Formaldehyde and Leukemia by Peter Morfeld

Formaldehyde and Leukemia: Critical Evaluation of Epidemiological Studies
Peter Morfeld
Institute for occupational Epidemiology and Risk Assessment (IERA) Institute for Occupational Medicine of Cologne University
University of Cologne Germany
University of Cologne
IARC 2004
Madrid, 19 April 2012| peter,morfeld@evonik.com
Seite 2
IARC Workshop, June 2004 on Formaldehyde and Leukemia

IARC 2006, Vol 88, Chapter 5.2, p. 276
In summary, there is strong but not sufficient evidence for a causal association between leukemia and occupational exposure to formaldehyde". Increased risks for leukemia has consistently been observed in studies of professional workers and in two out of three of the most relevant studies of industrial workers. These findings fall slightly short of being fully persuasive because of some limitations in the findings from the cohorts of industrial and garment workers in the USA because they conflict with the non-positive findings from the British cohort of industrial workers.
Seite 3
NCI study, up to 2004: Formaldehyde Exposure and Leukemia

Hauptmann et al (2003) - NCIs formaldehyde industry cohort 10 plants, 25,619 workers, mortality follow-up through 1994, 65 cases (incl 30 deaths from myeloid leukemia); no trend with cumulative exposure and (weak with) duration of exposure, positive trends with average and peak exposure intensity in internal (>0 ppm) comparisons: leukemia peak: RR=2.0 at 2-3.9 ppm and RR=2.5 at 4 ppm, relative to >0-2 ppm average: RR=1.5 at 0.5-0.9 ppm and RR=1.7 at 1 ppm, relative to >0-0.5 ppm myeloid leukemia peak: RR=2.4 at 2-3.9 ppm and RR=3.5 at 4 ppm, relative to >0-2 ppm average: RR=1.2 at 0.5-0.9 ppm and RR=2.5 at 1 ppm, relative to >0-0.5 ppm Marsh and Youk (2004) - re-analysis external comparisons revealed that the elevated leukemia and myeloid leukemia RRs and associated trends reported by NCI for peak and average intensity occurred because null (or slight) to moderate mortality excesses were compared with statistically significant baseline category deficits in deaths.
Seite 4
Other Studies on Formaldehyde Exposure and Leukemia up to 2004

Meta-Analysis (incl. NCI study) Collins and Lineker (2004) 8 studies, industrial workers: 7 studies, embalmers: 3 studies, pathologists/anatomists: all 18 studies:
RR=0.9 (0.95-CI: 0.8-1.0) RR=1.6 (0.95-CI: 1.2-2.0) RR=1.4 (0.95-CI: 1.0-1.9) RR=1.1 (0.95-CI: 1.0-1.2)
Studies by exposure levels of formaldehyde or duration of exposure Coggon et al (2003), UK - MRC study 6 plants, 14,014 workers, 31 cases; no positive trend: SMR=1.0 at 2 ppm, SMR=0.7 at > 2 ppm Pinkerton et al (2004), USA - NIOSH study 3 plants, 11,039 workers, 24 cases; no clear trend: SMR=1.0 at < 3 yrs, SMR=0.7 at 3-9 yrs, SMR=1.5 at > 10 yrs
Seite 5
After IARC 2004 Meta Analyses
Seite 6
After IARC 2004: Formaldehyde and Leukemia

Three Meta-Analyses: Bosetti et al (2008), Bachand et al (2010), Schwilk et al (2010) [superseded Zhang et al (2009)]
New NCI epi study on funeral directors and embalmers: Hauptmann et al (2009)
included in Bachand et al (2010) and Schwilk et al (2010)
Updated NCI epi study on 10 industrial plants: Beane Freeman et al (2009) updating Hauptmann et al (2003) by a longer mortality follow-up through 2004
included in Schwilk et al (2010)
Seite 7
Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Bosetti et al (2008) The overall RR was (professionals / industry workers): for all lymphatic and hematopoietic cancers 1.31 / 0.85, and 1.39 / 0.90 for leukemia. For lymphohematopoietic neoplasms there were modestly elevated risks in professionals, but not in industry workers.
Seite 8

