Use of this information is subject to copyright laws and may require the permission of the owner of the information, as described in the ECHA Legal Notice.
EC number: 200-001-8
CAS number: 50-00-0
Formaldehyde has local carcinogenic activity in experimental animals; there is evidence for a threshold effect for tumors involving cytotoxicity and regenerative cell proliferation as the mode of action.There is no evidence for systemic or local carcinogenic effects after oral exposure in rats. In dermal initiation/promotion studies formaldehyde did not initiate or promote skin tumorigenesis in mice. There is clear evidence from chronic inhalation studies in rats that formaldehyde causes tumors in the nasal cavity.The results from epidemiological studies are highly contested but may suggest an increased risk of cancer only at two tumor sites: nasopharyngeal cancers (NPCs) and leukaemias. The most recent review covering both relevant endpoints (Bachand et al 2010) concluded that summary risk estimates for NPCs were not elevated after excluding a single plant with an unexplained cluster of NPCs. Moreover, the most influential epidemiological study, i.e., the NCI industrial cohort study that includes this cluster plant (Hauptmann et al 2004) is unreliable regarding NPCs because of a missing robustness and an incomplete vital status follow-up (Marsh et al 2007a, Marsh et al 2007b, Marsh et al 2010). Bachand et al 2010 concluded about leukaemia risk: “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 leukaemia.” Furthermore, current analyses of the NCI cohort (Beane Freeman et al 2009a) did not confirm the statistically significant exposure-response relationships between formaldehyde exposure and leukaemia mortality reported before (Hauptmann et al 2003). The studies from 2009 up to 2015 did not strengthen an association between formaldehyde exposure and NPC or leukemia.
Effects of oral formaldehyde in the drinking water of rats on
incidences of lesions in the forestomach, fundus and pylorus of
glandular stomach and anterior duodenum with and without initiation by
oral MNNG and sodium chloride
Initiation & promotion with 0.5% FA
No initiation but exposure to 0.5% FA
No initiation and no exposure to FA
Adenomatous hyperplasia in fundus
Adenocarcinoma in pylorus
Hyperplasia in pylorus
Adenocarcinoma in duodenum
FA: formaldehyde; *: significantly different from initiation only,
p<0.05; **: p<0.01
Study meets scientific standards with acceptable restrictions concerning
initiation/promotion in forestomach and glandular stomach; not valid
concerning formaldehyde control (low number of animals, short treatment
period, one dose level [clearly irritant]; possible confusion between
papillary hyperplasia and papilloma in the forestomach; partly limited
documentation, e.g. no details about test substance).
In an initiation/promotion study male Wistar rats (n=10-30) were
initiated with 100 mg/L MNNG (N-methyl-N'-nitro-N-nitrosoguanidine) in
the drinking water (ad libitum) plus 10% sodium chloride in the diet (ad
libitum) for 8 weeks followed by promotion for 32 weeks with 0.5%
formaldehyde in the drinking water (n=17 for evaluation). Additional
control groups received only initiation (n=30) or only 0.5% formaldehyde
(n=10) or only vehicle (n=10). All groups were sacrificed after 40
weeks. Body weight gain was decreased in all treatment groups compared
to vehicle control suggesting toxic effects of initiation and promotion
procedure. In accord with other long-term drinking water studies local
lesions in forestomach (papilloma, erosion and ulcer at the limiting
ridge of fundic mucosa) and glandular stomach (hyperplasia) were induced
by formaldehyde but no malignant tumours. Promotion in initiated rats
increased hyperplastic lesions in the glandular stomach but did not
significantly increase the total number of adenocarcinomas in the
gastroduodenal epithelium. Conclusion: In rats oral exposure to 0.5%
formaldehyde in the drinking water induced local lesions in stomach but
there was no clear evidence for promoting activity.
Incidence of hyperplasia and other lesions in the forestomach and glandular stomach of rats after chronic exposure via the drinking water
Type of lesion
Males, dose in mg/kg bw
Females, dose in mg/kg bw
Number of rats examined
Focal papillary epithelial hyperplasia
Chronic atrophic gastritis
In a combined chronic toxicity/carcinogenicity study (all
parameters tested as recommended in OECD 453) 70 male and 70 female
Wistar rats per dose (subgroups of 10 rats/sex/dose killed 12 or 18
months after start of exposure) received via the drinking water 0, 1.2,
15, 82 mg/kg bw/day (males) or 0, 1.8, 21, 109 mg/kg bw/day (females)
for 105 weeks (concentration: 0, 20, 260, or 1900 mg/L or 0, 0.002,
0.026, 0.19%). At the high dose body weight gain was decreased in males
& females and food and water consumption decreased. Other parameters
(except pathology) were not altered. Pathological alterations in the
kidney like the renal papillary necrosis detected in high dose males and
females is discussed as an effect of the reduced water intake and is
indirectly treatment related. Treatment related lesions were detected in
the forestomach (focal papillary epithelial hyperplasia, ulceration and
hyperkeratosis) and the glandular stomach (chronic atrophic gastritis,
ulceration and hyperplasia) of males and females in the high dose group.
No gastric tumours were induced. This study did not provide any evidence
of carcinogenicity in rats after oral administration of formaldehyde.
Oral exposure via the drinking water induced non-neoplastic local
effects in the stomach of rats at a concentration of 0.19% corresponding
to 82 mg/kg bw/d in males and 109 mg/kg bw/d in females; the NOAEC is
0.026% corresponding to 15 mg/kg bw/d in males and 21 mg/kg bw/d in
Non-neoplastic effects including hyperplasia in the nasal
cavity of mice exposed for up to 24 months including interim sacrifice
of sub-groups and mice found dead/moribund
Dose level in ppm (nominal)
Total number of animals
Incidences of lesions
Squamous metaplasia of epithelium
Squamous epithelial hyperplasia of the nasolacrimal duct
Atrophy of olfactory epithelium
Data on rhinitis were not given in the Summary Table and not
readable in other Tables in CIIT (1981; insufficient documentation in
Kerns et al., 1983); no data about statistical significance but clear
effects at the mid and high dose level; it is not clearly stated that
all animals necropsied are included in the number of animals evaluated
for the specific organ.
Study is comparable to OECD guideline 453 with acceptable
restrictions (adrenal weights not measured, no histopathology of
sternum, larynx and pharynx).
Inhalation exposure of male and female B6C3F1 mice for 2 years at
dose levels of 0, 2, 6, 15 ppm (120 mice per sex per dose; 24 months
exposure, 6 h per day, 5 days per week; post exposure period 3 months
[only female]; interim sacrifices) resulted only in local effects in the
nasal cavity of males and females, predominantly in the respiratory
epithelium. Rhinitis, dysplasia, squamous metaplasia were induced, but
also atrophy of olfactory epithelium with increasing exposure periods at
the high dose level. Except reduced body weight no further effects were
detected. The NOAEC was 2 ppm and the LOAEC 6 ppm. In contrast to rats
only a few tumours were detected in the nasal cavity at 15 ppm [However,
in males the high mortality rate (not treatment related but due to
housing conditions) might influence the outcome of this study].
Generally the lesions in mice were less severe than in rats (see also
CIIT 1981 & Kerns et al. 1983 in Section 7.5.3 & 7.7).
Conclusion: Chronic formaldehyde inhalation induced local effects
in the nasal cavity of mice; the effects are less severe than in rats.
TheNOAEC is 2 ppm and the LOAEC 6 ppm. Ambiguous test results
were obtained concerning carcinogenic effects, the low incidence of
squamous cell carcinoma in the nasal cavity of males at the high dose
was not significant but presumably of toxicological relevance.
Neoplastic and non-neoplastic effects in the nasal cavity of
rats exposed for up to 24 months including interim sacrifice of
sub-groups and rats found dead/moribund
Goblet cell hyperplasia
Olfactory epithelial metaplasia goblet/ciliated cells
Olfactory epithelial squamous metaplasia
Squamous cell carcinoma
Squamous epithelium hyperplasia
Squamous epithelium papillary hyperplasia
Data on rhinitis were not given in the Summary Table and not
readable in other Tables in CIIT (1981; insufficient documentation of
these data in Kerns et al., 1983); no data about significance; it is not
clearly stated that all animals necropsied are included in the number of
animals evaluated for the specific organ.
Study is comparable to OECD guideline 453 with acceptable restrictions
(no histopathology of sternum, larynx and pharynx).
exposure of male and female F344 rats for 2 years at dose levels of 0,
2, 6, 15 ppm (120 rats per sex per dose; 24 months exposure, 6 h per
day, 5 days per week; post exposure period 6 months; interim sacrifices)
resulted exclusively in local effects of the nasal cavity and of the
proximal trachea in males and females. Only non-neoplastic effects were
reported in the high dose group in the proximal trachea but a high
incidence in squamous cell carcinomas of the nasal cavity
(51/120 males and 52/120 females).
effects in the mid dose group (squamous cell carcinomas in nasal cavity
of 1 male and 1 female) are not of statistical significance but possibly
of toxicological relevance. At 2 ppm purulent rhinitis, epithelial
dysplasia, and squamous metaplasia were present in the anterior part of
the nasal cavity. Dysplasia occurred earlier than metaplasia. Non-neoplastic
effects in the mid dose group like dysplasia and squamous cell
metaplasia in the nasal cavity occurred also in more posterior parts or
the nasal cavity.
Conclusion: Non-neoplastic local effects in the nasal cavity of rats
were reported even at the low dose level of 2 ppm in a long-term
inhalation study in rats; tumours in the nasal cavity were found at a
dose level of 15 ppm.
The study meets generally accepted scientific standards
with acceptable restrictions concerning local effects in experiment A:
Partly no data or only limited documentation (e.g. test substance,
control). Only males tested; restricted to local effects; no data on
sacrifice of moribund animals (possible autolysis of tissues in hamsters
found dead); Methods in experiment B-E insufficient (only dissecting
microscope used). Data
in experiment B-E are not suitable for evaluation of toxicity and
carcinogenicity of formaldehyde.
Syrian Golden hamster were exposed to 0 or 10 ppm, 5 h/day, 5 days/week
for lifetime (experiment A; 132 controls and 88 treated hamster).
Histopathology of the respiratory tract revealed local
effects of the nasal cavity (metaplasia
and hyperplasia) in
only 4 out of 88 treated animals which were not found in controls. The
survival rate is significantly decreased suggesting further evidence for
toxicity at 10 ppm. No tumours have been found. Although no data are
available on higher dose levels, there is some evidence that hamsters
seemed to be less sensitive than rats.
Conclusion: No local carcinogenic effects in hamsters at
10 ppm lifetime exposure but metaplasia and hyperplasia in the nasal
cavity in 4 out of 88 animals.
Study meets scientific standards and is with restrictions
comparable to OECD guideline 453 (Partly limited documentation; e.g. no
data about purity of the test substance; ophthalmoscopic examination &
urine analysis not done; blood clotting not examined in haematology;
limited number of organs examined [systemic toxicity can not be fully
evaluate]; low number of animals concerning carcinogenicity; only males).
In this inhalation study 32 male F344 rats per dose were exposed
to 0, 0.3, 2.2, or 15 ppm formaldehyde 6 h/day, 5 days per week for 28
months; 5 rats per dose were sacrificed after 12, 18, or 24 months of
exposure (no post exposure observation period). No effects were detected
concerning any measured parameter at the low and mid dose except
histopathological alteration at the mid dose level in the nasal
cavity.Rats exposed to 0.3 ppm for 28 months did not show any
histopathological changes in the nasal cavity. At 2.2 ppm a significant
increase was noted in the incidence of squamous cell metaplasia with and
without hyperplasia but no carcinogenic effects. Clear local
carcinogenic effects were noted at a dose level of 15 ppm. No lesions
were detected in any organ other than nasal cavity. There is some
evidence that the study of Kamata et al. is less suitable for deriving a
LOAEL (more variable exposure concentrations than in other studies).
