Standard setting in the Netherlands:approaches and examples Prof Dick Heederik, PhD Institute for Risk Assessment Sciences, Division of Environmental Epidemiology Utrecht University, The Netherlands firstname.lastname@example.org
Standard setting in the Netherlands1. Health Council, independent, protected by law: proposal for a standard2. Socio Economic Council (tri-partite): feasibility of the standard3. Ministry of Social Affairs and Employment: sets the standard4. Procedures described in guidelines of health councilStandard may be lower than an European Limit (ScientificCommittee Occupational Exposure Limits (SCOEL)
Standard setting in the Netherlands1. Committee consists of toxicologists, epidemiologists, hygienists 1. General trend is that epidemiological information (human studies) is becoming more and more important2. Regular collaboration with other organizations (Nordic Expert Group) 1. Literature evaluation by a combined subgroup 2. Hazard evaluation per country
Standard setting in the Netherlands: carcinogens1. Non-genotoxic carcinogens (threshold)2. Genotoxic carcinogens with a non-stochastic mechanism (threshold)HEALTH BASED RECOMMENDED OCCUPATIONAL EXPOSURE LIMIT3. Genotoxic carcinogens with a stochastic mechanism (no threshold)4. Genetoxic carcinogens with an unknown mechanism (assume no threshold)Often linear E-R, more often data drivenRISK CALCULATIONS exposure at on average 10-4 and 10-6 extra risk peryear (4.10-3 and 4.10-5 per 40 years exposure)
Differences with other organizations1. EU REACH regulatory framework: 1. DMEL comparable with a risk calculation on the basis of a linear model 2. Large assessment approach (based on margin of exposure principle as applied by European Food Safety Agency), not relevant for occupational standards2. SCOEL linear models, no guidance document
Steps in the risk assessment proces- Selection of information useful to derive an HBROEL or risk figures (complete review)- Selection of critical study/studies: Quality review/pooled- analysis/meta-analysis- Estimation of carcinogenic activity of an agent (exposure- response)- Calculation of risk in relation to exposure/derivation of a NOAEL or nowadays a BMD(L)- Obtaining exposure level at which a certain absolute risk is realized or Health Based Occupational Exposure Limit (HBROEL)
Exposure assessment and evidence basedmedicine✽Recent Dutch Health Council report 2009✽App. 200 papers identified✽50% major quality problems, exposure assessment component, underpowered, measurement endpoint, design✽These studies were not considered in the evaluation
Stepwise exclusion of studies with qualityissues (Lenters et al., EHP 2011, AOH 2012)
Benzene- carcinogenicity is a complex mechanism including genotoxic damage, inhibition of DNA repair and altered oncogenic signalling- leukaemia develops from genotoxic effects in the progenitor cells in the bone marrow, a primary target in benzene-toxicity.- Do effects on bone marrow cells have a threshold?- Is this an initial and required step to neoplastic disease?
Deriving occupational exposure limits:comparison of approaches1. NOAEL analysis2. BMDL analysis3. Calculation of risks based on exposure response relations
Decreased WBC, Granulocyte and Lymphocyte Counts and Benzene Exposure in Previous Month (Lan Q et al., Science 2004) Factory B (k=213, n=2667) B e n z e n e e xp o s u re F a c to ry B 100 9000 2000 2001 Controls (140) 8000 <1 ppm (109) 7000 1-10 ppm (110) Peripheral blood cell count >10 ppm (31)B e n z e n e (P P M ) 6000 10 5000 4000 3000 2000 1 1000 0 White Blood Cells Granulocytes Lymphocytes N= 214 172 209 20 209 194 210 209 207 171 248 138 450 16 0 .1 A l A l N t M r M r ar ar ay ay ov ec n n n ug ep ug eb eb Ju Ju c p p Ju Ja Ju O A A M M D F F S Monthly Personal Benzene Exposure Distributions by Factory (Vermeulen et al., 2004) 12
exposure limits: NOAEL analysis1. Lan et al., (2004) study high quality study with a balanced design. Data obtained from a representative working population, exposed up to 16 months.2. mean concentration of 0.57 ppm (1.8 mg/m3) a reduction of neutrophils was reported in the main study.3. subgroup of 30 workers, not directly exposed to other solvents, with the lowest benzene exposure level measured of 0.29 ± 0.15 ppm (mean ± SD) and a significant reduction in blood cells.
