www.monash.edu.au THE HEALTH WATCH STUDY Australian petroleum industry cohort A/Prof Deborah Glass and Prof Malcolm Sim Monash Centre Occupational Environmental Health, Department of Epidemiology and Preventive Medicine
Jan 21, 2015
www.monash.edu.au
THE HEALTH WATCH STUDYAustralian petroleum industry cohort
A/Prof Deborah Glass and Prof Malcolm SimMonash Centre Occupational Environmental Health, Department of Epidemiology and Preventive Medicine
www.monash.edu.au
2
Health Watch
• Set up 1980• Prospective cohort study of mortality and
cancer incidence• Australian petroleum industry workers
– Upstream sites– Refineries– Terminals– Airports
• Funded Australian Institute of Petroleum (AIP)– Large companies not small independents
www.monash.edu.au
3
Health Watch Cohort
• 95% of blue collar employees interviewed– except those at sites with <10 employees
• >5 years in industry• Actively followed & re-interviewed every 5 years
until 2000
• Surveys inc. job histories, smoking and drinking
www.monash.edu.au
4
Cohort is ageing
• Over 30 years
• 16,623 men and 1,375 women
• 2004: 1,473 men and 34 women died
– 289,275 person-years of observation in men
– 19,347 person-years in women
www.monash.edu.au
5
Update to mortality and cancer incidence
• Matched to national death data – end 2004
• Matched to Cancer Registry data– end 2002
www.monash.edu.au
6
Strong healthy worker effect
All major causes of death are low
SexOverall SMR
(95% C.I.)Cancer SMR
(95% C.I.)Cancer SIR
(95% C.I.)
Male 0.72 (0.68-0.76)
0.81 (0.75-0.88)
0.99 (0.94-1.04)
Female 0.65(0.45-0.91)
0.88 (0.54-1.34)
0.89 (0.68-1.15)
www.monash.edu.au
7
Women in Health Watch
Too few women to do many analyses
• 21/34 deaths were from cancer– SMR for cancer as expected
• 58 cancers– SIR for cancer as expected
www.monash.edu.au
8
Mortality among men in Health Watch
Cause SMR (95% C.I.)Cancer (Malignant) 0.81 (0.75-0.88)
Ischaemic heart disease 0.77 (0.69-0.85)
Stroke 0.60 (0.46-0.77)
Respiratory disease 0.73 (0.59-0.89)
All diseases of the digestive system 0.57 (0.42-0.77)
External Causes (accidents, violence, suicide) 0.64 (0.53-0.77)
All other causes 0.55 (0.47-0.64)
All causes 0.72 (0.68-0.76)
www.monash.edu.au
9
For men in Health Watch
There is no evidence of increasing cancer incidence or increasing cancer mortality with:
• increasing duration of employment; • increasing time since first employment;• time period of first employment.
www.monash.edu.au
10
Cancer among men in Health Watch
• Significantly excess:– Mesothelioma - 1.29 (1.13 - 1.48) – Melanoma - 1.76 (1.12 - 2.65)
• Leukaemia, prostate cancer and bladder cancer are no longer in excess
• Kidney cancer raised but not in significant excess in cohort or drivers
www.monash.edu.au
11
Health Watch lymphohaematopoetic (LH) cancers over time
0.2
0.7
1.2
1.7
2.2
2.7
3.2
3.7
1987 1990 1993 1996 1999 2002
Year of analyses
SIR
for
men
non Hodgkin lymphoma (NHL)Multiple myeloma (MM)Leukaemia
www.monash.edu.au
12
Nested case-control questions
• Is benzene exposure associated with increases in:– Leukaemia & sub-types?– Non-Hodgkin lymphoma (NHL)?– Multiple myeloma (MM)?
• Is there a latent period?
• Does exposure rate (peaks) matter?
• Are smoking and alcohol risk factors?