Conclusion: Inhalation exposure for 28 months induced in rats
non-neoplastic effects in the respiratory epithelium of the nasal cavity
at a dose level of ≥ 2.2 ppm, the NOAEC was 0.3 ppm. Clear local
carcinogenic effects in nasal cavity of rats were reported after chronic
exposure to 15 ppm.
The study meets scientific standards with acceptable restrictions
(partly limited documentation, e.g. no details about the test substance
and generation of gas); restricted to effects in the nasal cavity.
In this long-term inhalation study 90 -147 male F344 rats per dose
were exposed (whole body) to 0, 0.7, 2, 6, 10, 15 ppm formaldehyde gas 6
h/day, 5 days per week for 24 months. Rats were sacrificed immediately
after the exposure period and histopathology of the nasal cavity
performed. Rats exposed to 0.7 or 2 ppm for 2 years did not show any
histopathological changes in the nasal cavity and no effects on cell
proliferation index. Non-neoplastic effects were detected at a dose
level of 6 ppm (mixed cell infiltrate, squamous metaplasia, hyperplasia
in the anterior part of the nasal cavity) as well as a minimal
carcinogenic response (squamous cell carcinoma in 1/90). A sharp
increase in tumour incidences in the nasal cavity (mainly squamous cell
carcinoma) was reported at 10 and 15 ppm. Dose dependent effects on cell
proliferation were detected also only at >=10 ppm.
Conclusion: Inhalation exposure for 2 years induced local
non-neoplastic effects in the nasal cavity of rats at a dose level of 6
ppm, the NOAEC was 2 ppm; clearly carcinogenic effects were found at a
dose level of 10 ppm.
Incidences of non-neoplastic and neoplastic effects in the
respiratory tract of male rats exposed lifetime to 15 ppm
Number of rats with specified effect
15 ppm formaldehyde (n=100)
Shame-exposed control (n=99)
Untreated control (n=99)
Larynx squamous metaplasia
Trachea squamous metaplasia
Effects in nasal mucosa
Epithelial or squamoushyperplasia
Polyps or papillomas
No data about significance (but clear increase in incidences)
Study meets scientific standards with acceptable restrictions (only one
dose tested; rhinitis also in controls suggesting infection; only males
; restricted to respiratory tract and a few other organs; partly limited
documentation); reliable concerning effects in the respiratory tract
posterior to the nasal cavity (not examined in other studies presented
in Section 7.7).
Onehundred male Sprague-Dawley rats per dose level were exposed lifetime
to 0 or 15 ppm formaldehyde (6 h/day, 5 days/week). A significant
decrease in body weight gain and significantly increased mortality rate
were reported. Histopathology revealed increased incidences in
epithelial hyperplasia as well as squamous metaplasia (low incidence) in
larynx and trachea. Malignant tumours were recorded in the nasal cavity
of 40% of exposed rats (1 fibrosarcoma, 1 mixed carcinoma and 38
squamous cell carcinomas in 100 rats) but none in controls. An incidence
of 10% polyps or papillomas (0 in controls) was also observed.
Incidences of other tumours were comparable in both groups.
Conclusion: After chronic inhalation exposure in male rats neoplastic
effects were restricted to the nasal cavity, no tumours were detected in
larynx, trachea or lung. Beside effects in the nasal cavity, hyperplasia
and squamous metaplasia (low incidence) was found in larynx and trachea.
No tumours (0/28) were observed in both the groups exposed to
formaldehyde (FA) as initiator plus acetone (promotor), or 10% FA
(initiator) plus 1% FA (promotor).Tumor incidences in groups
initiated with BaP and treated with FA as promoter were 1/25 (4%), 2/28
(7%), and 7/27 (26%) at FA concentrations of 1%, 0.5%, and 0.1%,
respectively. Initiation with BaP followed by promotion with acetone as
well as initiation with acetone and promotion with TPA resulted in
tumour incidences of 3/27 (11%) in both cases. Five of 28 mice (18%)
treated with FA (initiator) and TPA (promotor) had skin nodules. The
highest tumour incidence (28/29; 97%) was observed in the group
initiated with BaP and treated with TPA as promotor.
Most of the nodules were benign tumours (keratocanthomas or papillomas;
malignant tumours were histopathologically diagnosed in the BaP+TPA
group, only (squamous cell carcinomas). No statistically significant
differences were observed between the treated groups and appropriate
controls in groups exposed to formaldehyde. According to the authors,
these results suggest that formaldehyde did not initiate or promote skin
tumorigenesis in minimally irritating concentrations (in a preliminary
test, a concentration of 10% FA was determined as moderately irritating,
1% caused mild irritation, 0.5% was slightly irritating, 0.1% induced no
local effects; no details given).
The average time to the first nodule was 110 days for mice
treated with BaP (initiator) plus TPA (promoter), 355 in acetone/TPA,
341 in FA/TPA, 370 in BaP/1% FA, 361 in BaP/0.5% FA, 338 in BaP/0.1% FA,
361 in BaP/acetone; no tumours were detected in FA/acetone or FA/1% FA.
There were no statistically significant differences found in the mean
days to first observed nodule in any test group when compared with the
appropriate negative control group.
Study meets scientific standards with acceptable restrictions (30 female
animals per group).
Thirty female CD-1 mice per group received a single dermal application
of an initiation solution followed two weeks later by application of
appropriate promoter solutions; the promoter treatments were
administered three times a week for 26 weeks followed by a post exposure
observation period of 26 weeks. Nine different dose regimens were
performed including appropriate negative and positive controls (BaP for
initiation and TPA for promotion). Initiation with formaldehyde (FA) was
5 mg FA in 50 µL vehicle (10% FA solution, irritant) and promotion
with 1, 0.5, or 0.1 mg FA in 100 µL vehicle (concentration 1, 0.5, 0.1%,
respectively; slight local irritation at 0.5%). Formaldehyde did
not initiate or promote skin tumorigenesis. Formaldehyde did not alter
the latency time to first tumour. The negative and positive controls
were valid. Malignant tumours were induced only in the positive control
(initiatorBaP/ promoter TPA).
Conclusion: Formaldehyde did not initiate or promote skin tumorigenesis
and did not alter the latency time to first tumour in an
initiation-promotion study in mice after repeated dermal application at
a concentration of 1%.
The studies from 2009 up to 2015 did not strengthen an association between formaldehyde exposure and NPC or leukemia.
Although NPC is a rare tumor, specific dietary habit and smoking have been associated with the incidence rates of this tumor. In the updates of the 3 large cohort studies no strengthening indications have been obtained: in the NCI study only one additional NPC was identified, but this occurred in the lowest exposure category. No additional NPCs were observed in the UK cohort and in the NIOSH garment workers studies. In addition, two further epidemiological investigations did not identify an association with NPC: an Italian study comprised a smaller cohort of 2750 employees with 457 deaths and a study on 1.2 million empoyees from the Finnish Cancer Registry with 149 cases of NPC. Thus, the former evidence, mainly from the NCI study, was not reinforced by the recent investigations.
In the updates of the 3 large cohort studies no further support for an association between formaldehyde and lymphohematopoetic neoplasms, especially for myeloid leukemias, were obtained. In the NCI study there was still a significant increase in some exposure categories, but the associations generally were weaker as compared to the former follow-up. The authors concluded that there was limited evidence for an association between formaldehyde and leukemia. Similarly, in the NIOSH garment worker study an association with formaldehyde still prevailed, but the former limited evidence was not strengthened by the update. In the UK cohort no significant associations with formaldehyde were seen for the different hematopoetic malignancies including a nested case-control study for all leukemias and myeloid leukemia. Also, the Italian study gave no indication in this respect. A metaanalysis focussing on high exposure groups reported an increased risk for leukemia, especially for myeloid leukemia. But this analysis suffered from methodological shortcomings by not using all available information. Some further analyse that included specifically the recent studies of Beane-Freeman et al. (2009) and Hauptmann et al. (2009), that were pivotal for the Group 1 decision of IARC, concluded that the overall ecvidence for leukemia was at most weak and may be due to chance or confounding.
For other tumor types, no clear associations were seen for lung and pharyngeal cancer in epidemiological studies from 2011, 2012, and 2013.
RAC (2012) did not consider the epidemiological evidence as sufficient for a classification as a human carcinogen regarding NPC and leukemia. RAC 2012 proposed the classification as Carc. Cat. 2; R45, according to Directive 67/548/EEC1999/45/ EC, and according to EU Classification, Labelling and Packaging of Substances and Mixtures (CLP) Regulation (EC) No. 1272/2008, Annex VI, the classification isCarc. Cat 1B (H350).
Data on experimental animals
Sufficient data are available on long-term drinking water studies in rats. No data are available on other species. In a combined chronic toxicity/carcinogenicity study (Til et al., 1989; all parameters tested as recommended in OECD Guideline 453) 70 male and 70 female Wistar rats per dose (subgroups of 10 rats/sex/dose killed 12 or 18 months after start of exposure) received via the drinking water 0, 1.2, 15, 82 mg/kg bw/day (males) or 0, 1.8, 21, 109 mg/kg bw/day (females) for 105 weeks (concentration: 0, 0.002, 0.026, 0.19%). At the high dose body weight gain was decreased in males & females and food and water consumption decreased. Other parameters (except pathology) were not altered. Pathological alterations in the kidney were related to reduced water intake and are indirectly treatment related. No systemic tumour formation was found. Treatment related lesions were detected in the forestomach (focal papillary epithelial hyperplasia, ulceration and hyperkeratosis) and the glandular stomach (chronic atrophic gastritis, ulceration and hyperplasia) of males and females in the high dose group. No gastric tumours were induced. In conclusion, hyperplasia in forestomach and glandular stomach but no tumour formation were detected after chronic exposure in rats via the drinking water. Non-neoplastic local effects in the stomach were reported at a concentration of 0.19% corresponding to 82 mg/kg bw/d in males and 109 mg/kg bw/d in females; the NOAEC is 0.026% corresponding to 15 mg/kg bw/d in males and 21 mg/kg bw/d in females. This study did not provide any evidence of carcinogenicity in rats after oral administration of formaldehyde.
These results were supported by the drinking water study of Tobe et al. (1989). Local effects in the forestomach were detected in rats after oral exposure to >= 0.10% in the drinking water (50 mg/kg bw/day) for 2 years; the NOAEC was 0.02% in the drinking water (10 mg/kg bw/ day). At the high dose level of 300 mg/kg bw/day corresponding to a concentration of 0.50% hyperplasia in forestomach and glandular stomach was found but no tumour formation in rats after chronic exposure (low number of animals).
Ambiguous results were presented in the drinking water study by Soffritti et al. (1989, 2002). This study is not sufficient for evaluation of systemic and local carcinogenic effects due to several deficiencies which are also mentioned by IARC (2006):
- no statistics given for the hemolymphoreticular tumours
- with regard to the Soffritti et al. (2002) the IARC working group performed a statistical analysis in comparison to the methanol control group. For hemolymphoreticular tumours a statistically significant increase was only found for the high dose males
- pooling of lymphomas and leukaemias (designated as hemolymphoreticular tumours)
- no reporting of non-neoplastic lesions
- absence of information on historical incidences of hemolymphoreticular tumours. This was also addressed in a letter to the editor by Feron et al., (1990) criticizing the report of Soffritti et al. (1989): the historical control data of the rat colony used revealed a broad variation and the finding may well have been a chance effect unrelated to treatment. No reply of Soffritti to this criticism is available
- in spite of the extensive histopathological evaluation done initially, in 1989 the total number of animals with hemolymphoreticular tumours increased from 79 (in 1989) to 150 (in 2002). But the investigators have not given any explanation/justification. for this massive discrepancy reported in the two publications on the same study.