Bachand et al (2010)
For leukemias, the summary relative risk (RR) was 1.05 (95% CI: 0.93, 1.20) for cohort studies, and the summary odds ratio (OR) was 0.99 (95% CI: 0.71, 1.37) for case-control studies. Based on cohort and case-control studies, no significant differences were seen by leukemia subtype, job type, publication period, or region. Previous meta-analyses showed elevated summary estimates for leukemia; however, these analyses included results from proportionate mortality studies and did not explore other factors that could influence or confound results. By limiting analyses to stronger case-control and cohort study designs, considering the effects of smoking, these meta-analyses provided little support for a causal relationship between formaldehyde exposure and leukemia.
Madrid, 19 April 2012| peter,morfeld@evonik.com Seite 9

In conflict to the meta-analyses presented so far:
Meta-Analysis: Zhang et al (2009) [and Zhang et al (2010)] Focusing on occupations known to have high formaldehyde exposure, summary relative risks (RRs) were elevated in 15 studies of leukemia: RR=1.54; 0.95-CI: 1.18-2.00 6 studies of myeloid leukemia: RR=1.90; 0.95-CI, 1.31-2.76
superseded by Meta-Analysis: Schwilk et al (2010)
Seite 10

Schwilk et al (2010) this meta-analysis includes two studies not used in previous analyses (Zhang et al 2009, 2010) 1) case-control investigation nested in a cohort of US funeral directors and embalmers (Hauptmann et al 2009)
replaced Wallrath and Fraumeni (1983, 1984) and Hayes (1990)
2) updated NCIs formaldehyde cohort study (Beane Freeman et al 2009)

replaced Hauptmann et al (2003)
Seite 11

Schwilk et al (2010) Meta-analysis focusing on high-exposure groups: increased risks of leukemia: RR = 1.53; 0.95-CI: 1.11 to 2.21; 14 studies increased risks of myeloid leukemia: RR = 2.47; 0.95-CI : 1.42 to 4.27; 4 studies Conclusion of authors These findings provide evidence of increased myeloid leukemia risk with exposure to formaldehyde.

Two studies performed by almost the same group of scientists Leukemia Zhang et al 2009: 15 studies; RR = 1.54; 0.95-CI: 1.18 to 2.00 Schwilk et al 2010: 14 studies; RR = 1.53; 0.95-CI: 1.11 to 2.21 (Harrington and Shannon 1975 split into 2 sub-studies, 3 studies replaced by 1 update) Myeloid Leukemia Zhang et al 2009: 6 studies; RR = 1.90; 0.95-CI: 1.31 to 2.76 Schwilk et al 2010: 4 studies; RR = 2.47; 0.95-CI: 1.42 to 4.27 (4 studies replaced by 2 updates) Observation: updating of studies caused a relevant change in RR estimate for myeloid leukemia only

Bachand et al 2010 (B): RR=1 for leukemia and RR=1 for myeloid leukemia vs. Zhang et al 2009 / Schwilk et al 2010 (S): RR=1.5 for leukemia and RR=2 for myeloid leukemia What are the reasons for these substantial differences? The updated studies (Hauptmann et al 2009, Beane Freeman et al 2009) had only an impact on the RR for myeloid leukemia. The inclusion of updated studies in S cannot explain the differences between B and S in the leukemia RR estimate The general methodology of computing overall relative risk estimates is similar and cannot explain the differences Does the set of eligible studies differ? Do the eligible data extracted from studies differ?

