide (IARC, 2010).

In vitro studies with fine and ultrafine titanium dioxide and purified DNA show induction of DNA damage that is suggestive of the generation of reactive oxygen species by both particle types. This effect is stronger for ultrafine than for fine titanium dioxide, and is markedly enhanced by exposure to simulated sunlight/ultraviolet light (IARC, 2010).

In vivo studies have shown enhanced micronucleus formation in bone marrow and peripheral blood lymphocytes of intraperitoneally instilled mice. Increased Hprt mutations were seen in lung epithelial cells isolated from titanium dioxide-instilled rats. In another study, no enhanced oxidative DNA damage was observed in lung tissues of rats that were intratracheally instilled with titanium dioxide (IARC, 2010).

Most in-vitro genotoxicity studies with titanium dioxide gave negative results (IARC, 2010).

A cohort of 3,607 workers employed in three DuPont titanium dioxide production facilities was followed from 1935 through 2006 (Ellis et al, 2010, 2013). The cohort included workers employed at least 6 months (183 days) on or after January 1, 1935 and prior to January 1, 2006 at any of the three DuPont TiO2 facilities. In addition, the worker had to have a job history, which resulted in exposure to TiO2 or TiCl4 based on the exposure assessment. If exposure to TiO2 or TiCl4 was unknown for more than 5 years or greater than 25% of the employment period, the worker was excluded from the study. For the Edgemoor plant, study eligibility was determined by considering work history prior to 1935, but entry into the study could not be earlier than January 1, 1935. Among the 8,359 workers identified as ever working at the three plants, 3,607 met the criteria for inclusion in the study. Workers with unknown gender or ethnicity were included in the study.

The exposure assessment was performed by an industrial hygienist experienced in historical occupational exposure assessment. Working with knowledgeable industrial hygienists at each site, we performed a walkthrough of each plant to gain knowledge of potential exposures throughout the processing operations and conducted interviews with long-term employees knowledgeable in plant operations to gain insight into changes in the plant operations over time. Detailed job descriptions which covered various time periods of plant operations as well as hard copy and electronic chemical monitoring data for a wide variety of materials were collected from each plant. The 3,488 industrial hygiene-monitoring records for TiO2 and TiCl4 collected from 1974 through 2002 were used in the exposure assessment.

Because of the focus on respiratory disease and the presence of asbestos in the plants, potential for asbestos was also determined for each worker as an ever/never exposure.

For TiO2 and TiCl4, the cumulative occupational exposure for each material was calculated for each job increment as the product of the number of calendar days in the job and the exposure level value specific for the job. The exposure level value was set as the midpoint of the range defined for each level of exposure. Annual cumulative occupational exposures were determined for each year of employment.

SMRs for all five outcomes of interest were less than 1 when compared to the US population. When stratified by plant, only the Edgemoor plant had any elevated SMRs for all malignant neoplasms, lung cancer, and all heart disease. None of these results was statistically significant. For the other two plants, all SMRs were statistically significantly reduced except the SMR for non-malignant respiratory diseases. In contrast to the results from the comparison to the US population, all the SMRs were elevated when the whole cohort was compared to other DuPont workers. Three outcomes had statistically significantly increased mortality: all causes, all malignant neoplasms, and lung cancer. The plant-specific comparison to the regional DuPont employee population produced significantly increased SMRs for four of the five outcomes among workers at the Edgemoor plant. For NJV and DL combined, all SMRs were less than 1.

When TiO2 exposure was not lagged, there was no evidence of an increase in the RR with increasing TiO2 exposure for any outcome. Although the RR was greater than 1 at every level for all five outcomes compared to the referent, the CIs were wide and overlapped across all levels. Only three of the 20 RR estimates were statistically significantly elevated. Two of the three occurred for all causes (exposure groups 15–35 and 80+ mg/m3year) and the third for heart disease (exposure group 15–35 mg/m3 year). There was no evidence of increasing risk with increasing exposure for any of the five outcomes. When exposure was lagged 10 years, results were similar with the RR estimates similar or slightly higher for each exposure level. For all causes, the RR estimates were statistically significantly increased for every exposure level except 35–80 mg/m3 year.

For heart disease, the RR estimates were statistically significantly increased at two exposure levels (5–15 and 15–35 mg/m3 year) and the lower bound was 1 for 80+ mg/m3 year. The only outcome with an indication of a positive relationship between risk and exposure was non-malignant respiratory disease with exposure lagged 10 years, but all the CIs overlapped and included unity.

For TiCl4 with no exposure lag, the trend with exposure was the same as TiO2, namely no positive association between exposure and disease outcome with CIs overlapping across exposure levels for each of the five outcomes. Looking at results for specific outcomes, lung cancer was the only outcome with increased RR estimates at all exposure levels. For all cancers similar results were seen except that one level had a reduced RR estimate (exposure group 1–5 mg/m3 year). For all causes, the RR estimates were less than 1 at all exposure levels. For non-malignant respiratory and heart diseases, three of the four RR estimates were less than 1. When the analysis was repeated using a 10-year lag, the overall results were the similar but the RR estimates tended to be higher.