- In addition, recently severe doubts were published concerning the scientific robustness of studies from this laboratory (Schoeb et al., 2009; Ward and Alden, 2009). The critique is mainly related to the observation of increased incidences of lymphomas after exposure to several chemicals reported in the decade between 1995 and 2005, and the validity of studies conducted by the Soffritti laboratory is questioned especially regarding the health status and historical control tumour incidences of the animals used.
In an initiation/promotion study (Takahashi et al., 1986) male Wistar rats (n=10-30) were initiated with 100 mg/L MNNG (N-methyl-N'-nitro-N-nitrosoguanidine) in the drinking water (ad libitum) plus 10% sodium chloride in the diet (ad libitum) for 8 weeks followed by promotion for 32 weeks with 0.5% formaldehyde in the drinking water (n=17 for evaluation). Additional control groups received only initiation (n=30) or only 0.5% formaldehyde or only vehicle. All groups were sacrificed after 40 weeks. Body weight gain was decreased in all treatment groups compared to vehicle control suggesting toxic effects of initiation and promotion procedure. In accord with other long-term drinking water studies, local lesions in forestomach (papilloma, erosion and ulcer at the limiting ridge of fundic mucosa) and glandular stomach (hyperplasia) were induced by formaldehyde but no malignant tumours developed. Forestomach papillomas were found in 8/10 formaldehyde-only treated rats and in none of the control animals. The occurrence of forestomach papillomas, especially after such a short treatment period, is in clear contrast to the findings of Til et al. (1989) and Tobe et al. (1989). This discrepancy was discussed by Til et al. (1989) as being possibly related to the use of different criteria for the classification of a lesion as papilloma or as papillary epithelial hyperplasia. At that time widely accepted criteria to distinguish both alterations were not available. Promotion by formaldehyde in initiated rats increased hyperplastic lesions in the glandular stomach but did not significantly increase the total number of adenocarcinomas in the gastroduodenal epithelium. Conclusion: In rats oral exposure to 0.5% formaldehyde in the drinking water induced local lesions in stomach but there was no clear evidence for promoting activity.
In summary, the weight of evidence shows that formaldehyde does not induce systemic or local carcinogenic effects after oral exposure.
No studies according to current guidelines were published. The available data on this endpoint showed local irritation but no carcinogenic effects. In an initiation/promotion study (Krivanek et al., 1983) 30 female CD-1 mice per group received a single dermal application of an initiation solution followed two weeks later by application of appropriate promoter solutions; the promoter treatments were administered three times a week for 26 weeks followed by a post exposure observation period of 26 weeks. Nine different dose regimens were performed including appropriate negative and positive controls (BaP for initiation and TPA for promotion). Initiation with formaldehyde was done with 5 mg formaldehyde in 50 μL vehicle (10% solution, irritant) and promotion with 1, 0.5, or 0.1 mg formaldehyde in 100 μL vehicle (concentration 1, 0.5, 0.1%, respectively; slight local irritation at 0.5%). Formaldehyde did not initiate or promote skin tumourigenesis. Formaldehyde did not alter the latency time to first tumour. The negative and positive controls were valid. Malignant tumours were induced only in the positive control (initiator BaP/ promoter TPA). In conclusion, formaldehyde did not initiate or promote skin tumourigenesis in mice and did not alter the latency time to first tumour in an initiation-promotion study at a concentration of 1%.
The initiation-promotion study of Iversen (1986) in mice (suggesting reduced latency time to first tumour) is insufficient for evaluation of carcinogenic or promoting effects.
In summary, the limited data indicate that formaldehyde does not act as a complete carcinogen or as a promoter or initiator on the skin after topical application.
Data on rats
There is clear evidence from chronic inhalation studies in rats that formaldehyde causes tumours in the nasal cavity of males and females; the most prominent tumour type identified in the nasal cavity was the squamous cell carcinoma, occasionally also other tumour types were found in the nasal cavity. A sharp increase of these tumour incidences was evident at concentrations >6 ppm indicating a non-linear dose response. In all studies cited, no systemic or carcinogenic effects were detected at other sites.
In a study comparable to OECD guideline 453 (CIIT, 1981; Kerns et al., 1983) inhalation exposure of male and female F344 rats for 2 years at dose levels of 0, 2, 6, 15 ppm (120 rats per sex per dose; 24 months exposure, 6 h per day, 5 days per week; post exposure period 6 months; interim sacrifices) resulted exclusively in local effects of the nasal cavity and of the proximal trachea in males and females. Only non-neoplastic effects were reported in the high dose group in the proximal trachea but a high incidence in squamous cell carcinomas of the nasal cavity (51/120 males and 52/120 females). Carcinogenic effects in the mid dose group (squamous cell carcinomas in nasal cavity of 1 male and 1 female) are not of statistical significance but possibly of toxicological relevance (extremely low background incidence, in IARC 1976: Pathology of tumours in laboratory animals; according to Borman et al. [Pathology of the Fisher rat, Academic Press, 1990] the spontaneous tumour incidence is 1/1936 in untreated male F344 rats and 2/1949 in corn oil controls, no spontaneous incidences in females). At 2 ppm purulent rhinitis, epithelial dysplasia, and squamous metaplasia were present in the anterior part of the nasal cavity. Dysplasia occurred earlier than metaplasia. Non-neoplastic effects in the mid dose group like dysplasia and squamous cell metaplasia in the nasal cavity occurred also in more posterior parts or the nasal cavity. Reversibility of these non-neoplastic effects have been shown after 3 months post exposure period. At month 27 there was a significant decrease in the incidence of squamous metaplasia at all exposure concentrations regressing predominantly from the more posterior parts of the nose to the anterior sections (see Section Repeated dose toxicity). In summary, non-neoplastic local effects in the nasal cavity of rats were reported even at the low dose level of 2 ppm in a long-term inhalation study in rats; tumours in the nasal cavity were found at a dose level of 15 ppm.
In the chronic study of Monticello et al. (1996; restricted to effects in the nasal cavity) in male F344 rats (0, 0.7, 2, 6, 10, 15 ppm, 6 h/day, 5 days/week) inhalation exposure for 2 years induced local non-neoplastic effects in the nasal cavity of rats at a dose level of 6 ppm, the NOAEC was 2 ppm; clearly carcinogenic effects were found at a dose level of 10 and 15 ppm.
In the inhalation study of Kamata and coworkers (1997; only 32 rats per dose and interim sacrifice) male F344 were exposed 6 h/day, 5 days per week for 28 months. Rats exposed to 0.3 ppm for 28 months did not show any histopathological changes in the nasal cavity. At 2.2 ppm a significant increase was noted in the incidence of squamous cell metaplasia with and without hyperplasia but no carcinogenic effects. Clear local carcinogenic effects were noted at a dose level of 15 ppm.
In male Sprague-Dawley rats (n=100 per dose level, lifetime exposure to 0 or 15 ppm, 6 h/day, 5 days/week) incidences in epithelial hyperplasia as well as squamous metaplasia (low incidence) in larynx and trachea were increased. Malignant & benign tumours were recorded in the nasal cavity of exposed rats but none in controls, no tumours were detected in larynx, trachea or lung (Sellakumar et al., 1985).
The results in Wistar rats (low incidence at 10 ppm, Feron et al., 1988; Woutersen et al., 1989) are limited by a small group size or short exposure duration. Nevertheless, there might exist strain differences in sensitivity comparing F344 with Wistar rats.
There is also some evidence for induction of nasal tumours at 20 ppm at the end of life even after subchronic exposure periods (Feron et al., 1988). Non-neoplastic lesions induced at 10 ppm partly persisted during the life time post observation period.
Other tumour types than the squamous cell carcinoma were found in the nasal cavity: squamous cell papilloma (3/32 rats at 15 ppm, 3/32 rats; Kamata et al., 1997), rhabdomyosarcoma (1/90 at 10 ppm, 1/147 at 15 ppm; Monticello et al., 1996), adenosarcoma (1/90 at 10 ppm, 1/147 at 15 ppm; Monticello et al., 1996), fibrosarcoma (1/100at 15 ppm; Sellakumar et al., 1985), polypoid adenoma (5/90 rats at 10 ppm & 14/147 rats at 15 ppm; Monticello et al., 1996), undifferentiated carcinoma or sarcoma (2/117 at 15 ppm, CIIT, 1981). The predominant localisation of tumours (mainly the squamous cell carcinoma) in the nasal cavity was the anterior portion of the lateral side of the nasal turbinate and the adjacent lateral wall or the mid-ventral septum (CIIT, 1981).
Data on other species
In repeated inhalation toxicity studies on mice it has been shown that the mouse is less sensitive than the rat. Generally the lesions in mice were less severe than in rats using the same experimental design (CIIT, 1981; Kerns et al., 1983). This might also be true for carcinogenic effects in long-term inhalation studies in B6C3F1 mice (CIIT, 1981; Kerns et al., 1983). Only 2 squamous cell carcinomas in the nasal cavity of male mice were detected in the high dose group (15 ppm). Location and morphology were similar to those observed in rats. However, only 25 male mice survived a minimum of 18 months in the high dose group (reduced survival not treatment-related). In females the survival was not reduced but no tumour was detected in the nasal cavity. Non-neoplastic lesions (see Section Repeated dose toxicit) like rhinitis, dysplasia, squamous metaplasia were induced, but also atrophy of olfactory epithelium with increasing exposure periods at the high dose level (NOAEC is 2 ppm and the LOAEC 6 ppm). At 27 months (3 months post exposure period) dysplastic epithelial lesions were only present in the 15 ppm group. Squamous metaplasia was not present at this time interval and the low- and intermediate-exposure groups were free of treatment-related effects.
Also hamsters were less sensitive than rats. In the long-term inhalation study on Syrian golden hamsters (Dalbey, 1982) at formaldehyde concentration of 10 ppm (5 h/d; 5 d/week) in 5% of the animals (4/88), metaplasia and hyperplasia in the nasal epithelium was found but no tumours were detected in the respiratory tract. In addition, treatment with 30 ppm (5 h/d; but only 1 day/week) did not lead to respiratory tract tumours (no further details available).
In a recent review by McGregor et al. (2007 & 2010) the authors summarised and evaluated the available experimental data. They concluded that sustained cytotoxicity and cell proliferation are key events in the proposed mechanism of action (MOA) for formaldehyde- induced nasal tumours in rats. Cytotoxicity, DPC formation, nasal epithelial cell regenerative proliferation, squamous metaplasia, and inflammation have been measured in rat studies and are considered to be site-specific, highly non-linear concentration response processes in concordance with the incidence of nasal tumours. Based on the weight of evidence, it is likely that this MOA is relevant to humans, at least qualitatively. Several epidemiological studies have indicated an increased risk of nasal cancers with formaldehyde exposure (see below). It seems unlikely for the authors that this MOA could result in increased tumour incidences at sites distant from first contact.
In a recent short term carcinogenicity study in genetically susceptible mice the potential role of the Trp53 gene in formaldehyde-induced nasal carcinogenicity, leukemia or lymphohematopoietic cancer, and potentially other neoplasms in genetically susceptible mice was investigated (NTP 2017). Mutations in the tumor suppressor gene Trp53 have been associated with formaldehyde-induced nasal tumors and might be a key mechanistic event in formaldehyde-induced leukemia. Male, Trp53 haploinsufficient (Trp53+) mouse strains (B6.129-Trp53tm1Brd and C3B6.129F1-Trp53tm1Brd) were exposed to 0-, 7.5- or 15-ppm formaldehyde (25/group) 6 h/d, 5 d/wk for 8 weeks, and then held for 32 weeks. Blood was collected for hematology, and major tissues and gross lesions were collected for histopathology. The primary formaldehyde-related finding was squamous metaplasia of the respiratory epithelium of the nose. Inhalation of a maximum tolerated dose of formaldehyde caused significant injury to the nasal mucosa and cell proliferation, but did not cause nasal tumors or an increased prevalence of leukemia or lymphohematopoietic cancer in Trp53+ mice. All observed neoplasms were considered background lesions for these mouse strains. The results of this short-term carcinogenicity study do not support a role for Trp53 in formaldehyde-induced neoplasia.