Differences in study selection B vs. S Dell and Teta 1995 was excluded in B but used in S for leukemia with RR= 2.65, 0.95-CI: 1.15 to 5.24 However, Dell and Teta 1995, wrote on p. 382: Among the R&D workers, the number of observed deaths (eight) from leukemia exceeded the expected deaths (three) There did not appear to be obvious common projects or exposures, except in general terms, that is, numerous references to solvents such as benzene and toluene.
Evaluation: Dell and Teta 1995 should be excluded (in favour of B)
Seite 15

Differences in study selection B vs. S Government Reports (Matonoski 1991, Robinson et al. 1987) used in B for leukemia with RR=1.35, 0.95-CI: 0.92 to 1.92 and RR=0.59, 0.95-CI: 0.01 to 3.28 but both excluded in S
Evaluation: both views appear to be acceptable
Seite 16

Differences in study selection B vs. S Marsh et al (2004) as alternative to NCI analyses (Beane Freeman et al 2009) of the NCI formaldehyde industry cohort additionally included in B for leukemia with RR= 0.79, 0.95-CI: 0.62 to 1.01 but not in S
Evaluation: exclusion appears to be reasonable (in favour of S)
Seite 17

Differences in study selection B vs. S Partanen et al (1993) used in B for leukemia with RR=1.40, 0.95-CI: 0.25 to 7.91 but excluded in S because: Partanen et al 1993 was excluded because exposure was assessed much differently in cases (personal interviews and company records) than in controls (only company records). Partanen et al (1993): This asymmetry in exposure data between cases and referents introduces a potential source of bias Evaluation: exclusion appears to be reasonable (in favour of S)
Seite 18

Differences in study selection B vs. S Sensitivity analysis by Schwilk et al (2010) 14 leukemia studies (FE): RR = 1.53, 0.95-CI: 1.18 1.98 18 studies, incl. the 4 studies additionally listed in Bachand et al (2010): RR = 1.45, 0.95-CI: 1.18 1.78 Evaluation: Difference in overall RR not caused by differences in study selection!
What caused the difference? Zhang et al (2009) and Schwilk et al (2010):
Meta-analyses focusing on high-exposure groups


Differences in selecting data from studies B vs. S Example: Beane Freeman et al (2009), NCI formaldehyde industry cohort S: myeloid leukemia at peak > 4 ppm: RR = 1.78, 0.95-CI: 0.87 to 3.64 B: all leukemias among all exposed: RR = 1.02, 0.95-CI: 0.85 to 1.22 B (Beane Freeman, Pinkerton, Stroup): myel leuk among all exp: RR = 1.09, 0.95-CI: 0.84 to 1.40 Beane Freeman et al (2009): myeloid leukemia, all leukemias observed unexposed exposed
SMR (0.95-CI) 0.65 (0.25 to 1.74), 0.48 (0.23 to 1.01) 0.90 (0.67 to 1.21), 1.02 (0.85 to 1.22)
Seite 20
4,
44, 116

Main methodological difference between Zhang et al (2009) and Schwilk et al (2010) vs. Bachand et al (2010) and Bosetti et al (2008)
Zhang et al (2009) and Schwilk et al (2010) selected the highest exposed subgroups from the eligible studies
Comment (Peter Morfeld): This is unusual in meta-analysis because highest cut points vary across studies metrics vary across studies (duration, peak exp, cum exposure, )

Results of Schwilk et al (2010) depend considerably on highest cut points chosen in each study even if the metric is the same Marsh and Youk 2004 (example): myeloid leukemia in NCI formaldehyde industry cohort NCI highest cat UPitt highest cat average intensity >1 ppm > 0.74 ppm SMR (US rates) 1.45 1.02
NCI cats: 60th and 80th percentiles of all exposed cancer deaths Upitt cats: tertiles of exposed leukemia deaths Methodological comment and advice: Royston P, Altman DG, Sauerbrei W. Dichotomizing continuous predictors in multiple regression: a bad idea. Statistics in Medicine 2006; 25:127-141

Main finding of Schwilk et al (2010)
All 14 leukemia studies (FE):
RR = 1.53, 0.95-CI: 1.18 1.98

Four myeloid leukemia studies (FE):
RR = 2.47, 0.95-CI: 1.57 3.86 Sensitivity analysis

