You are on page 1of 9

IJE vol.33 no.4 International Epidemiological Association 2004; all rights reserved.

International Journal of Epidemiology 2004;33:894902


Advance Access publication 20 May 2004 doi:10.1093/ije/dyh164
Health effects of dental amalgam exposure:
a retrospective cohort study
Michael N Bates,
1,2
Jackie Fawcett,
1,3
Nick Garrett,
1,4
Terry Cutress
3
and Tord Kjellstrom
5,6
Accepted 11 February 2004
Background Whether dental amalgam fillings (containing mercury) are hazardous is a long-
standing issue, with few epidemiological investigations. Allegations have particularly
involved nervous system disorders, such as multiple sclerosis, Alzheimers disease,
and chronic fatigue syndrome. This retrospective cohort study, the largest of its
kind, contained people in the New Zealand Defence Force (NZDF) between 1977
and 1997. The NZDF has its own dental service, providing all personnel with regular
and consistent treatment. Comprehensive treatment records are maintained and
archived.
Methods Yearly dental treatment histories, including amalgam filling placements, were
compiled from individual records. To minimize amalgam exposure misclassifi-
cation the cohort was restricted to people who, at NZDF entry, were aged 26 years
and had all their posterior teeth. The cohort was linked with morbidity records.
Data were analysed with a proportional hazards model, using a time-varying
exposure unit of 100 amalgam surface-years.
Results The final cohort contained 20 000 people, 84% males. Associations with medical
diagnostic categories, particularly disorders of the nervous system and kidney,
were examined. Of conditions allegedly associated with amalgam, multiple sclerosis
had an adjusted hazard ratio (HR) of 1.24 (95% CI: 0.99, 1.53, P = 0.06), but
there was no association with chronic fatigue syndrome (HR = 0.98, 95% CI: 0.94,
1.03), or kidney diseases. There were insufficient cases for investigation of
Alzheimers or Parkinsons diseases.
Conclusions Results were generally reassuring, and provide only limited evidence of an
association between amalgam and disease. Further follow-up of the cohort will
permit investigation of diseases more common in the elderly.
Keywords Chronic fatigue syndrome, cohort studies, dental amalgam, kidney diseases,
mercury, multiple sclerosis, New Zealand
Mercury, combined with other metals to form solid amalgams,
has long been used in reconstructive dentistry. Dental amalgam
currently contains about 50% mercury, with the remainder mainly
silver. Although alternative dental materials are increasingly
available for posterior fillings, amalgam has advantages that main-
tain its popularity as a filling material. These include relatively
low cost, durability, and less sensitivity to clinical technique than
other materials.
The use of mercury in dentistry has been controversial since
at least the middle of the 19th century, as inorganic mercury can
cause a wide variety of health effects, particularly to the neuro-
logical and renal systems.
1,2
This controversy has intensified over
894
1
Institute of Environmental Science and Research Ltd. (ESR), PO Box
50348, Porirua, New Zealand.
2
Current affiliation: School of Public Health, 140 Warren Hall, University of
California, Berkeley, CA 947207360, USA.
3
Current affiliation: Wellington School of Medicine, PO Box 7343,
Wellington, New Zealand.
4
Current affiliation: Faculty of Health, Auckland University of Technology,
Private Bag 92006, Auckland 1020, New Zealand.
5
Department of Community Health, University of Auckland, Private Bag
92019, Auckland, New Zealand.
6
Current affiliation: National Centre for Epidemiology and Population
Health, The Australian National University, Canberra, ACT, 0200 Australia.
Correspondence: Dr Michael Bates, 140 Warren Hall, University of California,
Berkeley, CA 947207360, USA. E-mail: m_bates@uclink.berkeley.edu