In a commentary to this NTP study, Thompson (2018) concluded that hese findings lend additional weight to the evidence that inhaled formaldehyde is not leukemogenic—including reanalysis of epidemiological studies and animal studies that indicate that inhaled formaldehyde does not distribute beyond the nasal cavity or reach the blood or bone marrow
Data on carcinogenicity in humans
The possible association between formaldehyde exposure and cancer has been investigated in numerous epidemiological studies in occupationally-exposed humans (e.g., pathologists, anatomists, embalmers, or industrial workers). Because the International Agency for Research on Cancer (IARC) evaluated formaldehyde in 1996, 2004 and 2009 an overview on the epidemiological knowledge about the carcinogenic potential of formaldehyde is ordered in accordance to these time points. A tabulated summary from cohort studies published before 1996 is given in IUCLID Section 7.10.2. New data on risk measures from cohort studies on cancers that were published after 1996 are summarised below. Epidemiological data are also available from case-control studies. With regard to toxicokinetic data and the results in long-term laboratory animal studies most epidemiological studies focused on carcinogenic effects in the respiratory tract, the site of first contact. In addition, in three out of four recent cohort studies all cancer types are investigated (Coggon et al., 2003; Hauptmann et al., 2003 & 2004, Pinkerton et al., 2004). The overall results suggested an increased risk of cancer only at two tumour sites: the upper respiratory tract and the haematopoietic system. However, recent reassessment of the Hauptman et. al. (2003, 2004) data by the U.S. National Cancer Institute suggest these findings may not be significant (Beane-Freeman, et. al. 2009 ab) as reviewed by Marsh et. al. (2010) and supported by Bachand, et. al. (2010)
Summary Table on new data on risk measures from recent cohort studies on cancer (after 1996)
Exposed cohort Referent group
Tumour site (ICD): number of deaths
Risk measure SMR (95% CI)
7359 male & female workers at a plastic-producing plant (2735 deaths) Local county population
Nasopharynx (147): 7 Pharynx (146-149): 22 Lung (162): 262
5.00(2.01-10.3) 2.23(1.40-3.38) 1.21(1.06-1.36)
Nasopharyngeal cancer only in exposed workers, but not related to DOE and weakly related to Doe, cumulative exposure and AIE; SMR increased with TSFE
Marsh et al., (2002)
14,014 male workers at 6 British chemical factories (5185 deaths) National population
Pharynx (146-149.1): 6 Larynx (161): 7
Lung (162): 272 Leukaemia (204-208):8 Multiple myeloma (203): 7
1.91 (0.70-4.17) 1.56 (0.63-3.22) 1.58(1.40-1.78) 0.71 (0.31-1.39) 1.18 (0.48-2.44)
SMR related to men with high exposure (> 2 ppm)
Differentiation in exposure categories only for lung cancer (exposure related increase in SMR)
Coggon et al. (2003)
11039 male & female workers at 3 garment factories (2206 deaths) General population
Pharynx (146-149): 3 Lung (162): 147 Leukaemia (204-08): 17 Myeloid leukaemia (205): 9
0.64 (0.13-1.86) 0.98 (0.82-1.15) 1.78(1.04-2.86) 2.24(1.02-4.25)
MCOD analysis used for leukaemia and myeloid leukaemia, SMR related to workers exposed for ≥ 10 years
Pinkerton et al. (2004)
25619 male & female workers at 10 plants (8486 deaths) SMR: general population; RR: low exposure group This study included as Plant 1 the cohort analyzed by Marsh et al. (2002)
Nasopharynx (147): 8 Lung (162): 641 Leukaemia (204-207):65 Nasopharynx (147): 7 Nasopharynx (147): 6 Nasopharynx (147): 3 Nasopharynx (147): 2 Leukaemia (204-07): 29 Myeloid leukaemia (205): 14 Leukaemia (204-07): 17 Myeloid leukaemia (205): 9
SMR 2.10(1.05-4.21) 0.97 (0.90-1.05) 0.85 (0.67-1.09) RR 1.83 1.67 4.14 4.18 2.46(1.31-4.62) 3.46(1.29-9.43) 1.68 (0.91-3.08) 2.49(1.03-6.03)
SMRs only for nasopharyngeal cancer increased but RR also for (myeloid) leukaemia peak exposure ≥ 4 ppm* average intensity ≥ 1 ppm cumulative exposure ≥ 5.5 ppm-years* duration of exposure >15 years peak exposure ≥ 4 ppm** peak exposure ≥ 4 ppm** average intensity ≥ 1 ppm average intensity ≥ 1 ppm
Hauptmann et al. (2003) (leukaemia) & (2004) (nasopharyngeal cancer)
Due to an incomplete follow-up these studies are not reliable although used as pivotal evidence by IARC (2006)
Continuation of the data presented in Hauptmann et al; 2003; follow-up of the 25619 male & female workers through December 31, 2004
all lymphohematopoietic malignancies (200-209) 286 Hodgkin disease (201) 25 multiple myeloma (203) 48 all leukaemia (204-7) 116 myeloid leukaemia (205)44 all lymphohematopoietic malignancies (200-9) 108 Hodgkin disease (201) 11
SMR 0.94 (0.84 to 1.06) 1.42 (0.96 to 2.10) 0.94 (0.71 to 1.25) 1.02 (0.85 to 1.22) 0.90 (0.67 to 1.21) RR 1.37(1.03 -1.81) 3.69(1.31-12.02)
exposed workers, all SMRs non-significant peak exposures ≥ 4.0 ppm vs lowest peak exposure dito all other malignancies RR non-significant
Beane Freeman et al. (2009a)
CI: confidence interval; SMR: standard mortality ratio; RR: relative risk; ICD: International classification of diseases code; DOE: duration of employment; Doe: duration of exposure; TSFE: Time since first employment; AIE: average intensity of exposure; MCOD: Multiple causes of death analysis; bold type: trend test significant
Indication for an excess of death from nasopharyngeal tumours in industrial workers (Blair et al., 1986; Collins et al., 1988; Marsh et al., 1996; all three studies refer to the same study population) and embalmers (Hayes et al., 1990) was presented in former cohort studies and a proportionate incidence study (Hansen & Olsen, 1995; SPIR 1.3 for nasopharynx and SPIR 2.3 for nasal cavity tumours); however, no increased risk was detected in other studies (see IUCLID Section 7.10.2 - Additional Information, Table 1).
Some evidence for an association between formaldehyde exposure and nasopharyngeal tumours was also reported by Marsh et al. (2002). Although the SMR for all pharyngeal cancer and nasopharyngeal cancer were significantly increased, no clear relation to the level of exposure was found when different exposure categories were considered (see Table above). Most pharyngeal cancer (PC) and nasopharyngeal cancer (NPC) occurred in workers hired between 1947 and 1956 (higher exposure levels presumed). But short-term workers showed a higher death rate of PC or NPC than workers employed for > 1 year. Neither PC nor NPC were associated with duration of employment but SMR (as well as number of deaths) increased with time since first employment. Only little evidence of increasing mortality risks with increasing duration of exposure, cumulative exposure or AIE (average intensity of exposure) was seen. There is some evidence for increasing risks for PC with increasing duration of exposure with average exposure >0.2 or 0.7 ppm. Concerning especially the NPC data it should be noted that subgroup size of 1-3 cases resulted in limitation of the statistical power.
Evidence for an association between formaldehyde exposure and nasopharyngeal cancers came from a publication of Hauptmann et al. (2004). The cohort study of Hauptmann et al. (2004) is the largest and most informative study in industrial workers concerning the number of study subjects in the cohort. It includes as Plant 1 the cohort analyzed by Marsh et al. (2002). Hauptmann et al .(2004) have shown a statistically significant exposure-response relationship for peak exposure and cumulative exposure compared with an internal control group (see Summary Table above). The peak exposures are defined as the highest peak exposure ever identified during work history (see the critical discussion of this unusual exposure metric in Marsh and Youk, 2005). In comparison with the general population, the SMR for nasopharyngeal tumour was also increased(SMR = 1.97; 0.95-CI: 0.94-4.14).The cohort from Hauptmann et al. (2004) has been re-evaluated by Marsh and Youk (2005). They point out that 6 of the 10 deaths due to nasopharyngeal cancer occurred only in Plant 1, while the remaining four cases occurred individually in four of the other nine plants. Furthermore a comprehensive exposure analysis for the cases in Plant 1 showed no clear relation to exposure as had already been pointed out in Marsh et al. (2002). Only three of the cases were exposed to formaldehyde longer than one year and each has had low average intensity of exposure.
Marsh and Youk (2005) hypothesised that the increased risks for nasopharyngeal cancer in Plant 1 are due to occupational or non-occupational exposure to risk factors outside Plant 1 (see study by Marsh et al 2007b below for follow-up on this hypothesis).
Similar criticism came from Tarone & McLaughlin (2005). In their re-evaluation of the data presented by Hauptmann et al. (2004) the calculated SMR for nasopharyngeal cancer in exposed workers in Plant 1 was 9.1 (95% CI: 3.3-19.8) (unexposed workers SMR 0.0) but in Plant 2-10 the SMR in exposed workers was 0.6 (95% CI: 0.1-2.3) (unexposed workers SMR 1.9). They argued that the absence of an increased risk in exposed workers of Plants 2-10 and the magnitude of the differences in SMRs raises serious questions about the interpretation of the results. However, Tarone and McLaughlin focussed on ever/never exposure comparison (all exposure categories combined) and compared with external rates. In their reply Hauptmann et al. (2005) suggested the internal comparison (used in Hauptmann et al., 2004) for comparison which is more informative without a healthy worker bias. However, even the comparison with an external group resulted in increased SMRs for Plant 1 and Plants 2-10 at the highest level for all four exposure categories (Hauptmann et al., 2005). Furthermore, the homogeneity of SMRs for Plant 1 versus Plants 2-10 was not rejected, except peak exposure indicating no clear pattern of risk heterogeneity between Plant1 and Plants 2-10 (Hauptmann et al., 2005).The authors stated that all six workers with nasopharyngeal cancer in Plant 1 had been in the highest peak exposure category of ≥ 4 ppm (compare with criticism presented by Marsh & Youk (2005, Section Human data).
Marsh and co-workers (Marsh et al., 2007a, Section Human data) substantiated their criticism in a recent re-evaluation of the Hauptmann study (Hauptmann et al., 2004).They performed two types of re-analyses with focus on peak exposure and NPC mortality. First, they performed a formal interaction analysis to examine whether the risk estimates varied considerably by plant (“effect modification”). The results substantiated the arguments of Tarone and McLaughlin (2005): the findings are heterogeneous (plant 1 vs. plant 2-10) and cannot be validly combined into one estimate and, thus, cannot be generalized to other populations. The interaction analysis also showed convincingly that the model applied by Hauptmann et al. (2004) was inappropriate. Marsh and co-workers suggested using these extended modeling techniques in future updates of the NCI cohort study to avoid model misspecifications. Second, they performed a sensitivity analysis to study whether the results reported by Hauptmann et al (2004) suffer from uncertainties not covered by the reported confidence intervals due to the low number of cases (“small sample bias”). Their Monte-Carlo sensitivity analyses revealed substantial instability problems with the exposure-response relationships related to Plant 1. The authors stated that their results do not support a causal association with formaldehyde exposure and nasopharyngeal cancer. Furthermore, Marsh et al. (2007b) investigated the possibility that the large NPC mortality excess among workers of plant 1 (see also criticism by Marsh & Youk, 2005) may be related to occupational factors external to this study plant (Wallingford cohort). The results in their nested case-control study suggested that the NPC mortality excess in the Wallingford cohort may reflect the influence of external employment in the local metal industries that entailed possible exposures to risk factors for upper respiratory tract cancer like sulphuric acid mist, mineral acids and metal dusts. Marsh and co-workers suggested further updates of the Hauptmann study that should include co-exposure data on silver smithing and other metal work for all 10 plants. This might help to explain the findings for NPC in plant 1 in comparison with the other 9 plants.