Schwilk et al (2010), Table 1, FE n = 18 studies, incl. the 4 studies additionally listed in Bachand et al (2010) RR = 1.45, 0.95-CI: 1.18 1.78 (no substantial diff. to RR=1.53 from 14 studies) Re-calculation of this meta-analysis, FE (Peter Morfeld) n = 18 studies, incl. the 4 studies with data as listed in Bachand et al (2010) RR = 1.13, 0.95-CI: 0.96 1.32 (remarkable diff. to RR=1.53 from 14 studies)
Results of Schwilk et al (2010) depend considerably on the specific data extraction process

Zhang et al (2009) and Schwilk et al (2010) Comment (Peter Morfeld): this approach depends considerably on definition of metric depends considerably on definition of highest cat does not use the full information available Methodological comment and advice: Greenland and Longnecker: Methods for trend estimation from summarized dose-response data, with applications to meta-analysis. Am J Epidemiol 1992;135:1301-1309 Orsini et al: Meta-Analysis for Linear and Nonlinear Dose-Response Relations: Examples, an Evaluation of Approximations, and Software. Am J Epidemiol 2012;175:6673
combined RRs of highest categories across studies

Peter Morfeld: Data combined are heterogeneous (no statistical test necessary to be convinced) However, Schwilk et al (2010) performed tests of heterogeneity (Table 2) they found large p-values, usually > 0.2 This appears to rule out heterogeneity Schwilk et al (2010) wrote in the discussion section: Nevertheless, the relatively low heterogeneity statistics (and the heterogeneity P > 0.05) highlight the overall consistency in these data.
Seite 25

Peter Morfeld: Data combined are heterogeneous (no statistical test necessary to be convinced) This obvious heterogeneity cannot be ruled out by global p-values that are large; the tests as presented in Schwilk et al (2010) are insensitive Example given in Erren and Morfeld 2012: considerable influence on the calculated overall OR= 1.26 even though the global test for heterogeneity was not significant (p=0.3). Indeed, when measuring heterogeneity due to one study against all other 16 studies we calculated an I2=89.5% (p=0.002). When omitting this study we found a remarkably reduced overall OR. Warning in a leading textbook (Rothman et al. 2008, p. 671): large P-values do not indicate that it [heterogeneity] can be safely ignored

Calculation of tests and statistics (Peter Morfeld)

Test(s) of heterogeneity for all 14 studies analyzed in Schwilk et al (2010): chi2 Leuk Myeloid Overall 9.70 4.50 20.66 df 9 3 13 P 0.375 0.212 0.080 I2 7.2% 33.4% 37.1%
Random effects meta-analysis with standard confidence intervals is no solution Al Khalaf MM, Thalib L, Doi SAR. Combining heterogeneous studies using the random-effects model is a mistake and leads to inconclusive meta-analyses. J Clin Epidemiol 2011;64:119-123.

A possible way out
Riley RD, Higgins JPT, Deeks JJ. Interpretation of random effects meta-analyses. BMJ 2011;342:d549 predictive intervals should be calculated in random effects meta-analyses if findings are positive and heterogeneity cannot be excluded for sure Graham PL, Moran JL. Robust meta-analytic conclusions mandate the provision of prediction intervals in meta-analysis summaries. Journal of Clinical Epidemiology 2012; 65: 503-510
Seventy-two meta-analyses from 70 articles were identified, containing between three and 80 studies each, with median nine studies. Substantial heterogeneity results in exceedingly wide PIs. Results imply that meta-analytic practitioners should be even more cautious about the conclusions of a meta-analysis than they have been previously.
Seite 28

University of Cologne Peter Morfeld: Data from 14 studies analyzed by Schwilk et al (2010) L= leukemia M= myeloid leukemia Predictive intervals taking heterogeneity of selected data seriously: RR (0.95-CI) Overall (n=14): 1.58 (0.63 - 3.98) Myeloid (n=4): 2.64 (0.39 18.1)

