Hydrocarbons, steam-cracking tar middle

Epidemiological data

Administrative data

Endpoint:

epidemiological data

Type of information:

migrated information: read-across based on grouping of substances (category approach)

Adequacy of study:

key study

Reliability:

2 (reliable with restrictions)

Rationale for reliability incl. deficiencies:

other: Large, well designed, human population study, with reliable exposure assessment methodology, adequate for assessment

Data source

Reference

Reference Type:

publication

Title:

Peripheral blood effects in benzene-exposed workers

Author:

Schnatter AR, Kerzic P, Zhou Y, Chen M, Nicolich M, Lavelle K, Armstrong T, Bird M, Lin l, Hua F and Irons R

Year:

2010

Bibliographic source:

Chem Biol Interact 184: 174-181.

Materials and methods

Study type:

cross sectional study

Endpoint addressed:

repeated dose toxicity: inhalation

Test guideline

Qualifier:

no guideline followed

Principles of method if other than guideline:

The study examined peripheral blood effects in 928 workers from five factories in and around Shanghai.

GLP compliance:

no

Test material

Test material information

Constituent 1

Reference substance name:

Benzene

EC Number:

200-753-7

EC Name:

Benzene

Cas Number:

71-43-2

IUPAC Name:

benzene

Details on test material:

Benzene

Method

Type of population:

occupational

Ethical approval:

confirmed and informed consent free of coercion received

Details on study design:

METHOD OF DATA COLLECTION
- Type: Questionnaire and Clinical tests:
- Details: a questionnaire was administered in-person to each study subject, who was questioned on a broad range of topics, including date of birth, tobacco and alcohol use, height, weight, hobbies, medication use, existing diseases, and exposure to livestock and/or pesticides.
- Clinical tests included urinary cotinine levels to validate responses to smoking. Ambient benzene, cotinine, creatinine and single nucleotide polymorphisms were determined. Blood samples were analysed for total white blood cells (WBC), neutrophils, basophils, eosinophils, monocytes, lymphocytes, total red blood cells (RBC), haemoglobin, mean corpuscular volume (MCV), and red cell distribution width (RDW), platelets, and mean platelet volume (MPV).

STUDY PERIOD: September 2003 to June 2007

SETTING: Five factories in Shanghai: two rubber factories (A and E), a shoe manufacturer (B), an insulation materials (rubber plank) factory (C), a pharmaceutical factory (D).

STUDY POPULATION
- Total population (Total no. of persons in cohort from which the subjects were drawn): 1078.
- Selection criteria: All workers in the selected workshops were eligible and were included in the final sample if they completed a consent form and questionnaire, and provided a blood sample.
- Total number of subjects participating in study: 1078
- Sex/age/race: Female - 34.6 years, male 42.6 years, Chinese
- Smoker/non-smoker: 11 female smokers, 372 male smokers
- Total number of subjects eligible for analysis after subjects were excluded because they were not currently working, or had hepatitis, a pre-existing blood disorder, or a blood transfusion in the prior six months: 1046 ( 563 males, 483 females). Two additional workers who did not have valid blood counts were excluded from the study. The benzene exposure index used in analysis was defined for 928 of the 1046 workers, and 855 of the 928 were categorised as exposed to benzene.

COMPARISON POPULATION
- Type: Workers from the same factories not exposed to benzene - 73 in total - Exposed and comparison subject counts by factory were as follows: Factory A: 328/13; Factory B: 321/18; Factory C: 108/1; Factory D: 80/41 and Factory E: 18/0.

HEALTH EFFECTS STUDIED
Twelve peripheral blood indices derived from a complete blood count (CBC) test. Peripheral blood was collected by venepuncture from all individuals, using standardized procedures and processed for routine CBC using an automated haematology analyzer (CellDyne 3700, Abbott, Park, IL). The twelve indices were selected based on biologic plausibility as well as previous literature on benzene. The indices were: total white blood cells (WBC), five WBC subtypes (neutrophils, basophils, eosinophils, monocytes, and lymphocytes), total red blood cells (RBC), three red cell-related measures (haemoglobin, mean corpuscular volume (MCV), and red cell distribution width (RDW)), platelets, and mean platelet volume (MPV).

Details on exposure:

Exposure was measured via 3M badges (n=734) or estimated based on workers in similar jobs (n=194). Benzene exposure was widely distributed (range = LOD/√2 = 0.07 mg/m3 (0.02 ppm) - 872 mg/m3 (273 ppm)) but right skewed. Interquartile values were 0.9 mg/m3 (0.3 ppm), 7.4 mg/m3 (2.3 ppm), and 29.5 mg/m3 (9.2 ppm).