Importantly, Beane-Freeman et al. (2009a) recently reported that 1006 deaths were not included in the data analysed by Hauptmann et al. (2004). An explanation could not be given why these 1006 deaths were not included. Marsh et al (2010) described and discussed this incomplete follow-up in the National Cancer Institute's formaldehyde worker study and the impact on subsequent re-analyses and causal evaluations.The addition of these deaths increased mortality by 10.5% for the study group. However, the deaths were not missed at random. The percent increase in corrected numbers of deaths among “unexposed” subjects is approximately two times greater than that observed among “exposed” subjects for all deaths, all cancer deaths and all solid cancers. This may have a pronounced effect on the dose-response relationships as they have been overestimated potentially by Hauptmann et al (2004). Furthermore, according to the tables published in Beane-Freeman et al (2009b) the corrected increase for total deaths by exposure status equals 995, not 1,006 as described by Beane-Freeman (2009a). The noted difference creates doubt whether all ascertainment errors were addressed by the NCI working group as the consistent differential change in cause-specific numbers of deaths by exposure category cannot likely be explained as a chance occurrence thereby raising serious doubt about the validity of the Hauptmann (2004) analysis. Aside from the combined category “all solid cancers”, the NCI working group has not reported corrected estimates for any site-specific cancer mortality corresponding to solid tumour types, particulary NPC.
Thus, Marsh and co-workers asked for a prompt corrigendum of the 2004 publication by the NCI working group since this study has played such a prominent role in causal evaluations. Finally it should be noted that an update of the Hauptmann et al. (2004) study on NPC has been initiated by NCI. Thereby the mortality from solid cancers that was analysed by Hauptmann through 1994 will be traced for another 10 years and include the results from the 995/1006 additional cases. A final assessment of the mortality by NPC should wait for the results of this NCI industry cohort update and clarifications of the NCI data base.
Other studies do not support a causal relationship between formaldehyde and NPC deaths. Collins et al (1997) summarized all studies published up to 1996 and found 14 cohort studies showing RR=1.0 (0.95-CI: 0.5-1.8). Aside from the NCI formaldehyde industry cohort study two other big cohort studies were performed after 1996. A slight but statistically not significant increase in pharyngeal cancer was found in the study of Coggon et al. (2003) in 14014 workers from 6 British plants. While only one death from NPC was observed with 2.0 expected, leading to an SMR=0.5 (0.95-CI: 0.01-2.79), the man had not worked in a job with high formaldehyde exposure. No nasal or nasopharyngeal tumours were found in the cohort (11039 workers from 3 plants) studied by Pinkerton et al. (2004) and only 3 deaths were related to cancer of the pharynx clearly limiting the statistical power of this study to investigate this endpoint . Indication for an association between formaldehyde exposure and nasopharyngeal tumours was presented in case-control studies. Several studies on nasopharyngeal cancer found an increased risk for an exposure to formaldehyde, mainly in subjects with the highest probability, level or duration of exposure (Vaughan et al., 1986b & 2000; Roush et al., 1987; West et al., 1993; Hildesheim et al., 2001) but two other studies did not show such an association (Vaughan et al., 1986a; Armstrong et al., 2000).
Collins et al (1997) summarized all studies published up to 1996 and found 15 case-control studies with an overall OR=1.3 (0.95-CI: 0.9-2.1).Vaughn et al 2000 (USA: 196 cases / 244 controls: OR= 1.3, 0.95-CI: 0.8-2.1) and Hildesheim et al 2001(Taiwan: 375 cases / 325 controls: OR= 1.4, 0.95-CI: 0.9-2.2) were the two case-control studies that were published afterwards and reported similar findings. However, although they showed some evidence of an increase in NPC risk the exposure levels in these studies were uncertain. Thus, these investigations are clearly limited.
In the meta-analysis presented by Bosetti et al. (2008) nine deaths from nasopharyngeal cancer in three cohorts of industry workers yielded a pooled SMR of 1.33, which declined to 0.49 after excluding six cases from one US plant (plant 1, see Hauptmann et al., 2004, Marsh et al 2007): the non-significantly increased SMR for nasopharyngeal cancer among industry workers is attributable to a cluster of deaths in a single plant. Thus, they confirmed the observation of Marsh et al. (2005), Tarone and McLaughlin (2005)and Marsh et al. (2007a). Similarly, after analysing cohort and case-control studies Duhayon et al. (2008) concluded that the human studies fail to raise a convincing conclusion concerning NPC mortality being related to formaldehyde.
Bachand et al. (2010) performed another statistical summary of epidemiological studies and found that estimates for NPCs were not elevated after excluding a single plant with an unexplained cluster (cohort RR = 0.72, 95% CI: 0.40, 1.28), again giving support to the observations by Marsh et al. (2005), Tarone and McLaughlin (2005) and Marsh et al. (2007a). The summary estimate was increased for case-control studies overall, but the summary OR for smoking-adjusted studies was 1.10 (95% CI: 0.80, 1.50). Bachand et al. concluded that there is little support for a causal relationship between formaldehyde exposure and nasopharyngeal cancer.
In a review summarizing the current knowledge regarding the epidemiology of NPC the major risk factors were identified by Chang and Adami (2006), i.e. elevated antibody titers against the Epstein-Barr virus, a family history of NPC,consumption of salt-preserved fish and certain human leucocyte antigen class I genotypes. They concluded that the epidemiological evidence for an increased risk of NPC following workplace exposure is limited and that occupational or non-occupational exposures other than formaldehyde may have been responsible for excess NPC mortality.
Summary and evaluation of Nasopharyngeal cancer (NPC)
Regarding the association between formaldehyde exposure and solid tumours IARC (2006) concluded:
- sufficient evidence for NPC was mainly based on the NCI industrial cohort study (Hauptmann et al. 2004) and supported by positive findings from other studies
- limited evidence for sinonasal cancer
- no evidence for other sites (incl. oral cavity, oro- and hypopharynx, pancreas, larynx, lung, brain).
The IARC (2006) evaluation for NPC was confirmed by an IARC working group in October 2009 (Baan et al., 2009).
A literature search after the last IUCLID update was carried out up to May 2022.
As NPC is one of the critical tumor types possibly associated with formaldehyde exposure, some studies on NPC not directly related to formaldehyde are briefly mentioned here.
Liu et al. (2012) used a screening program for NPC based on detection of antigens against the Epstein-Barr virus, examination of the nasopharynx and lymphatic palpation. Among 28,688 individuals from a high risk area in China, NPC was detected in 41 persons (0.18%) and in 28 of these in an early stage. Xu et al. (2012) carried out a case-control study to investigate potential NPC risk factors in a high risk area of China with 1316 NPC cases and 1571 controls from the same region, 1657 controls from another high risk area, and 1961 controls from a low risk area. Smoking was the only life style factor significantly associated with NPC and was also linked to Epstein-Barr virus seropositivity among the control populations. Specific workplace related analyses were not carried out. Fachiroh et al. (2012) studied life style and genetic factors among 681 cases and 1078 controls from Thailand. Again this study did not include analyses of the workplaces. Frequent consumption of fermented vegetables, smoking, and some specific genetic variants were significantly associated with NPC. Polesel et al. (2013) studied the risk for NPC in relation to dietary habits in 198 NPC cases and 594 controls from Italy, a low risk region for NPC. Elevated vegetable consumption was significantly inversely related to NPC and this effect was particularly strong for yellow and red pigmented vegetables. Elevated egg consumption was significantly associated with NPC. Xue et al. (2013) conducted a meta-analysis on studies over the last 30 years and identified 28 case-control and 4 cohort studies with a total of 10274 NPC cases and 415266 controls. A significant risk for NPC was identified with an OR of 1.6 (95% CI 1.38; 1.87) with a robust association for smokers compared to never smokers.
Beane-Freeman et al (2013, key) further extended the follow-up of the NCI cohort of workers in FA industries (n = 25 619) through 2004. During 998 239 person-years, 13 951 deaths occurred. With one additional death, albeit occurring in the lowest exposure category, previously observed excesses for nasopharyngeal cancer (n = 10) persisted for peak, average intensity and cumulative exposure; RRs in the highest exposure categories were 7.66 (95 % CI: 0.94–62.34), P-trend = 0.005, 11.54 (95 % CI: 1.38–96.81), P-trend = 0.09, and 2.94 (95 % CI: 0.65–13.28), P-trend = 0.06, respectively. For all cancer, solid tumours and lung cancer, standardised mortality ratios (SMRs) among exposed workers were reported to be elevated, but internal analyses, as described by the authors, revealed no positive associations with FA exposure.
A follow-up through December 2012 was conducted in the British (UK) cohort from six factories comprising 14 008 men in the period 1941–2012 (Coggon et al 2014, key). In the period, 7 378 men had died. 3,991 were at some time highly exposed. In the whole population, the standardised mortality ratio [SMRs (95 % CI)] for all cancers [1.10 (1.06–1.15)], stomach [1.29 (1.11–1.49)], rectum [1.23 (1.01–1.49)], and for lung cancer [1.26 (1.17–1.35)] was significantly increased based on the national death rate for England and Wales. There was no excess mortality from NPC; the only death occurred in a man with low/moderate exposure (1.7 deaths expected for exposures above background). Additionally, the authors included a nested case-control analysis of cancer in the upper airways, larynx, mouth, pharynx, tongue, and for all leukaemia and myeloid leukaemia. ORs for these cancers were independent of the duration of the exposure. The authors ascribed the increases in risk estimates to non-occupational confounding factors, which may include smoking and socioeconomic factors and concluded that the study provided no evidence that FA posed an increased hazard either of upper airway cancer. It was noted that the study was not able to take smoking and socioeconomic factors into account.
A follow-up has been conducted on the US NIOSH garment industry cohort (Meyers et al 2013, key), which is one of the three largest prospective cohorts. The study comprised 11 043 workers. Causes of death were obtained from 99.7 % (3 904) of the identified deaths. About 77 % had year of first exposure in 1970 or earlier. In the early 1980s, personal FA sampling was performed among 549 employees. The geometric mean FA concentration was 0.15 ppm with a geometric standard deviation of 1.90. The SMRs were similar to that of the US population for all cancers, for buccal cavity and pharyngeal cancers and for respiratory cancers. The authors concluded that the study showed little evidence for an increased risk of mortality from buccal cavity, pharyngeal (including nasopharyngeal) or respiratory cancer.
An Italian cohort with subjects employed in a factory producing laminate plastic, decorative papers and craft papers, using phenolic and melamine resins, has been established Pira et al. (2014, supporting). The major risk was considered to be FA exposure, but FA concentrations were not reported. The cohort comprised 2750 employees from the period 1947 to 31 May 2011, who have been employed at least 180 days. Data on survival (80.3%) and death (16.6%, N=457) were collected. Cause of death could not be retrieved for 26 out of 457 (5.7%) deceased employees. Person-years of observation were 70,933. Expected number of death (E) and SMRs were obtained by comparison with the regional deaths rates. The risk for all cancers was not increased (149; 0.80 (0.68-0.94) and there was a non-significantly increased risk for oral and pharynx cancer (9; 1.49 (0.68-2.82) and for bladder cancer (10; 1.51 (0.72-2.77)). The study has a long follow-up period, but a limitation is the lack of quantitative FA exposures and the relatively small cohort.