exposure limits: NOAEL analysis3. uncertainty factor of 3 to take into account the use of a minimal effect level instead of a no-effect level.4. A LOAEL of 0.3 ppm, results in an HBROEL of 0.3/3 = 0.1 ppm
exposure limits: BMDL analysis1. Lan et al. (2004) biomarker study was used to estimate exposure levels that are expected to result in a drop of 5% or more in white blood cell populations.2. Generalized Additive Model (GAM), and was adjusted for age, body mass index (BMI), sex, smoking, alcohol consumption, and presence of infections
Spline Regression Analyses of WBC Count and Benzene Exposure ✽ Modeling of data from 247 exposed and 139 control subjects ✽ No apparent threshold ✽ Evidence of supralinear responseStavanger 2007 16
exposure limits: BMDL analysis3. Using the BMDL10 based on the best fitting model (lin- log) for the most sensitive endpoint would result in an estimate of 0.1 ppm as well
Issues in the interpretation of the results- Other studies that made use of routine benzene exposure and haematological response data (Swaen et al., Chem Biol Interactions, 2010; Tsai et al., Reg Tox Pharmacol, 2004)- when is a negative an informative negative study? (Ahlbohm et al., 1990)- study quality issues?- genetic differences?
What if benzene is considered a no thresholdcarcinogen? B enzene D ose -R esponse R elationships✽Benzene example ✽Low dose risk 1 0 0 0 n= 1 0 2 3 identification P liofilm 1 0 0 ✽Low dose response 1 0 R isk for AML 1 modeling 0 .1 ✽Identification of < 4 0 4 0 -2 0 0 2 0 0 -4 0 0 > 4 0 0 ppm -yrs susceptible groups n=4 7 7 14 W ong, 1995 100 C hina 10 R isk for A N LL/M D S 1 0.1 none <40 40-99 100 + H ayes et al., 1997 ppm -yrs 19
Use of evidence from occupational studies in benzene dose-response assessment• Currently most dose-response assessments for benzene are based on one study conducted among rubber hydrochloride workers (PLIOFILM)• Linear extrapolation of findings to exposure levels relevant to the general public• Problems: 1. Is linear extrapolation justified? 2. Discussion on the quality of the exposure assessment in the PLIOFILM study
Quality of the exposure assessment in the PLIOFILM study✽ Few exposure measurements✽ Data gaps filled with ‘expert judgment’✽ Large potential for exposure misclassification: ✽ Assigned exposure levels were either to high or too low ✽ Assigned exposure levels not very accurate✽ Difficult to decide what the actual value of this study is for dose-response assessment
Studies included in the meta-regression (Vlaanderen et al., 2010)Study Country Study design Risk estimates Year Nested OR 2003Australian Health Watch Australia case- controlCAPM-NCI China Cohort, RR 1997 Nested OR 1996Canada petrol Canada case-controlCostantini Italy Cohort SMR 2003DOW USA Cohort SMR 2004Wong USA Cohort SMR 1987 Nested OR 1997UK-Petrol U.K. case-controlPliofilm USA Cohort SMR 2002Swaen Netherlands Cohort SMR 2005
Resulting Regression models Deviance linear model (1): 29.3 (28 df) Deviance natural spline (2): 25.8 (27 df) 3 Flexible model is not linear 1 2ln R R 2 1 Both models predicted an intercept 0 Linear intercept: RR = 1.65 0 Spline intercept: RR = 0 0 100 200 3 1.33 400 500 600 C u m u la tive e xp o su re (p p m -ye a rs)