The apparent conflict between published meta-analyses Bossetti et al (2008) and Bachand et al (2010) vs. Schwilk et al. 2010 [Zhang et al (2009) and Zhang et al (2010)] can be resolved. The latter studies selected the highest exposed subgroups from the eligible studies (not recommended: Altman, Greenland, ) metrics vary across studies (duration, peak exp, cum exposure) highest cut points vary across studies even if the metric is the same results are difficult to interpret (even if p is large: Al Khalaf, Rothman, ) no significant excess risk if heterogeneity is accounted for by predictive intervals (robust procedures necessary: Riley, Graham, )
IARC 2009 New Epi Studies
Seite 31
IARCs recent decision on Formaldehyde and Leukemias

IARC 2009 100F workshop - Baan et al (2009): The epidemiological evidence has become stronger: a recent study* found that embalming was significantly associated with an increased risk for myeloid leukemia, with significant trends for cumulative years of embalming (ptrend=0.020) and for increasing peak formaldehyde exposure (ptrend=0.036)
The Working Group concluded that, overall, there is sufficient evidence for leukemia, particularly myeloid leukemia
* new NCI study on embalmers: Hauptmann et al (2009)

New NCI study on embalmers: Design

Main hypothesis acc to Hauptmann et al (2003) and Bean Freeman et al (2009) Myeloid leukemia risk is linked to peak exposure in internal analysis Hauptmann et al (2009) Case-control study on professionals employed in the funeral industry 168 lymphohematopoetic deaths (incl 34 myeloid leukemias) [=cases of major interest] + 48 brain tumors 265 controls with death attributed to other causes (excluding respiratory and nervous system), matched to cases by data source, sex, dates of birth and death
Seite 33
New NCI study on embalmers: Exposure assessment

Hauptmann et al (2009) Interviews 1990-1992 with next of kin and co-workers on work practices, characteristics and tobacco use of study subjects Exposure assessment by a model that links rate of embalming, effect of ventilation, and concentration of formaldehyde solution to concentration in the air Qualitative interview data (guesses) transformed into quantitative data to apply the model [details in about 50% of subjects missing: no data to calculate peak, average or cumulative exposures for these subjects]
Seite 34
New NCI study on embalmers: Results

Hauptmann et al (2009) Unconditional logistic regression adjusting for matching variables and smoking Results focus on myeloid leukemia and embalming Ever vs. never: OR = 11.2 (0.95-CI: 1.3 to 95.6) Positive trends observed with duration and number of embalmings, not with cumulative exposure but with average and peak formaldehyde exposure effect of duration of embalming and peak exposure were reported by IARC as important findings, Baan et al (2009)
New NCI study on embalmers: Limitations of myeloid leukemia findings (1)

Hauptmann et al (2009) No incidence density sampling of controls, deceased controls only [Peter Morfeld: weak design, text book: Rothman et al 2008] Table 2: Mean average and peak exposures are reported to be identical among myeloid leukemias and controls, just duration and number of embalmings and cumulative exposures are higher among cases [Peter Morfeld: this does not support the modelling results reported] Race distribution clearly different between myeloid leukemias and controls (see Table 1) but ignored in analyses and discussion [Peter Morfeld: blacks should be dropped from analyses, note that Beane Freeman et al 2009 adjusted for race] Calendar year of first employed clearly different between myeloid leukemias and controls but not adjusted for in analyses [Peter Morfeld: cases began work more often before WWII]
Seite 36
New NCI study on embalmers: Limitations of myeloid leukemia findings (2)

Hauptmann et al (2009) Risks were attenuated when missing data were taken into account Trends no longer significant or positive in internal models the model type emphasized by Hauptmann et al (2003) and Beane Freeman et al (2009) Trend in peak exposure depend on only one case among controls the model type emphasized by Hauptmann et al (2003) and Beane Freeman et al (2009). The authors performed an analysis using a larger referent group: again no trend in internal analysis
Peter Morfeld: Hauptmann et al (2009) clearly limited
Seite 37
Embalmers = very informative? Industrial Workers Have Much Higher Average Formaldehyde Exposures
Job TWA Peak
Foundry
2.8
10
Source: IARC (1995)