Airway cancers associated with FA exposures were studied in a Finnish cohort with 1.2 million employees (Siew et al 2012, key). All men born between 1906 and 1945, and employed during 1970 were included. The follow-up was in the Finnish Cancer Register for nasal cancer (292 cases), cancer of the nasopharynx (149 cases) and lung cancer (30 137 cases) during the period 1971–1995. The Finnish job-exposure matrix was used to estimate exposures. Duration of exposure was estimated from census data. A latency period of 20 years was accepted. Number of exposed cases (N), relative risk (RR) obtained by comparison with unexposed, and 95 percent confidence intervals were estimated (N; RR (95 % CI)). The risks of nasal cancers (17; 1.1 (0.6–1.9), nasal squamous cell carcinoma (9; 1.0 (0.4–2.0)) and nasopharyngeal cancer (5; 0.9 (0.3–2.2)) were not increased. The risk was slightly increased for lung cancer (1 831; 1.2 (1.1–1.3)). However, the risk in the highest exposure group (≥ 1 ppm) was not increased. Thus, the authors considered the increased risk to be due to residual confounding effects of smoking and co-exposures, including asbestos and crystalline silica. FA exposures were below 1 ppm in most occupations. Only flour layers, and varnishers and lacquers had average exposures at 1 ppm. Overall, this study found no increase in portal-of-entry cancer at low FA concentrations in occupational settings.
In conclusion, the weight of evidence from the most recent studies does not give any additional support for an association between formaldehyde and NPC and the classification of IARC as a group 1 carcinogen for this tumor type.
Therefore, the strength of the epidemiological data for NPC is not sufficient for a classification of formaldehyde as a human carcinogen within the EU regulatory framework taking into account that
- only in one of the three large cohort studies an association between formaldehyde and NPC was found
- this association was restricted to one specific plant
- there is evidence that pre-employment factors may have been the cause for NPC formation in this plant
- the stability of the statistical analysis and the appropriateness of the statistical model of Hauptmann et al. (2004) may be questioned
- unaccounted deaths raise doubt about the mortality analysis of Hauptmann et al. (2004)
- two recent meta-analyses and a general review on NPCs (not cited in IARC 2006) do not substantiate the association between formaldehyde exposure and NPC
- the ongoing update by NCI for another 10 years survival should be waited for before coming to a final conclusion.
In a pooled analysis of 12 case control studies conducted in 7 countries and adjusted for age, study and other occupational exposure an exposure related increase in the risk of sinonasal cancer, particularly adenocarcinoma, was reported (Luce et al., 2002). Similar results were presented by T’Mannetje et al. (1999). In a Danish study, the odds ratios for squamous cell carcinoma in the nasal cavity or para-nasal sinus were increased (Olsen & Asnaes, 1986). In contrast to these case-control studies, the recently published cohort studies (Coggon et al., 2003; Pinkerton et al., 2004; and Hauptmann et al., 2004) reported no excess of this cancer type.
In the literature search after the last IUCLID update carried out up to May 2022, some further studies on lung cancer were identified:
Regarding lung cancer, IARC (2012) noted that there are statistical significant associations but the results are inconsistent. Therefore, no classification for lung cancer was assigned to formaldehyde.
In most cohort studies no association between exposure and lung cancer was found. However, in two recent studies (Coggon et al., 2003; Marsh et al., 2002 Section Human data) some evidence for an association was provided by SMR analysis. However, both studies did not adjust for smoking and Coggon et al 2003 could not identify any trend in lung cancer mortality across metrics of formaldehyde exposure. No excess in mortality from lung cancer was recorded by Pinkerton et al. (2004, Section Human data) and also by Hauptmann et al. (2004, Section Human data), representing the most detailed and informative investigations concerning this tumour site. In conclusion: there is no evidence on exposure-related increase in tumours of the lung. Furthermore, in case-control studies there has been no increase in lung cancer except for one study (Gérin et al., 1989) which reported an association between lung adenocarcinoma and formaldehyde exposure.
There are numerous epidemiological studies among textile workers on lung cancer indicating reduced risks related to cotton dust exposure presumably due endotoxin exposure. Checkoway et al. (2011, supporting) investigated by a nested case-cohort study within 267400 women textile workers potential associations with other exposures like wool and synthetic fibres, formaldehyde, silica, dyes, and metals. 628 lung cancer cases were compared to a reference subcohort of 3188 workers. A job-exposure matrix was constructed and age/smoking-adjusted hazard ratios (HR) were calculated. No associations for lung cancer were found for most agents, but increased risks, although imprecise, were observed for ≥10 years of exposure to silica (HR 3.5, 95% CI 1.0-13) and for ≥10 years of exposure to formaldehyde (HR 2.1, 95% CI 0.4-11). The authors note that there is only weak prior evidence supporting an association between lung cancer and formaldehyde.
Mahboubi et al. (2013, supporting) conducted two population-based case-control studies in Montreal, Canada, on lung cancer. Identical interviews by questionnaire for the 2 studies were carried out in 1976-1986 and 1996-2002. In all, 2060 lung cancer cases were compared to 2046 population controls. About 25% of the subjects were exposed to formaldehyde. The adjusted OR for formaldehyde was 1.06 (95% CI 0.89-1.27) comparing ever vs. never exposed subjects. No marked increase in lung cancer risk related to occupational formaldehyde exposure was observed. But the authors note that the subjects were mainly exposed to low concentrations: for example, only 3.7-8% of the subjects were exposed to substantial exposure levels as defined by authors for the cumulative exposure index.
In a recent metaanalysis, Kwak et al. (2020) searched for articles on occupational formaldehyde exposure and lung cancer in PubMed, EMBASE, Web of Science, and CINAHL databases. In total, 32 articles were selected and 31 studies were included in a meta-analysis. Subgroup analyses and quality assessments were also performed.The risk of lung cancer among workers exposed to formaldehyde was not significantly increased, with an overall pooled risk estimate of 1.04 (95% confidence interval [CI], 0.97-1.12). The pooled risk estimate of lung cancer was increased when higher exposure studies were considered (1.19; 95% CI, 0.96-1.46). More statistically robust results were obtained when high quality (1.13; 95% CI, 1.08-1.19) and recent (1.13; 95% CI, 1.07-1.19) studies were used in deriving pooled risk estimates. No significant increase in the risk of lung cancer was evident in the overall pooled risk estimate; even in higher formaldehyde exposure groups. Our findings do not provide strong evidence in favor of formaldehyde as a risk factor for lung cancer. However, since risk estimates were significantly increased for high-quality and recent studies, the possibility that exposure to formaldehyde can increase the risk of lung cancer might still be considered.
A study obtained in the literature search up to April 20 2015, is mentioned here. Paget-Bailly et al. (2012, supporting) carried out a metaanalysis on cancer of the laryx based 99 publications on chemicals with at least 10 available publications. Significantly increased meta-relative risks were identified for polycyclic aromatic hydrocarbons, engine exhaust, textile dust, and working in rubber industry. Formaldehyde was not significantly associated with laryngeal cancer on the basis of 11 studies.
Most studies published before 1996 focused on cancer of the respiratory tract because a potential effect of formaldehyde on the respiratory organs appeared to be biologically plausible. On the same grounds no or only weak evidence for an association between formaldehyde exposure and increased mortalities from systemic cancer in humans was expected. However, in two out of three recent studies on industrial workers [see Section Human data and Summary Table above] some evidence for an association between formaldehyde exposure and leukaemia has been reported by two US investigations (Pinkerton et al. 2004; Hauptmann et al. 2003 - however, see discussion below on the continuation of this study by Beane Freeman et al., 2009a,b) but not by the large British study (Coggon et al. 2003).
In an update of the garment worker study (Pinkerton et al. 2004) leukaemia (all types, 24 deaths observed) and myeloid leukaemia (15 deaths observed), identified by a multiple-cause mortality assessment, were non-significantly increased in comparison to the general population. Investigating the exposure metric surrogates “duration of exposure” and “time since first exposure” there was no clear trend in all leukaemias but some indication of a trend in myeloid leukaemia. In summary, the study of Pinkerton et al. (2004) lent some support to the hypothesis of a possible link between formaldehyde exposure and the occurrence of leukaemia, in particular myeloid leukaemia. However, no exposure concentration data was available for analysis in this study.
Coggon et al. (2003) identified 31 leukaemia deaths and reported a decreased risk in the high exposure group, i.e., in jobs with formaldehyde concentration levels estimated to be higher than 2 ppm (SMR = 0.71, 95% confidence interval 0.31-1.39). For comparison, the SMR for all leukaemias was calculated to be 1 for those jobs with exposures less than 2 ppm. However, the exposure assessment was crude and myeloid leukaemias were not evaluated separately.
Influential results from the NCI cohort study analyses were published by Hauptmann et al. (2003) who were able to make extensive use of individual estimates of exposure levels to formaldehyde in their analyses. The authors documented 69 leukaemia cases including 30 deaths from myeloid leukaemia. For all leukaemias, the authors reported a significant increase in the relative risk for peak exposure and the trend was highly significant (referent group: low exposure level; but no analysis was carried out by comparison to the general population). This trend was driven by clearly elevated relative risks for death from myeloid leukaemia (again, highly significant across peak exposure categories). In 2004 a working Group at IARC concluded (IARC 2006): “In summary, there is strong but not sufficient evidence for a causal association between leukaemia and occupational exposure to formaldehyde. 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 and because they conflict with the non-positive findings from the British cohort of industrial workers. Thus, this judgement by the working group at IARC was driven a lot by the results published by Hauptmann et al. (2003).
As already described and discussed in the chapter on nasopharyngeal cancers Beane-Freeman et al. (2009a, b) reported that 1006 / 995 deaths were not included in the data analysed by Hauptmann et al. (2003). An analysis of the corrected data by Beane-Freeman et al. (2009b) showed that the results and conclusions published in Hauptmann et al (2003) were invalid. The addition of 995 deaths to the NCI cohort increased total mortality by 10.5 percent and produced notable corrections to the results published by Hauptmann et al (2003). A detailed discussion is given in Marsh et al (2010). For leukaemia, the correction of number of deaths by 11 percent resulted from seven additional deaths that occurred only among subjects in the lowest category (>0 to 1.9 ppm) of “peak exposure” used by NCI as the referent for relative risk calculations. Hauptmann et al.(2003) reported that the relative risk for leukaemia mortality among those workers classified as having the peak exposure to formaldehyde was 2.46 (95% confidence interval (CI): 1.31-4.62). Upon reanalysis, the corrected relative risk estimate for this outcome was 1.60 (95% CI: 0.90-2.82) (Beane Freeman et al., 2009b). This addition of seven leukaemia deaths to the referent category (Beane Freeman et al., 2009b) considerably attenuated the exposure-response relationship for highest ever peak exposure to formaldehyde and leukaemia, resulting in a trend test p-value of 0.0942 for exposed groups only (the original reported p-value for this trend test based on exposed groups only was 0.004). These new, corrected findings for leukaemia (Beane Freeman et al., 2009b) supported the results of a reanalysis of the 1994 NCI data (Hauptmann et al., 2003) published in Marsh and Youk 2004. This reanalysis showed that the exposure-response relationship for leukaemia originally reported by Hauptmann et al. (2003) was due largely to a deficit in deaths among the low or unexposed subgroups. The uncorrected standardised mortality ratios (SMRs) calculated by Marsh and Youk (2004) were inordinately low and statistically significant for both the unexposed category (SMR=0.38, 95% CI= 0.10-0.97) and lowest peak exposure (>0-1.9 ppm) category (SMR=0.50, 95%CI=0.28-0.81). The most recent NCI results for leukaemia yielded an even more attenuated exposure-response relationship for leukaemia resulting in a trend test p-value of 0.12 among exposed workers only for mortality follow-up through 2004 (Beane Freeman et al., 2009a, the authors now documented 123 leukaemia cases including 48 deaths from myeloid leukaemia). Clearly, the incorrect data for leukaemia in the original 1994 follow-up of the NCI cohort study created misleading evidence for a highly significant exposure-response relationship between peak formaldehyde exposure and leukaemia. All findings about myeloid leukaemia are non-significant in the updated study (Beane Freeman et al., 2009a) although the number of cases available for analysis increased by more than 150%. Only the original (i.e. uncorrected) NCI cohort data were available in 2004 when the International Agency for Research on Cancer (IARC, 2006) reclassified formaldehyde as a known human carcinogen (Group 1) based on findings for NPC and to a lesser extent on NCI’s findings for leukaemia reported by Hauptmann et al.(2003, 2004).