Workers from each study factory were allocated to one of 133 similar exposure groups (SEGs), which were defined as groups of workers with similar job, locations, assignments, tasks, work patterns, schedules and materials used. A random sample of workers was then selected from each SEG to obtain individual-level measurements, resulting in 2,973 benzene samples with an average of four samples (range: 1-14) per worker. Monitored workers wore 3M® organic vapour badges on their lapel for one to sixteen hours during their work shift. Badges were analysed for benzene and toluene with limits of detection (LOD’s) of 0.1 and 0.16 mg/m3, respectively. For each monitored worker in an SEG, the individual-level benzene measurements were adjusted for the time spent performing different tasks during a typical work week schedule to calculate an individual weekly average exposure value i.e. the average 8 hour time-weighted arithmetic mean exposure for a typical working week.

For each SEG the arithmetic mean of the time-weighted averages was calculated and assigned to each individual within a SEG. The benzene exposure index that was ultimately used in the blood count analysis was based on individual weekly average readings and imputed values from the SEG weekly average reading if the latter was from a homogeneous SEG. Homogeneous SEG’s were defined (n=88, 66% of SEG’s) as having a ratio of the upper and lower 95% confidence limits for the between worker variance of 4 or less as determined by a random effects ANOVA model. Exposure distributions were compared for homogeneous vs. all other SEGs, with no patterns to suggest that a bias would result by restricting exposure data to sufficiently homogenous SEG’s. 1046 workers were assigned to an SEG and the benzene exposure index used in blood count analyses was defined for 928 of the 1046 workers, and was based on either individual weekly average readings (n=734) or the SEG weekly average if the subject was from an homogeneous SEG and was not individually monitored (n=194). A similar measure of toluene exposure was estimated for 1,008 of the 1,046 workers.

Statistical methods:

Initially correlations and simple regressions between blood parameters and three benzene indices: SEG weekly averages, individual exposure weekly averages, and individual weekly averages with imputed values from SEG averages of homogeneous SEG’s were examined. The three metrics were highly correlated (range 0.70 to 0.90) therefore, benzene exposure was estimated using individual-level weekly averages with imputed averages from sufficiently homogenous SEGs for individuals who had not been monitored. Benzene exposure levels at or below the limit of detection (LOD) for benzene (0.10), and toluene (0.16) were transformed with the formula LOD/√2. Generalized linear models were used to evaluate the associations between benzene exposure and peripheral blood parameters and urinary metabolite concentrations. Each candidate covariate was sequentially assessed by comparing the Akaike’s Information Criterion (AIC) and Bayesian Information Criterion (BIC) of nested models fitted with benzene exposure. A p-value of 0.20 was chosen to evaluate entry and exit into models. Parsimony and change in the benzene coefficient were considered in model development. Candidate confounders of the benzene/blood parameter relationships included worker age, gender, BMI, current smoking status, current alcohol use, SNP’s and SNP combinations, and co-exposure to toluene. SNP data were treated as categorical values. Change point regressions were conducted to identify benzene concentrations that produce blood count changes distinguishable from background levels for each parameter. To evaluate clinically relevant endpoints, logistic models with blood parameters using high or low out-of-range values were fitted. All analyses were performed in SAS.

Results and discussion

Results:

In models examining the relationship between logged benzene exposure and each of the twelve blood parameters, adjusted for potential covariates other than toluene exposure, there were statistically significant decrements for RBC count, haemoglobin, total platelet count, MPV, total WBC counts, monocytes lymphocytes and neutrophils per log of benzene concentration and a statistically significant increase in MCV with increasing log benzene concentration. Among the three types of blood elements (WBCs, RBCs and platelets), stronger patterns of association emerged for total RBCs and its constituents. Toluene exposure reduced the benzene effect to non-significance for the following outcomes: neutrophils, monocytes, RBC’s, haemoglobin, MCV, and platelets. Toluene and benzene show a moderately strong correlation (r=0.66), and this may have resulted in co-linearity when logged benzene and toluene were included in the same regression model. In change point regression models, the lowest estimated change points were 7.77 ppm (95% CI 2.97-20.1 ppm) for neutrophils and 8.24 ppm (95%CI 2.41-28.7 ppm) for MPV, while lymphocyte count had the highest estimated change point of 38.1 ppm (95% CI 19.7 – 74.3 ppm) (calculated from change point confidence intervals reported for log benzene mg/m3). It is noted that while there is some support in the literature for an effect on neutrophils at this level of exposure, there is no support for MPV being a particularly sensitive endpoint, although there are few other studies that have examined this endpoint.
The proportion of workers whose values fell outside the range of clinically normal limits was generally small, with a maximum of fewer than 8% for haemoglobin. There were too few workers with abnormally low readings of monocytes (N=1), MPV (n=1) or basophils (N=0) for meaningful statistical analyses via logistic regression, while no clinical norms were available for RDW. Results for logistic regression models that examined out-of-range values generally supported the continuous models in that stronger effects were seen for red cell parameters, although only abnormally high MCV values and abnormally low RBC, haemoglobin, platelets and WBC values were significantly associated with benzene exposure. To further assess dose response for the out-of-range indices that showed effects in logistic regressions, benzene exposures were categorised as <1 ppm, 1-<10ppm, and 10+ ppm. Stronger effects were seen for MCV and RBC, with highly significant OR’s in the >10 ppm exposure category. MCV showed a monotonic risk, but the dose response for RBC was irregular.