b
y

g
u
e
s
t

o
n

M
a
r
c
h

2
4
,

2
0
1
3
h
t
t
p
:
/
/
i
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

HEALTH EFFECTS OF DENTAL AMALGAM EXPOSURE 895
the last 20 years or so, because highly sensitive analytical tech-
niques have shown mercury to be continuously released from
dental amalgam fillings and absorbed into the body.
35
Reviews
of the evidence have consistently stated that there is a deficiency
of adequate epidemiological studies addressing this issue.
68
The New Zealand Defence Force (NZDF) is the only large
organization in New Zealand that has its own dental service and
maintains consistent and comprehensive adult dental records,
which are centrally stored and archived. Free dental care has
been provided to all members of the regular New Zealand military
forces since 1918. Treatment is mandatory, regular, and consistent
across ranks. This study is a retrospective cohort study of NZDF
personnel, with an internal comparison. The cohort study
design was selected because there is no particular disease which
has been clearly identified as associated with dental amalgam
use. However, various claims have been made, particularly
about associations with autoimmune and neurological diseases,
such as multiple sclerosis (MS),
9
Alzheimers disease,
10
Parkinsons disease,
11
and chronic fatigue syndrome (CFS).
12
This paper presents overall results, with additional focus on
diseases of the nervous system and kidney, not including cancer.
Methods
The study design was approved by the Wellington Ethics
Committee.
Identification and definition of the cohort
Potentially, the cohort comprised all members of the regular
NZDF (army, navy, and air force) meeting the following criteria:
(1) a member of the regular arm of the NZDF for at least one
day between 1 January 1977 and 31 December 1997. These
dates were selected for the availability of health outcome data;
(2) NZDF dental records for the period of service were available.
Individuals with incomplete records were excluded.
Confirmation of vital status
For cohort members not recorded by the New Zealand Health
Information Service (NZHIS) as deceased, a variety of methods
were used to confirm that they were alive at, or as close as possible
to, the final date of follow-up (31 December 1997). These were
similar to methods previously described
13
and included NZDF
service dates, records of the New Zealand Drivers License Register,
dates of last contact with the New Zealand public hospital system,
pension and retirement fund records, and electronic voter registra-
tion lists. Person-time contributions to the cohort terminated at
death, end of the follow-up period, or the date closest to end of
follow-up at which they could be confirmed as alive and living
in New Zealand.
Outcome data
Each person in New Zealand is allocated a unique National Health
Index (NHI) number on first contact with the national health
system. All public hospital records (hospital discharge records)
and deaths are linked to NHI numbers. All mortality records since
1988 are identifiable through the associated NHI numbers. Earlier
mortality linkage is by name and birth date. Hospital discharge
information was not universally linked to an NHI number until
the late 1980s.
Once an NHI number for a cohort member was obtained, linked
hospital discharge and mortality data, by dates and disease codes,
were obtained up to 1997. When there were duplicate hospital
admissions for the same condition, only the first admission date
was used in the analysis.
Amalgam exposure data and exposure index
Amalgam exposure data were obtained from NZDF Dental Service
records. Treatments are recorded by tooth surface on dental record
cards. Bite-wing X-rays are taken on entry and periodically
thereafter.
Dental record charts for retirees from the NZDF before 1992
were stored in the NZDF personnel archives. Navy records for
personnel retiring up to about 1989 were stored elsewhere and
were unavailable. For the purposes of this study, the dental
records, including X-rays, of all other NZDF personnel dis-
charged during 1977 to 1991 were systematically examined by
dental therapists who were part of the study team. A data entry
template was used for consistent recording of amalgam place-
ments by tooth, tooth surface, and year, for each year of military
service. For each person for whom dental records were available,
all amalgam placements and tooth extractions, from year of
entering the NZDF until 1991, were recorded.
Dental treatment data for people in service 19921997 were
available as an electronic database. This database was combined
with the database created from the archive-based records to
provide the final amalgam exposure database.
For each subject for whom we had dental records, a dental
amalgam exposure index was created based on the cumulative
weighted sum of times since placement of fillings in tooth surfaces.
Because they are larger, amalgam-filled, pre-molar and molar
occlusal (top) tooth surfaces were weighted more highly than
other surfaces. Relative surface area weightings were based on
the evaluation of Saxe et al.
10
The highest weighting was 2.9
for amalgam-filled occlusal surfaces on molar teeth with at least
two other amalgam-filled surfaces.
Exposure index assumptions were: (1) amalgam fillings existing
on entry into the NZDF had been in place since age 15. The aver-
age period from age 15 to NZDF entry was 4.3 years; (2) amalgam
filling burden at NZDF discharge remained constant until death
or the end of the follow-up period; (3) contribution of an amalgam
filling to exposure ceased if the amalgam was replaced with
a filling of other material or when the tooth with the amalgam
was extracted.
Covariate data
Available covariate data were date of birth, sex, calendar period
of NZDF service, and military rank at either exit from the
NZDF or end of follow-up. Since the three different services
have corresponding ranks, a combined variable for equivalent
rank was used.
Statistical analysis
The primary method of analysis was the Cox proportional
hazards model. The selected unit of amalgam exposure was
100 amalgam-filled surface-years. This arbitrary unit is equivalent
to having 10 non-occlusal amalgam-filled surfaces for 10 years.
This was entered into models as a time-varying cumulative
exposure index. The model was stratified by sex, year of birth
(five calendar-year blocks), and equivalent rank (four levels).