Plywood Manufacture 2.2 10
Resins Operation
2.0
10
Garment Manufacture
1.9
Monomer Production
0.7
Pathologists
0.4
Embalmers
0.2
Seite 38
Additional New Data: Update of NCI cohort leukemia study

Updating Hauptmann et al (2003) by a longer mortality follow-up through 2004. Bean Freeman et al (2009): new leukemia risk estimate is lower than before!
4 3,5 3
Relative Risk
Hauptmann et al (2003) Beane Freeman et al (2009)
2,5 2 1,5 1 0,5 0 0 >0-<2.0

Category
2.0-<4.0
4.0+
Seite 39
Additional New Data: Update on NCI cohort leukemia study

An important problem detected in Hauptmann et al (2003) = NCI cohort study on leukemia with mortality follow-up until 31 Dec 1994:
Beane Freeman et al (2009) Hauptmann et al (2003) missed 1006 deaths due to an invalid tracing of vital status. Surprisingly, one calculates 995 missing deaths by comparing Table 1 (Hauptmann et al. 2004) and Table S1 (Beane Freeman et al. 2009b).
This difference with the published number of 1006 missing deaths creates doubt whether all errors were fixed by the NCI working group.
Seite 40
Additional New Data: Revised # of deaths in NCI cohort

Observed Deaths (Hauptmann et al., 2003) All Causes Unexposed Exposed Total All Cancer Deaths Unexposed Exposed Total All Solid Neoplasms Unexposed Exposed Total All LHP Unexposed Exposed Total Leukemia Unexposed Exposed Total
Revised Deaths (Beane-Freeman et al., 2009b)
Change in Observed Deaths
Percent Cange
827 7659 8486 183 1916 2099 166 1755 1921 17 161 178 4 65 69
1002 8479 9481 228 2147 2375 208 1967 2175 20 180 200 4 72 76
175 820 995 45 231 276 42 212 254 3 19 22 0 7 7
21 % 11 % 12 % 25 % 12 % 13 % 25 % 12 % 13 % 18 % 12 % 12 % 0% 11 % 10 %
Source: Marsh et al (2010)
Seite 41
NCI cohort study: differential missing of deaths

1. Differential missing among the unexposed The percent increase in revised deaths among unexposed" persons is twice that of the exposed" for all deaths, all cancer deaths, and all solid neoplasm deaths
2. Differential missing among the exposed? The increase of 7 leukemia deaths is only seen among the exposed. However
Seite 42
Leukemia deaths were missed in the lowest exposed category (>0 to 1.9 ppm) of the highest ever peak exposure
Table 2. Correction of NCI 1994 Data and Additional Follow-up Attentuated Exposure-Response Relationship for Highest Ever Peak Formaldehyde Exposure and Leukemia. NCI* Original 1994 Follow-up a,b Highest Ever Peak Category Unexposed > 0 to 1.9 ppm* (referent) 2.0 to 3.9 ppm 4.0 ppm Trend Test (p-value) All groups Exposed only
0.001 0.004 0.0208 0.0942 0.02 0.12 Deaths, RR* (95% CI*) 4, 0.78 (0.25, 2.43) 16, 1.00 20, 2.04 (1.04, 4.01) 29, 2.46 (1.31, 4.62)
NCI Revised analyses for 1994 Follow-up c

Deaths, RR (95% CI) 4, 0.52 (0.17, 1.57) 23, 1.00 20, 1.36 (0.73, 2.51) 29, 1.60 (0.90, 2.82)
NCI 2004 Follow-up d
U Pitt Analysis of Original 1994 Follow-up e