Supralinear shape✽ observed low-dose supralinearity biologically relevant? • Saturation of benzene-metabolite enzymatic pathways might have induced a supralinear shape 2000 BO-Alb, pmol/g 1500 1000 500 0 0 10 20 30 40 50 Benzene, ppm✽ Attenuation of the ERC at higher exposure levels might also have played a role: • Depletion of susceptible individuals at high exposures • Exposure measurement error • Healthy worker survivor effect • High disease background rates
Sensitivity analysis: Effect of leaving one study out 3 .0 Without CAPM-NCI 2 .5 2 .0 ln R R 1 .5 1 .0 Without Pliofilm 0 .5 0 .0 0 100 200 300 400 500 600 C u m u la tive e xp o su re (p p m -ye a rs) Study quality is associated with a n.s. increase in slope of the ERR (Vlaanderen et al., 2012)
Impact of substituting benzene with a measure ofbenzene metabolism (Vlaanderen Am J Epid 2011) Exposure metric Health Watch AIC Pliofilm AIC Fold (*10-3) (*10-3) difference Benzene 98.9 103.7 5.35 207.0 18 p = 0.0007 p < 0.0001 Sum 2.09 104.4 0.13 207.5 16 of metabolites p = 0.0007 p <0.0001 PBPK model Sum 1.22 103.9 0.12 207.6 10 of metabolites p = 0.0006 p < 0.0001 MML model Sum 1.32 103.3 0.11 207.2 12 of metabolites p = 0.0005 p < 0.0001 Regression splines model
Risk calculations Health Council- Meta-analysis benzene as described in Vlaanderen et al. (2010)- Repeated for leukemia and AML- Cumulative risks of mortality from leukemia and AML were compared for simulated (hypothetical) cohorts of exposed and unexposed subjects in a life-table analysis
Use of exposure response to calculate exposurelimit: Life-table analysis- All-cause and cause-specific (leukemia/AML) mortality rates identical in the exposed and non- exposed cohorts- All-cause mortality rates for males: Statistics Netherlands (Statline)- mortality rates for leukemia and AML: Comprehensive Cancer Centre (IKC).- Mortality rates modeled using Generalized Additive Model to obtain smooth estimates of average mortality rates by age.
Benzene exposure levels (ppm) that result in anexcess mortality of 4/100.000 (1/1 000 000 per year)at age 75 after occupational exposure from 20-65 Meta-regression model Decay function Leukemia AML Linear, no intercept None 0.011 0.031 Linear 15yrs 0.020 0.057 Exponential 10yrs 0.013 0.038 Exponential 15yrs 0.013 0.038 Linear, intercept None 0.017 0.044 Linear 15yrs 0.032 0.080 Exponential 10yrs 0.021 0.053 Exponential 15yrs 0.021 0.054 Spline, no intercept None 0.003 0.007 Linear 15yrs 0.005 0.013 Exponential 10yrs 0.003 0.008 Exponential 15yrs 0.003 0.008 Spline, intercept None 0.004 0.009 Linear 15yrs 0.007 0.016 Exponential 10yrs 0.004 0.011 Exponential 15yrs 0.004 0.011
Issues in the calculation of risks risk diff.✽ Complete follow-up versus age 80: 1-6-1.9✽ Use of average rates versus male rates only 0.8-0.9✽ continuous risk after age 65 versus trapezium model 1.5-2.0✽ NHL included versus not included 2.6-3.6✽ Mortality rates UK, EU, NL: no effect
ConclusionAny approach (NOAEL), BMDL, risk calculations, on the basis of thepresent information, results in a clear downword pressure for anexposure standardStrong interplay between epidemiological and toxicologicalinformationEpidemiological information crucial for risk calculations (allapproaches make use of human data)Toxicological interpretation (mechanisms) drives the choice for one ofthe approaches