Combined and plant-specific cohort mortality was compared with the overall US population and other DuPont employees. The relationships between selected causes of death and annual cumulative exposures to titanium dioxide and chloride were investigated using Poisson regression methods to examine trends with increasing exposure.

Among the 833 deaths, no causes of deaths were statistically significantly elevated either overall or plant-specific when compared to the US population. Compared to DuPont workers, statistically significantly elevated SMRs for all causes, all cancers, and lung cancers were found driven by the workers at the oldest plant. Comparing increasing exposure groups to the lowest group, disease risk did not increase with exposure.

There was no indication of a positive association between occupational exposure and death from all causes, all cancers, lung cancers, non-malignant respiratory disease, or all heart disease.  The results of this study were similar to those in previous studies of these DuPont workers. The 3,607 workers in the current study overlap with the 5,054 workers at these three plants from our earlier study (Ellis et al, 2010). While the previous study criteria only required work in the process area at any of these three DuPont plants for 6 months, the current study was more restrictive. Although these restrictions reduced the size of the study cohort substantially, they were applied in an effort to reduce the uncertainty in the exposure estimates. Retaining the excluded workers and assigning their unknown exposure years to the exposure control group would likely have resulted in bias towards the null, which may have masked any increased risk from exposure.

The restrictions were also made to facilitate comparison to results from the two other cohort studies of TiO2 workers (Fryzek et al, 2003; Boffetta et al, 2004). Similar to these two studies, the current study was restricted to workers with employment of at least 6 months, a job that had potential for exposure to TiO2, and not more than 25% or 5 years missing job history. Although a large percentage of the workers in our previous study were excluded from this study, the workers in the current study had similar characteristics to our earlier study with respect to gender, ethnicity, birth year, age at entry, length of follow up, and plant of employment. The results of the exposure analysis were also similar to those for the previous studies of Chen and Fayerweather (1988) and Fayerweather et al (1992), which found no association between lung cancer and TiO2 or TiCl4 regardless of whether TiO2 or TiCl4 exposure assessment was based on a time-weighted average, exposure duration or cumulative exposure index.

The statistically significant SMRs seen in the study cohort when the referent was other DuPont workers did not result from the truncation of the analysis to 1955. When we repeated the SMR analysis using the US population as the referent truncating the rates at 1955, the results were similar to the analysis using US rates beginning at 1935. The increased SMRs using a DuPont referent were reported in the previous study of the Edgemoor by Chen and Fayerweather (1988). Similar results have been seen for studies of other DuPont worker cohorts when US and DuPont rates were used as referents in the same study (Leonard et al, 2007, 2008). This difference in the SMRs using a general population versus an employed population as the referent demonstrates the bias associated with the healthy worker effect (Monson, 1986).

The increased SMRs consistently observed for the Edgemoor workers using the DuPont referent rates may result from the way these rates were assembled. Although the DuPont death registry began in 1957, only deaths for pensioned and active workers were included prior to 1979. With the inception of the  NDI in 1979, deaths for non-pensioned former workers were added. Since our SMR study began with DuPont reference rates for the 1955–1959 time period and includes deaths for all workers regardless of pension status, using the DuPont referent rates would underestimate the number of expected deaths for our study cohort which would increase the estimates of the SMR.

The limitations include lack of smoking history information when the outcomes of interest include lung cancer and non-malignant respiratory disease, reconstruction of work histories resulting from different practices in recording and retaining work history both within and between plants, lack of information on ethnicity, and reliance on work history and cause of death data from Chen and Fayerweather (1988) without complete documentation on how the data for the study were assembled.

Strengths include the completeness of vital status ascertainment and collection of cause of death information as well as the availability of multiple sources of information to capture the data needed to complete this study.

With no exposure data available prior to 1975, the likelihood of misclassification of exposure was increased. This lack of data prior to the 1970s is common in occupational epidemiologic studies since monitoring was limited prior to the imposition of regulatory guidelines in that era.

However, the hazards assessment was done by an industrial hygienist experienced in historical occupational exposure reconstruction with input from knowledgeable individuals at each plant. There is no reason to believe that any systematic bias occurred since the exposure assignment for each worker was done without knowledge of the worker’s vital status or cause of death. As a result, the effect of any exposure misclassification would be to bias the results toward finding no effect.

In conclusion, the results of this study are consistent with those of other studies of TiO2 workers. There is no indication of a positive association between occupational exposure to TiO2 or TiCl4 and death from all causes, all cancers, lung cancer, non-malignant respiratory disease or all heart disease.