SMR* (95% CI) 0.38 (0.10, 0.97) 0.50 (0.28, 0.81) 1.04 (0.63, 1.60) 1.31 (0.88, 1.89)
Deaths, RR (95% CI) 7, 0.59 (0.25, 1.36) 41, 1.00 27, 0.98 (0.60, 1.62) 48, 1.42 (0.92, 2.18)
* NCI, National Cancer Institute, RR, relative risk; CI, confidence interval; SMR, standardized mortality ratio; ppm, partsper-million.
a b
Hauptmann et al., 2003 Used by IARC in 2004 reclassification c Beane Freeman et al., 2009b d Beane Freeman et al., 2009a e Marsh and Youk, 2004

Seite 43
NCI cohort study: open questions

1. The percent increase in revised deaths among unexposed" persons is twice that of the exposed" for all deaths, all cancer deaths, and all solid neoplasm deaths (however, no new risk estimates provided by NCI yet) 2. The increase of 7 leukemia deaths is only seen among the low exposed: corrected risk estimates are lower than the estimates published before, and still reflect a deficit of deaths in the baseline category (Marsh and Youk 2004) 3. The difference between 995 missing deaths according to the tables and the published number of 1006 missing deaths creates doubt whether all errors were fixed by the NCI working group.
Described and discussed in:
Marsh, Youk, Morfeld, Collins, Symons 2010: Incomplete follow-up in the National Cancer Institute's formaldehyde worker study and the impact on subsequent reanalyses and causal evaluations. Regulatory Toxicology and Pharmacology 2010; 58: 233236
Seite 44
NCI cohort study: Hauptmann et al 2003 in perspective

Figure. Exposure-Response Relationship for Highest Ever Peak Formaldehyde Exposure and Leukemia Based on Original, Corrected and Updated NCI Data and Reanalysis of Original NCI Data 3 NCI-1994 Original 2.5 NCI-1994 Revised NCI-2004 Relative Risk or SMR 2 SMRs NCI-1994*
* Marsh and Youk (2004). Based on incomplete data (Beane Freeman et al., 2009a,b)

Used by IARC in 2004 Reclassification
1.5
0.5
0 Unexposed >0 - <1.9 2.0 - 3.9 4.0+ Highest Ever Peak Formaldehyde Exposure (ppm)
NCI cohort study: Is the reduction of relative risk a reflection of leukemia latency periods?
Leukemia Patterns by Duration and Latency not Consistent with Causal Association:
Source: Marsh and Youk (2004)
Seite 46
Summary and Conclusions
Seite 47
Summary: Epidemiology of Formaldehyde and Leukemia

Most important medical endpoints discussed nasopharyngeal cancers (NPCs)
leukemia(s)
Of major relevance: US National Cancer Institute (NCI) studies. Embalmers: Hauptmann et al. 2009, Industry: Beane Freeman et al 2009 on leukemias,. IARC decided (2004, 2009) that formaldehyde exposure is a human carcinogen causing NPCs and leukemia Critical discussions and re-evaluations of the NCI cohort study (Marsh and coworkers 2004, 2007), errors in former studies of NCI (Hauptmann et al 2003, 2004), no support by a large British study (Coggon et al 2003) and clear limitations of the embalmer case-control study (Hauptmann et al 2009) shed doubt on IARCs decision Recent meta-analyses (Bossetti et al 2007, Bachand et al 2010) lend support to this critical point of view. Schwilk et al (2010) is not convincing due to the specific data extraction method applied.
Seite 48
Summary: Formaldehyde and Leukemia by ANSES

Proposal for Harmonised Categorization and Labelling based on Regulation (EC) No 1272/2008 (CLP Regulation), Annex VI, Part 2
ANSES (on behalf of the French MSCA): CLH REPORT FOR FORMALDEHYDE, Sep 2011, p.170: Conclusion about Formaldehyde and leukemia Overall, some positive observations have emerged in industrial populations but meta-analyses generally show a discrepancy in the results between industrial and professional populations in which several studies indicate an increased risk of leukaemia and especially myeloid leukaemia. Therefore, it is considered that available data do not provide causal evidence for formaldehyde as the aetiological factor as a bias specific to professionals cannot be ruled out.
Seite 49
Broader Context: CoI = Conflict of Interest