The study of Beane Freeman et al. (2009a) is a continuation of the cohort study presented in Hauptmann et al. (2003). The authors extended the follow-up through December 31, 2004, for 25619 workers employed at 10 formaldehyde-using or formaldehyde producing plants before 1966. Relative risk (RR) estimates and 95% confidence intervals (CI) were calculated to examine associations between quantitative formaldehyde exposure estimates (peak exposure, average intensity and cumulative exposure) and death from lymphohaematopoietic malignancies. Confounding by exposure to other possible leukaemiogenic substances was assessed. There were statistically significant increased risks for the highest vs lowest peak formaldehyde exposure category ( ≥ 4 parts per million [ppm] vs >0 to < 2.0 ppm) and all lymphohaematopoietic malignancies (RR = 1.37; 95% CI = 1.03 to 1.81, P trend = .02) and Hodgkin lymphoma (RR = 3.96; 95% CI = 1.31 to 12.02, P trend = .01). Statistically nonsignificant associations were observed for multiple myeloma (RR = 2.04; 95% CI = 1.01 to 4.12, P trend > .50), all leukaemia (RR = 1.42; 95% CI = 0.92 to 2.18, P trend = .12), and myeloid leukaemia (RR = 1.78; 95% CI = 0.87 to 3.64, P trend = .13). However, the unexposed showed a significantly elevated multiple myeloma mortality when compared to the low exposed (RR=2.74, 95% CI = 1.18 to 6.37) and SMRs were below 1 among the exposed for all lymphohaematopoietic malignancies, multiple myeloma and myeloid leukaemia. An unexceptional SMR was found among the exposed for all leukaemias and a non-significantly elevated SMR for Hodgkin lymphoma. There was little evidence of association for any lymphohaematopoietic malignancy with average intensity or cumulative exposure at the end of follow-up in 2004. Time period analyses have shown that disease associations varied over time. For peak exposure, the highest formaldehyde-related risks for myeloid leukaemia occurred before 1980, but trend tests attained statistical significance in 1990 only. After the mid-1990s, the formaldehyde-related risk of myeloid leukaemia declined. Data suggested that largest risks occurred closer in time to relevant exposure. However, Marsh and Youk 2004, Fig. 2 described the distribution of leukaemia deaths by time since highest peak exposure (data of Hauptmann et al 2003) and showed that the majority of cases occurred after more than 20 years of latency. In conclusion, the updated analyses of the NCI cohort (Beane Freeman et al. 2009a, b) reduced concern about a possible causal association between exposure to formaldehyde and lymphohaematopoietic malignancies, in particular for myeloid leukaemia. Due to open questions regarding data validity and appropriate statistical modeling re-analyses should be waited for that may help clarify the situation (Marsh et al 2010).
Collins and Lineker (2004) performed an overview of all epidemiological studies that investigated formaldehyde exposure and leukaemia risk. They evaluated 8 studies on industrial workers, 7 studies on embalmers and 3 studies on pathologists/anatomists. The authors calculated a summary RR of 1.1 (0.95-CI: 1.0-1.2). The embalmer studies showed a significantly elevated risk of RR=1.6 (0.95-CI: 1.2-2.0) and a borderline statistically significant elevation was found for pathologists/anatomists (RR =1.4, 95%CI 1.0-1.9). Industrial workers showed no excess: RR = 0.9 (95% CI 0.8–1.0). However, according to IARC 1995 higher peak and time weighted average exposure concentrations of formaldehyde were expected to occur among industrial workers. The authors discussed that increased leukaemia rates in embalmer, pathologist or anatomist could be related to exposures other than formaldehyde.
An updated overview of all epidemiological studies published through February 2007 was performed by Bosetti et al (2008).They differentiated between studies of industrial workers and professionals (pathologists, anatomists and embalmers) exposed to formaldehyde. The overall RR for industrial workers was 0.85 (95%-CI: 0.74-0.96) when analysing all lymphatic and hematopoietic cancers, and 0.90 (95%-CI:0.75-1.07) for leukaemia. The overall RR for professionals was 1.31 (95%-CI: 1.16 -1.48) for all lymphatic and hematopoietic cancers, and 1.39 (95%-CI: 1.15-1.68) for leukaemia. Therefore, the authors found modestly elevated risks for lymphohematopoietic neoplasms in professionals, but not industry workers – in support of the result presented by Collins and Lineker (2004).
Zhang et al (2009a) focused on occupations known to have high formaldehyde exposure, analysing a subset of studies used in Collins and Lineker (2004) and Bosetti et al. (2008). Summary relative risks (RRs) were elevated in 15 studies of leukaemia (RR=1.54; confidence interval (CI), 1.18-2.00) with the highest relative risks seen in six studies for myeloid leukaemia (RR=1.90; 95% CI: 1.31 -2.76). All six studies of myeloid leukaemia had relative risks of ≥ 1.4 for the highest exposure categories. The authors concluded that these results of the meta-analysis suggested that formaldehyde causes leukaemia, specifically myeloid leukaemia. However, in the discussion the authors stated that the question of biological plausibility remains and requires further investigation. They proposed potential mechanisms and hypothesized that formaldehyde may act on bone marrow directly or, alternatively, may cause leukaemia by damaging the hematopoietic stem or early progenitor cells that are located in the circulating blood or nasal passages, which then travel to the bone marrow and become leukemic stem cells.
In contrast, other scientists stated that no plausible mechanism for the induction of leukaemia in humans is found (Section Mode of action). According to Golden et al. (2006), there is no evidence that formaldehyde reaches the bone marrow or has toxic effects in the bone marrow and there is no credible evidence that formaldehyde causes leukaemia in experimental animals. In accord, Pyatt et al. (2008) concluded that existing science does not support the proposed hypothetical mode of action as a logical explanation for proposing that formaldehyde is a realistic etiological factor for any lymphohematopoietic malignancy. Mechanistic considerations in carcinogenicity were also evaluated in IARC (2006); a summary is presented in IUCLID Section 7.12. The IARC working group considered several possible mechanisms for the induction of human leukaemia such as clastogenic damage to circulating stem cells. But they were not aware of any good rodent models that simulate the occurrence of acute myeloid leukaemia in humans. Therefore it was not possible to identify a mechanism for this tumour type. In addition to these mechanistic considerations, the epidemiological approach chosen by Zhang et al. (2009a) appears to be difficult to interpret because the authors combined highest exposure categories of very different meaning from independently designed studies. See the methods paper by Berlin et al. (Berlin JA, Longnecker MP, Greenland S. 1993. Meta analysis of epidemiologic dose-response data. Epidemiology 4(3): 218-228) for a critical discussion of such an approach and for possible solutions how to combine results from studies with differently categorized exposure metrics.
The most recent meta-analysis on epidemiological studies investigating the potential link between formaldehyde and cancers was published by Bachand et al (2010). The authors tried to understand differences between already published overviews and identified characteristics of the original investigations that were used to explore biases. They investigated leukaemia (overall) and three subcategories (myeloid, lymphatic/lymphocytic, and other/unspecified). In contrast to the other meta analyses described and discussed before these authors included the new NCI follow-up study by Beane Freeman et al (2009a) and did not rely on the misleading results of Hauptmann et al. (2003). Therefore, this detailed analysis appears to be the most reliable. They reported the following findings:
- For leukaemias, 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 leukaemia subtype, job type, publication period, or region.
Previous meta-analyses showed elevated summary estimates for leukaemia; however, these analyses included results from proportionate mortality studies and did not explore other factors that could influence or confound results. Proportionate mortality studies have clear methodological limitations when used to evaluate causality (Rothman KJ, Greenland S, Lash TL.,2008. Modern epidemiology. 3 edn. Philadelphia: Lippincott Williams & Wilkins, p.97-99).
- By limiting analyses to stronger case-control and cohort study designs, considering the effects of smoking, the meta-analyses provided little support for a causal relationship between formaldehyde exposure and leukaemia.
However, following a meeting on 20-27 October 2009 in Lyon, an IARC working group now stated (Baan et al 2009) that there is sufficient evidence in humans for a causal association of formaldehyde exposure and myeloid leukaemia. They argued that the epidemiological evidence has become stronger: a recent study (Hauptmann et al. 2009) found that embalming was significantly associated with an increased risk for myeloid leukaemia, with significant trends for cumulative years of embalming and for increasing peak formaldehyde exposure. Moreover,Zhang et al (2010) observed haematological changes in peripheral blood and chromosomal aberrations in formaldehyde exposed Chinese workers. Whereas the study by Zhang et al. (2010) is discussed in the endpoint summary on genetic toxicity Hauptmann et al (2009) will be described and critically discussed in the following.
Hauptmann et al (2009) reported on a NCI case-control study based on three registries listing professionals employed in the funeral industry. Vital statistics offices documented 6808 deaths between 1960 and 1986 for the subjects enumerated on these three lists. In addition to vital status cause of death was reported to NCI. The case-control study was based on 168 lymphohematopoetic deaths (incl 34 myeloid leukaemias), 48 deaths from brain tumours and 265 deceased matched control subjects among these 6808 deaths. The controls were chosen in such a way that their death was attributed to other causes of death than those of interest while excluding respiratory and nervous system as causes. Furthermore, controls were matched to cases by data source, sex, date of birth and date of death. Interviews with next of kin and co-workers were performed between 1990 and1992 to collect basic exposure information for the deceased cases and controls. In these interviews data on work practices, tobacco use and characteristics of the study subjects were gathered. Quantitative exposure assessment was based on a model developed at the Cincinnati College of Mortuary Sciences that linked rate of embalming, effect of ventilation, and concentration of formaldehyde solution to formaldehyde concentration in the air. When the qualitative data from the interviews with next of kin and former co-workers were transformed into quantitative input data with the aim to apply the model it became obvious that necessary details in about 50% of the subjects were missing: there were no data available to calculate peak, average or cumulative exposures to formaldehyde for about half of the study group. The authors used unconditional logistic regression adjusting for matching variables and smoking to analyze the data. The reported results focused on myeloid leukaemia and embalming:
- Ever vs. never embalming was associated with a clearly elevated odds ratio of 11.2 although the significantly elevated estimate was rather imprecise (95%-CI: 1.3 - 95.6)
- Positive trends were observed with duration and number of embalmings
- Positive trends were not found for cumulative exposure but were found withaverage and peak formaldehyde exposure metrics.
No noteworthy associations were observed with other lymphohematopoietic malignancies or brain cancer. The estimated effects of duration of embalming and of peak formaldehyde exposure on myeloid leukaemia mortality were emphasized by the IARC working group as important findings (Baan et al 2009).
However, the design of the investigation suffered from obvious weaknesses: no incidence density sampling of controls was performed and only deceased controls were enrolled for study (see for a discussion the textbook by Rothman et al. 2008; reference see above). Moreover, mean average and peak exposures were reported to be identical among myeloid leukaemias and controls according to the description given in Table 2 in Hauptmann et al (2009), just duration and number of embalmings and cumulative exposures were described as higher among cases. Thus, it is unclear how robust the reported findings of an association between peak formaldehyde exposures and myeloid leukaemia are. Next, the race distribution was clearly different between myeloid leukaemias and controls (see Table 1 of Hauptmann et al. 2009) but this difference was ignored in the analyses and the discussions although race was always adjusted for in the NCI cohort analyses (Hauptmann et al. 2003, Beane Freeman et al. 2009a). Furthermore, the calendar year of first employment in the funeral industry clearly differed between myeloid leukaemia cases and controls but this was not adjusted for in the analyses (cases began work more often before WWII when exposures may have been qualitatively different than later). Hauptmann et al. (2009) performed sensitivity analyses to estimate the effect of the large amount of missing data in exposure assessment on the study findings: all risk estimates were attenuated when missing data were taken into account. Moreover, trends were no longer significant or positive ininternal analyses - that is the model type emphasized by Hauptmann et al (2003) and Hauptmann et al (2004). In addition, the reported trend in formaldehyde peak exposure was based on only one case among the controls. Therefore, the authors performed an additional analysis using alarger referent group. Again, there was no trend found in internal analyses.Therefore, this case-control study of Hauptmann et al. (2009) is clearly limited and should not be understood as proof for a causal association of formaldehyde exposure and myeloid leukaemia mortality.