www.monash.edu.au
13
Nested case-control study
79 LH Cancer 395 Controls5:1 age matched
Health WatchCohort
(~16,000 men)
www.monash.edu.au
14
Quantitative exposure assessment
• Detailed job histories from cohort records– Interview
– Company records
• Site history
• Contemporary colleague– Structured case-blind interview
> tasks> products> technology
www.monash.edu.au
15
Exposure model
• Exposure measurements – Company & supplementary data
Base Estimates for tasks (ppm)
• Exposure modifiers– eg technology factors
• Individual exposure estimates– work history + algorithm
individual exposure estimates (ppm & ppm-years)
www.monash.edu.au
16
Base estimates
• 54 BEs, 49 used in study
• 36 based on local data– Based on measured personal exposure to benzene– Data from Australian petroleum industry – Data from Australian sites – More than 3870 data points– Identified task/job – Routine exposure– Used AM of data
www.monash.edu.au
17
Rail car loading
Observed Value
6420-2-4-6
Exp
ect
ed
No
rma
l3
2
1
0
-1
-2
www.monash.edu.au
18
Laboratory worker (lubes, R&D)
Observed Value
10-1-2-3-4-5-6
Exp
ect
ed
No
rma
l3
2
1
0
-1
-2
www.monash.edu.au
19
Exposure metrics
• Duration (years)• Intensity (average daily ppm)
– Highest or longest job
• Cumulative exposure (ppm-years)
www.monash.edu.au
2020
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Other Upstream
Office
Upstream Operations
Supervision
Other Refinery
Other Terminal
Road Tanker Driving
Refinery Operations
Aircraft Refuelling
Wharf & JettyOperations
Tank FarmOperations
Laboratory
Road Tanker Loading
Fitting
Drum Laundry & Preparation
Vehicle Maintenance
Rail Car Loading
Drum Filling
Average Intensity of Exposure(ppm)
Job group and exposure
www.monash.edu.au
21
Years of employment
mean exposure period of 20 years (range 4-42)
0%
10%
20%
30%
40%
<10 10-20 20-30 30-40 >40
Cases
Controls
www.monash.edu.au
22
Health Watch case-control study
4-8 > 16
0.1
1.0
10
100
1000
NHL/MM
1-2<1 8-16 > 16
0.1
1.0
10
100
1000
Cumulative Lifetime Benzene Exposure midpoint (ppm-years)
Odd
s ra
tio (
log
scal
e) 86 63
Leukaemia
7
98
www.monash.edu.au
23
Combining two lowest exposure groups
Exposure (ppm-years)
Cases
OR (95% CI)
< 2 9 1.0
2 - 4 8 2.9 (1.0 – 8.5)
4 - 8 3 1.2 (0.3 – 5.0)
8 - 16 6 3.1 (0.9 – 10.6)
> 16 7 51.9 (5.6 – 477)
www.monash.edu.au
2424
Leukaemia latency
Odd
s R
atio
(lo
g sc
ale)
Cumulative Exposure (ppm-years)
Leukaemia < 15 years
0.05
1
10
100
1000
Leukaemia > 15 years
0 5 10 15 0.05
1
10
100
1000
34.12
6.18
www.monash.edu.au
25
Od
ds
Ra
tio (
log
sca
le)
Cumulative Exposure (ppm-years)0 20 40 60
0.5
1
2
5
10
20
50
100
Leukaemia exposure groups
Horizontal bars indicate the range of exposure in each group
www.monash.edu.au
26
Leukaemia risk by sub-type
OTHER2 ALL1 hairy cell leukaemia 2 unspecified lymphocytic
Exposurequintile
Odds Ratio( 95%CI)
AML 1-3 1.004 0.00 (0.00)5 8.89 (0.95-82.84)
ANLL 1-3 1.00(AML+AUL) 4 1.11 (0.18-6.96)
5 8.29 (1.31-52.36)CML 1-3 1.00
4 0.00 (0.00)5 0.78 (0.07-9.06)
CLL 1-3 1.004 2.25 (0.34-14.76)5 7.15 (1.29-39.70)
Cases405425501425
www.monash.edu.au
27
What is a peak?
• Highest job?
• Highest day?
• Highest hour?
• Highest 15 minutes?
www.monash.edu.au
28
Evidence for effects of peak exposure
0 10 20 30 40
0.5
1
10
100
500
1000
Odd
s R
atio
(lo
g sc
ale)
With CB/BTX cases
Without CB/BTX cases
Cumulative Exposure (ppm-years)
98
39
www.monash.edu.au
2929
High exposures
• 12 subjects were exposed to concentrated
benzene or BTX
• 5 developed leukaemia, no NHL or MM
• 5/12 exposed >32 ppm-years
• 4 developed leukaemia
• No cases among office workers
www.monash.edu.au
30
Adding possible high exposures (PHEs)C
umul
ativ
e E
xpos
ure
with
PH
Es
adde
d lo
g sc
ale
(ppm
/yea
rs)
12
48
1632
6412
8
1 2 4 8 16 32 64 128
Cumulative Exposure log scale (ppm/years)
www.monash.edu.au
31
Odds ratios reduced
7.8 (2.3 - 25.9)7>16
2.7 (0.7 - 10.1)6>8-16
1.2 (0.3 - 5.2)3>4-8
3.1 (1.0 - 9.3)8>2-4
1.09≤2
Cumulative exposure and PHEs
OR (95% CI)
Number of cases
Benzene exposure
(ppm-years)
www.monash.edu.au
3232
Summary of case-control results
• NHL MM - not associated with benzene exposure • Leukaemia - strongly positive
– ANNL & CLL ~ positive
• Significant excess risk at >16 ppm-years– Cum exp and intensity too close to separate
• Latency period 10-15 years • Effect of “peaks” – some evidence• No association with smoking or alcohol
www.monash.edu.au
THE POOLED PETROLEUM INDUSTRY CASE-CONTROL STUDY
www.monash.edu.au
34
Pooled study (published online JNCI 30/10/12)
• 3 case-control studies (Canadian, UK, Australian) nested in petroleum industry cohorts
• Each updated with new cases and pooled• more power for leukaemia subtypes• use WHO classification of LH cancers
• Similar design, case and control identification, exposure assessment and analytical methods
www.monash.edu.au
35
Aims of the study
To investigate the relationship between exposure to benzene and risk of leukaemia
– Evaluate dose-response overall– Evaluate by WHO subtype– Include leukaemias, MDS and MPD
– Explore influence on dose-response relationships of study, site type, job, lag/latency, exposure metrics
www.monash.edu.au
36
Pooling three nested case control studies
U.K.2
inconsistent dose response, depending on subgroups and different
exposure metrics
based on 31 LH cancers based on 90 leukaemias based on 79 LH cancers
before pooling data update studies
Canada1
no consistent dose response, but
small study
Australia3
strong dose response, for ANNL & CLLRefs:
1. Schnatter et al. 1996 53:773-781.
2. Rushton et al. 1997 54: 152-166.
3. Glass et al. 2003 14: 569-577.
60 LH cancersIncl. 5 MDS
193 LH cancersIncl. 11 MDS
117 LH cancersIncl. 13 MDS
370 LH cancers
www.monash.edu.au
37
Pooled study steps
• Ethical approvals, identify new cases & controls, obtain work histories, carry out exposure assessment
• Ensure consistency of disease classification• Certainty of diagnosis
• Assess consistency of exposure assessments– Development of common job groups
– Development of peak and skin exposure metrics
– Certainty of exposure
– Comparability of background exposure
– Rationalization of differences across studies
• Statistical Analyses
www.monash.edu.au
38
Traditional Paradigm: Anatomy• LEUKAEMIAS (in peripheral blood)• LYMPHOMAS (in lymph system)
New Paradigm: Cell of Origin• MYELOID tumours
– Myeloproliferative Disease (MPD)– Myelodysplastic Syndrome (MDS)– Acute Myeloid Leukaemias (AML)
• LYMPHOID tumours– B-cells – T-cells
Checked lymphohaematopoietic (LH) cancer classification
(leukaemias and lymphomas)
www.monash.edu.au
39
Quantitative exposure assessment
• Individual job histories
• Site histories
• Exposure of new cases and controls
assessed by original study method
• Exposure intensity (ppm) for each job• Confidence score for each estimate
– L, M, H
www.monash.edu.au
40
Pooled study exposure assessment
• Team of study hygienists and external hygienist• Peak exposure metric
– Prob. >3ppm for 15-60 mins at least weekly
• Skin exposure metric– 0, L, M, H prob of at least weekly skin exposure
• Prepared common list of job categories– Allocated each job for each individual
www.monash.edu.au
41
Exposure assessment rationalisation
• Compared job estimates between studies• Estimates in each job category by era
– pre 1945, 1945-1960, 1960s & 1970s and 1980+ – AM, GM, n, max & min
• If mean were same (within 10%)- no change• If different
– Checked job/technology/products/conditions– If no apparent local explanation, adjust
• Some job cats. had no other study comparison– Different industry sector or period
www.monash.edu.au
42
Study exposure estimates
Little change from original estimates
Little difference between original and revised
www.monash.edu.au
43
Statistical analyses
• Risk as ORs by exposure : – cumulative benzene (ppm-years)
> ppm-years within relevant window (lag/latency)
– average & maximum intensity (ppm)– peaks & skin
• Job category, industry sector
• Sensitivity analyses: study, job confidence, exposure confidence
www.monash.edu.au
44
www.monash.edu.au
45
1.73
4.333.47
11.6
0.1
1
10
100
>0.348 and<2.93
>2.93 >0.348 and<2.93
>2.93
MDS, cumulative benzene exposure and certainty of diagnosis
Odd
s R
atio
s
All Subjects More Certain Diagnoses
Cumulative Exposure (ppm-years)
www.monash.edu.au
46
Pooled analysis MDS cases and controls
Suggests MDS cases over-represented at approx 1+ ppm
Currentexposure zone
www.monash.edu.au
47
CONCAWE study findings
MDS associated with low benzene exposure MDS may be a more sensitive outcome than AML
AML: several ORs were >1, few statistically sig. perhaps higher benzene exp. for sig. risks of AML some cases formerly classified as AML were MDS
CML: several ORs >1, but no clear exposure response pattern
CLL and MPD: no strong relationship
www.monash.edu.au
48
Putting the evidence together
• Epidemiology assesses risk for the group Can it be applied to other groups?
• Risk estimates from single studies
• Pooled data or meta-analyses needed for conclusions about causation
wobbly
www.monash.edu.au
49
Population risk vs individual risk
• Attribution at a population level– Benzene exposure increases the incidence of
leukaemia
• Attribution for an individual– Benzene exposure caused leukaemia in this person
www.monash.edu.au
50
Acknowledgements
• Australian Institute of Petroleum• Institute of Petroleum (UK)• American Petroleum Institute• Conservation for Clean Air and Water
Europe (CONCAWE)• Aromatic Producers Association• Energy Institute• Canadian Petroleum Products Institute