Boffetta P, McLaughlin JK, Veccia CL, Tarone RE, Lipworth L, Blot W.: False-positive results in cancer epidemiology: A Plea For Epistemiological Modesty, JNCI (2008) Letters to the Editor: Cogliano and Straif (2009) Hauptmann and Ronckers (2009) and reply of authors and reply of authors
Issues: false positive findings: overstated degree of evidence conflict of interest of experts at IARC workshop: careerism, advancing own research results in discussions, increasing prestige and possibility of future funding One discussed example is formaldehyde epidemiology
Please read the paper and the letter exchanges and make up your mind!
Seite 51

Formaldehyde and Leukemia by Peter Morfeld

Enviado por

Dados do documento

Direitos autorais

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formaldehyde and Leukemia by Peter Morfeld

Enviado por

Direitos autorais:

Formaldehyde and Leukemia: Critical Evaluation of Epidemiological Studies

University of Cologne Germany

Madrid, 19 April 2012| peter,morfeld@evonik.com

IARC Workshop, June 2004 on Formaldehyde and Leukemia

IARC 2006, Vol 88, Chapter 5.2, p. 276

Madrid, 19 April 2012| peter,morfeld@evonik.com

NCI study, up to 2004: Formaldehyde Exposure and Leukemia

Madrid, 19 April 2012| peter,morfeld@evonik.com

Other Studies on Formaldehyde Exposure and Leukemia up to 2004

Madrid, 19 April 2012| peter,morfeld@evonik.com

After IARC 2004 Meta Analyses

Madrid, 19 April 2012| peter,morfeld@evonik.com

After IARC 2004: Formaldehyde and Leukemia

Madrid, 19 April 2012| peter,morfeld@evonik.com

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Madrid, 19 April 2012| peter,morfeld@evonik.com

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

superseded by Meta-Analysis: Schwilk et al (2010)

Madrid, 19 April 2012| peter,morfeld@evonik.com

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

2) updated NCIs formaldehyde cohort study (Beane Freeman et al 2009)

Madrid, 19 April 2012| peter,morfeld@evonik.com

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Evaluation: Dell and Teta 1995 should be excluded (in favour of B)

Madrid, 19 April 2012| peter,morfeld@evonik.com

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Evaluation: both views appear to be acceptable

Madrid, 19 April 2012| peter,morfeld@evonik.com

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Evaluation: exclusion appears to be reasonable (in favour of S)

Madrid, 19 April 2012| peter,morfeld@evonik.com

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Madrid, 19 April 2012| peter,morfeld@evonik.com

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

What caused the difference? Zhang et al (2009) and Schwilk et al (2010):

Meta-analyses focusing on high-exposure groups

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

RR = 1.53, 0.95-CI: 1.18 1.98

RR = 2.47, 0.95-CI: 1.57 3.86 Sensitivity analysis

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

combined RRs of highest categories across studies

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Madrid, 19 April 2012| peter,morfeld@evonik.com

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Calculation of tests and statistics (Peter Morfeld)

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Madrid, 19 April 2012| peter,morfeld@evonik.com

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

Meta-Analyses after IARC 2004: Formaldehyde and Leukemia

IARC 2009 New Epi Studies

Madrid, 19 April 2012| peter,morfeld@evonik.com

IARCs recent decision on Formaldehyde and Leukemias

* new NCI study on embalmers: Hauptmann et al (2009)

New NCI study on embalmers: Design

Madrid, 19 April 2012| peter,morfeld@evonik.com

New NCI study on embalmers: Exposure assessment

Madrid, 19 April 2012| peter,morfeld@evonik.com

New NCI study on embalmers: Results

New NCI study on embalmers: Limitations of myeloid leukemia findings (1)