In summary, the extensions and re-evaluations of the NCI cohort study, the results of a large British study and clear limitations of the NCI embalmer case-control study shed doubt on the decision of an IARC’s working group that there is a causal association between formaldehyde exposure and leukaemia mortality, in particular myeloid leukaemia mortality (Baan et al 2009). Results from a recent meta-analysis (Bachand et al 2010) support this point of view.
The strength of the epidemiological leukaemia data is not sufficient for a classification of formaldehyde as a human carcinogen within the EU regulatory framework taking into account that there is no consistency between the different epidemiological studies:
- the statistically significant trends in the NCI cohort study as originally reported by Hauptmann et al. (2003) were misleading and were withdrawn by Beane Freeman et al (2009 a/b)
- the NCI embalmer case-control study is clearly limited
- an association between formaldehyde and leukaemia generally was not found for industrial workers within the meta-analyses, but some indication of a relationship for professionals (e.g. embalmers, pathologists). However, for these professionals confounding factors cannot be excluded. And these findings could not be repeated when focusing on the stronger designs of case-control and cohort studies.
- there is no mechanism for induction of leukaemia that is broadly accepted by the scientific community: the mechanism proposed by Zhang et al. (2010; see IUCLID Section 7.10.2 and mode of action in IUCLID Section 7.12) should only be considered as being hypothetical.
- in carcinogenicity bioassays with inhalation exposure leukaemia was not observed in any species
A literature search after the last IUCLID update was carried out up to April 20, 2015.
In the follow-up on the US NIOSH garment industry cohort described abovewith 11,043 workers (Meyers et al 2013, key) the SMRs were similar to those of the US population for all cancers as well as for lympho-haematopoietic cancers (leukaemias, Hodgkin disease, non-Hodgkin lymphoma, and multiple myeloma). Stratifying SMRs for “year of first exposure” (< 1963, 1963–1970, ≥ 1971) did not show significant associations for lympho-haematopoietic cancers, Associations between “duration of FA exposures” (< 3, 3–9, ≥ 10 years) and risks of cancer were studied with SMRs and SRRs. There was no exposure-dependent increase in risks for lympho-haematopoietic cancers and non-Hodgkin lymphoma. The risks increased with the length of the exposures for leukaemia, myeloid leukaemia and acute myeloid leukaemia, but were not statistically significant. For multiple myeloma, the SMRs for the exposure groups were 1.16 (0.50–2.29), 2.03 (1.01–3.64) and 0.64 (0.17–1.64), respectively, and the SRRs were 1.00 (reference), 1.22 (0.46–3.26) and 0.28 (0.08–099), respectively. Nevertheless, among persons with ≥ 10 years of exposure and ≥ 20 years since first exposures, the risk for leukaemia (23 deaths, SMR: 1.74 (1.10–2.60)) was significantly increased when multiple causes of death were considered. Additionally, the association between duration of exposure and leukaemia (36 cases) and myeloid leukaemia (21 cases) was studied using four multivariate Poisson regression models (adjusted for age, year of birth and years since first exposure), where exposures were either untransformed or transformed (log, square root, and categorical (< 1.6 (reference), 1.6– < 6.5, 6.5– < 16, 16– < 19 and ≥ 19 years)). Only the untransformed model for leukaemia and the categorical model for myeloid leukaemia were statistically significant. Nevertheless, for leukaemia and myeloid leukaemia, the rate ratio was significantly increased in the fourth category (4.56 (1.30–16.2) and 6.42 (1.40–32.2), respectively), but not for the other (2th, 3th and 5th) categories. In total,the myeloid leukemia mortality was elevated but not the overall leukemia mortality compared to the US population. When analyzing SMRs by year of first exposure, exposure duration, and time since first exposure, the overall leukemia mortality showed a significant association with increasing exposure duration. The authors concluded that that this update continued to show limited evidence for leukemia, but the extended follow-up did not strengthen the formerly observe associations.
In the follow-up of the British (UK) cohort study described above in the period 1941–2012 (Coggon et al 2014, key), no significant increase was seen for the different hematopoietic malignancies. The cohort was stratified for levels of exposure, where the exposure was >2 ppm in the high exposure group. No increase was seen for non-Hodgkin lymphoma [0.90 (0.48–1.55)], multiple myeloma [1.18 (0.57–2.18)], leukaemia [0.82 (0.44–1.41)] and myeloid leukaemia 0.93 [0.40–1.82)]. Exposure in the high exposure group was further stratified for duration of exposure (<1 year, 1–14 years and ≥ 15 years) and the SMRs were largely independent of the length of the exposure period. Additionally, the authors included a nested case-control analysis for all leukaemia and myeloid leukaemia. ORs for these cancers were independent of the duration of the exposure.
Also, in the Italian cohort study described above (Pira et al. 2014, supporting) the incidences for lymphoma, myeloma, leukaemia, and for all lympho-haematopoietic neoplasms were not increased. The study has a long follow-up period, but a limitation is the lack of quantitative FA exposures. Several studies analysed the epidemiological evidence for lymphohematopoetic cancer including leukemia. A meta-analysis of Schwilk al. (2010, supporting) focussed on high-exposure groups and myeloid leukaemia. The analysis included two large studies in particular: one involving > 25 000 workers in US FA industries and the other involving a cohort of > 13 000 funeral directors and embalmers. FA was found associated with increased risks of leukaemia (RR = 1.53; 95 % CI = 1.11–2.21; p = 0.005; 14 studies), specifically myeloid leukaemia (RR = 2.47; 95 % CI = 1.42–4.27; p = 0.001; 4 studies). This study was interpreted by the authors to provide evidence of an increased myeloid leukaemia risk with high exposures to FA. The analysis has been considered to suffer from methodological shortcomings similar to the analysis of Zhang et al. (2009a) because it did not use all available information. The chosen highest exposure cut points varied across the combined studies, which introduced heterogeneity; the homogeneity tests used in the study were considered insensitive. Predictive intervals are recommended instead of confidence intervals and the findings of elevated leukaemia and myeloid leukaemia risks were far from significant if using these techniques in the data analyses (Morfeld 2013).
Cole et al. (2010, supporting) analysed the epidemiological findings and the statistical methods used in the cohort study of Beane-Freeman et al. (2009 a/b) on industrial workers and in the case control study of Hauptmann et al. (2009) on embalmers. These 2 studies were pivotal for the Group 1 classification of IARC for leukemia. The authors concluded that neither study substantiated a significant excess of mortality from any lymphohematopoetic cancer although both were interpreted as positive for an association between formaldehyde and myeloid leukemia. In industrial workers there was only a weak and transitory relationship with peak exposure and for embalmers the limited exposure response relationship was not analysed for statistical significance. It is concluded that these 2 studies do not provide clear evidence between formaldehyde exposure and myeloid leukemia.
Rhomberg et al. (2011, supporting) used a hypothesis-based weight-of-evidence approach to assess the overall evidence for the formation of leukemia in humans after inhalation exposure to formaldehyde.They used the database from studies in humans, animals and mechanistic approaches and concluded that the case for a causal association is weak and strains biological plausibility. To their interpretation the association between formaldehyde exposure and leukemia in some human studies is rather due to chance or confounding.
Checkoway et al (2012, supporting) reviewed and summarised the total published epidemiological literature during 1966–2012. The literature was categorised according to study design and population: industrial cohort studies, professional cohort studies and population-based case-control studies. It was found that findings from occupational cohort and population-based case-control studies were very inconsistent for lympho-haematopoietic malignancies, including myeloid leukaemia. Apart from some isolated exceptions, relative risks were close to one, and there was little evidence for dose-response relations for any of the lympho-haematopoietic malignancies. It was concluded that at present, there is no consistent or strong epidemiologic evidence that FA is causally related to any lympho-haematopoietic malignancy. The absence of established toxicological mechanisms was found to further weaken the arguments for causation.
Checkoway et al. (2015, supporting) evaluated the association between cumulative and peak formaldehyde exposure and mortality from acute myeloid leukemia and other lymphohematopoetic malignancies observed in the NCI cohort (Beane-Freeman et al., 2009 a/b). When using the Cox proportional hazards analyses, acute myeloid leukemia was unrelated to cumulative exposure while for chronic myeloid leukemia a suggestive association with peak exposure was observed, albeit based on very small numbers. For Hodgkin lymphoma relative risk estimates were increased in the highest cumulative and peak exposure categories. No other lymphohematopoetic malignancy was related to chronic or peak exposure. It was concluded that this reanalysis does not support formaldehyde as a cause for acute myeloid leukemia. As no prior epidemiological evidence exists for Hodgkin leukemia and chronic myeloid leukemia, any causal interpretation for these malignancies was considered at most tentative.
Mundt et al. (2018) reviewed the epidemiological evidence published after the Checkoway et al. (2012) review and identified several epidemiological studies including the NIOSH garment workers study (Meyers et al. 2013), the UK industrywide formaldehyde producers and users study (Coggon et al. 2014), a case-control study in the Nordic countries (Talibov et al. 2014), an occupational study in Italy (Pira et al. 2014) and a European occupational study (Saberi Hosnijeh et al. 2013). In addition, an extended analysis of the National Cancer Institute’s (NCI) cohort study (Beane-Freeman et al. 2009) was conducted by Checkoway et al. (2015) which included data from the NCI formaldehyde industrial workers cohort to further investigate specific types of LHM, especially AML, and a more conventional definition of peak exposure. Following a review of the results of the available epidemiological data, Mundt et al. (2018) concluded that there is no evidence of significant increases in LHM, specifically AML, among cohorts of workers exposed to formaldehyde.
1) A statistically increased risk for nasopharyngeal cancer in workers exposed to formaldehyde has only been detected at one out of ten plants in one cohort study but not in others. The validity and robustness of this specific observation is unclear as it suffers from small numbers of cases and an incomplete mortality follow-up.
2) A non-significant correlation between formaldehyde exposure and leukaemia, especially myeloid leukaemia was seen in some studies but not all. This tumour type is considered to be biologically not plausible taking into account all information from animal and human data in a weight of evidence approach.
A summary and evaluation of data on carcinogenicity available up to 2009 is reported in reviews of the IARC working group (IARC, 2006; 2012; Baan et al., 2009); details on epidemiological data are presented in Section 7.10.2 and carcinogenicity in experimental animals in Section 7.7 of the technical dossier.
Information on Registered Substances comes from registration dossiers which have been assigned a registration number. The assignment of a registration number does however not guarantee that the information in the dossier is correct or that the dossier is compliant with Regulation (EC) No 1907/2006 (the REACH Regulation). This information has not been reviewed or verified by the Agency or any other authority. The content is subject to change without prior notice.Reproduction or further distribution of this information may be subject to copyright protection. Use of the information without obtaining the permission from the owner(s) of the respective information might violate the rights of the owner.
Deze website maakt gebruik van cookies om het surfen zo aangenaam mogelijk te maken.
Welcome to the ECHA website. This site is not fully supported in Internet Explorer 7 (and earlier versions). Please upgrade your Internet Explorer to a newer version.
Do not show this message again