Confounding factors:

Age, gender, BMI, current smoking status, current alcohol use, SNP’s and SNP combinations, and co-exposure to toluene.

Strengths and weaknesses:

Strengths: Large study of 928 workers exposed across a wide range of benzene concentrations (interquartile range, 0.3 ppm - 9.2 ppm) and 73 unexposed workers. Reliable exposure assessment methodology based on more than 2900 individual benzene readings, supplemented by use of average measurements from other workers in the same SEG if exposure measurements for the SEG were reasonably homogeneous. Potential heterogeneity in exposure readings was accounted for by using worker time-weighted average exposures. Other possible influences on blood values were accounted for by collecting information on SNP’s for key genes, toluene exposure, demographics and lifestyle habits as well as other medical information.
Weaknesses: The investigation was cross-sectional and it was not possible to directly account for historical effects on blood parameters. No information is given about the duration and level of past benzene exposure. Outliers may have had an effect on estimates of the change-point in regression models. The five factories produced different products with potentially different co-exposures and consistency of effects between the factories was not investigated.

Any other information on results incl. tables

The study showed that a wide range of benzene exposures affect urinary metabolites and some, but not all, peripheral blood cell indices. Metabolite levels for HQ and ttMA were clearly affected at exposure concentrations of 0.5 ppm and above, with toluene co-exposure also enhancing levels of all benzene metabolites. The data suggest a 2-3-fold enhancement of metabolite formation per unit benzene exposure under 1 ppm, but also suggest that correction factors for background exposure are critical in assessing such biotransformation rate effects. For blood indices measured continuously, effects were seen on RBC, WBC, PLT, HGB, MCV, lymphocytes, neutrophils, monocytes and MPV. Toluene exposure may be a confounder for many of these reported relationships.

Table of Change point concentrations for selected blood parameters (based on Schnatter et al 2010, Table 4):

Blood Parameter	Change-point (95% CI) log benzene mg/m^3
WBC	4.18 (3.46, 4.90)
Lymphocytes	4.80 (4.14, 5.47)
Lymphocytes^1	4.78 (4.10, 5.46)
Neutrophils	3.21 (2.25, 4.16)
RBC	3.70 (3.21, 4.19)
Haemoglobin	3.43 (2.65, 4.21)
MCV	4.47 (4.12, 4.82)
Platelets	3.62 (2.97, 4.28)
MPV	3.27 (2.04, 4.52)

 ¹ Model includes term for toluene exposure

Applicant's summary and conclusion

Conclusions:

The one-sided 95% lower confidence limit for the change-point concentration for neutrophils was calculated to be 3.5 ppm (11.2 mg/m3) and is regarded as the NOAEC.

Executive summary:

The study investigated peripheral blood effects in 928 benzene exposed workers in five factories. The study was well designed and controlled, with subjects exposed across a wide range of benzene concentrations. The exposure assessment methodology employed was reliable.

Based on change point regression analyses, absolute neutrophil counts and reduced MPV were the most sensitive blood parameters for benzene exposure with estimated change points of 7.77 ppm (95% CI 2.97-20.1 ppm) and 8.24 ppm (95%CI 2.41-28.7 ppm), respectively. Lymphocyte count had the highest estimated change point of 38.1 ppm (95% CI 19.7 – 74.3 ppm).

Higher benzene exposures were more strongly related to red cell parameters, including MCV, than white cell parameters. The one-sided 95% lower confidence limit for the change-point concentration for neutrophils was calculated to be 3.5 ppm (11.2 mg/m³), and is the NOAEC for haematological effects in this study.