Overview of the Epidemiology of Formaldehyde

1
Overview of the Epidemiology of Formaldehyde

• Gary M. Marsh, Ph.D., F.A.C.E.
• Department of Biostatistics
• Graduate School of Public Health
• University of Pittsburgh
• 2005 European Toxicology Forum
• Brussels, Belgium
• November 8, 2005

2
Outline

• Key epidemiological research
• 2004 IARC Reclassification
• UPitt reanalysis of leukemia and nasopharyngeal
  (NPC) cancer mortality data from U.S. NCI cohort
  study
• UPitt study of NCIs Plant 1
• Conclusions

3
Epidemiologic Research
4
Three Large Cohort Studiesof Industrial Workers

• U.S. National Cancer Institute (NCI)
• (Blair et al., 1986 Hauptmann et al., 2003
  2004)
• British Medical Research Council (MRC)
• (Gardner et al., 1993 Coggon et al., 2003)
• U.S. Natl. Institute for Occupational Safety
  Health (NIOSH)
• (Stayner et al, 1988 Pinkerton et al., 2003)

5
U.S. NCI Cohort Study

• 1994 follow-up of cohort of 25,619 workers in 10
  U.S. industrial plants that made or used HCHO
• 6/10 nasopharyngeal (NPC) cancers deaths at one
  plant (Plant 1)
• Study Conclusions
• Lympho-hematopoietic cancers (Hauptmann et
  al., 2003)
• Suggestion of a causal association with HCHO and
  leukemia, particularly myeloid
• Solid tumors (Hauptmann et al, 2004)
• A possible causal association between HCHO and
  NPC and possibly other upper respiratory sites

6
British Cohort Study

• 2000 follow-up of cohort of 14,014 workers from 6
  British chemical plants (Coggan et al., 2003)
• Study Conclusions A small effect on sino-nasal
• or NPC cannot be ruled out a possible increase
  in
• lung cancer risk no leukemia excess

7
U.S. NIOSH Cohort Study

• 1998 follow-up of cohort of 11,039 garment
  workers from 3 U.S. plants (Pinkerton, et al.,
  2003)
• Study Conclusions A possible relationship
• between HCHO and myeloid leukemia no nasal
• cancers or NPCs observed

8
UPitt - NCI Plant 1 Study

• 1998 follow-up of expanded cohort of 7,328
  plastics producing workers
• Nested case-control study of 22 pharyngeal
  cancers including 7 NPCs
• Includes 6 of 10 NPCs in NCIs 10 plant study
• Study Conclusions
• Corroborated Plant 1 NPC excess in NCI study
• Little evidence of HCHO exposure-response
• Possible role of risk factors external to study
  plant

9
IARC Evaluation

• June 2004 - IARC upgraded formaldehyde from Group
  2a (probably carcinogenic to humans) to Group 1
  (known human carcinogen)
• Used strength of the evidence approach
• Based on new epidemiology findings available
  since last review in 1994
• Nasopharyngeal cancer was judged sufficient based
  largely on findings of the NCI U.S. study (moved
  from limited)
• Leukemia was judged nearly sufficient largely
  based on the NCI and NIOSH (garment workers)
  studies

10
Current Issues NPC

• The upgrading to sufficient from limited was
  based on a single studys results for NPC
• (NCI Study)
• 6 of 10 NPCs in Plant 1 with SMR 7.39 (plt.01)
  (confounding?)
• UPitt reanalysis shows Plants 2-10 SMR .98
• UPitt independent study of Plant 1 revealed no
  consistent associations with formaldehyde and
  suggested external factors

11
Current Issues NPC (contd)

• UPitt independent Plant 1 study found median
  average intensity of exposure 1/10 that of NCI
• A significant trend with exposure drove this
  assessment however, very small numbers for a
  rare cancer
• 8 observed and 4 expected deaths among exposed
• UPitt reanalysis shows trend entirely driven by
  Plant 1
• Four cases drove the IARC upgrading

12
Reanalysis of NCI Cohort Data (Marsh
and Youk, 2004 2005)

• Cohort data file obtained from NCI
• Examined robustness of NCI findings for leukemia
  NPC
• Alternative categorizations of HCHO exposure
• Internal vs. external mortality comparisons
• Analysis of duration and time since first highest
  peak
• Alternative metrics of NCIs peak exposure data
• Latency period analysis
• Plant-specific analysis (e.g., Plant 1 v. Plants
  2-10)

13
UPitt Value
14
NCI Findings for NPC Driven Entirely by
Plant 1
All Workers
n6
n19
n4
n50
15
NPC among HCHO-Exposed Workers
From Tarone and McLaughlin, Am J Epid 2005
16
NPC - NCIs Highest Peak Association Driven by
Plant 1
UPitt Finding for Plant 1
Basis of NCI Finding for Plants 1-10
UPitt Finding for Plants 2-10
zero deaths
17
NPC and Pharyngeal Cancer in UPitt Study of Plant
1
18
Majority of NPC and pharyngeal cancer cases
worked less than one year
19
Pharyngeal Cancer SMRs by Worker TypeCounty
Rates, 1945-98

p lt .05


Baseline
20
Pharyngeal Cancer SMRs by Average Intensity of
HCHO Exposure (through 1995)Total Cohort, County
Rates, 1945-98
p lt .01


Baseline
21
Pharyngeal Cancer Odds Ratios (OR) by Average
Intensity of Exposure to HCHO (through
1995)(model adjusted for smoking and year of
hire)
Global p .51
Baseline
22
Evidence that NPC or pharyngeal cancer risk is
related to HCHO or other factors at Plant 1
with or w/o consideration of co-exposures to
particulates
23
Most likely explanations for the increased
pharyngeal cancer risk in Plant 1 . . .
The increased risks for NPC and other pharyngeal
cancers may reflect the influence of (1)
Occupational factors associated with employment
outside Plant 1 (e.g., hardwood metal
manufacturing in the area) (2)
Non-occupational risk factors such as those
associated with unknown responses in
case-control study (3) Unmeasured
occupational factors at Plant 1that existed in
the past history of plant
24
Current Issues - Leukemia

• Leukemia findings do not appear biologically
  plausible
• UPitt reanalysis of NCI leukemia data revealed
• Findings not robust w.r.t. alternative
  categorizations of formaldehyde exposure
  methods of data analysis
• Serious limitations of highest peak exposure
  metric
• Key findings for highest peak and average
  exposure driven by inordinately large deficits in
  deaths in baseline categories used for RRs

25
Different Exposures Assigned Same Highest Peak
26
Exposure-Response in RRs Due to Inordinately Low
Baseline Mortality Rates for Leukemia
SMRs based on local county comparisons
Observed Deaths
29
NCI Baseline Category
NCI Baseline Category
NCI Baseline Category
Inordinately Low Rates
20
4
16
27
Pattern of Leukemia RRs Not Consistent with
Causal Association
28
Leukemia Latency Periods Too Long ?
29
Conclusions

• That NCIs suggestion of a causal association
  with HCHO and leukemia is not robust with respect
  to alternative characterizations of exposure and
  methods of data analysis casts considerable
  additional uncertainty about the validity of this
  association
• The findings of the independent UPitt study
  suggests that the large NPC excess in Plant 1 is
  not associated with HCHO and may reflect the
  influence of nonoccupational risk factors or
  external occupational risk factors
• IARCs reclassification of HCHO as Group 1 did
  not adequately account for the anomalous findings
  for NPC in Plant 1 or the findings from the
  independent UPitt Plant 1 study

30

• Questions?

31
NPC Latency Periods Too Short?