b
y

g
u
e
s
t

o
n

M
a
r
c
h

2
4
,

2
0
1
3
h
t
t
p
:
/
/
i
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

896 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
All models were tested for compliance with the proportionality
assumption.
14
The hazard ratios (HR) represent the risk associated with one
unit of amalgam exposure, relative to the risk associated with
having no amalgam exposure.
Results
Of the 40 366 individuals who potentially could have been in
the cohort because of their period of service, 1716 had missing
electronic dental records and 8970 either did not have available
paper-based dental records or they were considered to be
insufficiently complete. Individuals with missing and incomplete
records appeared to be randomly distributed across the potential
cohort. After excluding individuals with missing and incomplete
records, we were left with 29 680 people with complete dental
records ( the initial cohort). However, particularly prior to 1970,
a substantial proportion of the cohort had missing teeth, other
than wisdom (third molar) teeth, at entry into the NZDF. Because
we had no information on prior amalgam treatments for missing
teeth, we minimized amalgam exposure misclassification by limit-
ing the main analysis to people aged 25 years at entry to the
NZDF and with no missing posterior teeth, except wisdom teeth.
This resulting final cohort contained exactly 20 000 people.
Since entry to the NZDF may have been before the beginning of
the follow-up period in 1977, some cohort members would
have been aged 25 at the time of their entry into the cohort.
Table 1 shows the outcome of the follow-up for the initial and
the final cohorts.
Of the theoretical total person-years of follow-up, 92.5% was
traced for the initial cohort, and 93.0% for the final cohort. The
theoretical total person-years of follow-up assumes that all cohort
members lost to follow-up were alive at the end of follow-up.
Table 2 shows demographic details for the initial and final
cohorts. The final cohort tended to be younger and began
service with the NZDF on average later than the initial cohort.
However, the average length of follow-up was comparable for
the initial and final cohorts (12.5 and 11.2 years, respectively).
Table 3 shows the weighted mean annual number of amalgam
fillings across approximate quartiles of the follow-up period,
and numbers of person-years of exposure for each of the age
calendar period categories. Generally, the number of fillings
increases with age, then declines in the older groups, probably
because of extractions. There is a progressive decline in person-
years contributed after age 25 years. The average cumulative
exposure summed over the years of follow-up for the final
cohort was 628 amalgam-filled surface-years.
The morbidity data analysis was undertaken on any occurrence
of an International Classification of Diseases, Ninth Revision
(ICD) code, or group of codes, within up to 15 codes that could
be recorded for every hospital discharge, after eliminating any
readmission for the same condition (ICD code).
Table 4 shows HR for broad disease categories, based on target
organs or types of illness, in the final cohort. Most HR were around
1.00. The only category that had an HR that was statistically
significant (i.e. P 0.05) was endocrine, nutritional, and meta-
bolic diseases, and immunity disorders (ICD-9: 240279).
However, this was in the direction of reduced risk. We carried
out a similar across-the-board analysis of mortality in the final
cohort. There were 189 deaths (HR = 1.01, 95% CI: 0.97, 1.05),
of which 131 were from injuries and poisonings (ICD-9:
800999) and 28 from neoplastic diseases (ICD-9: 140239). No
causes of death showed evidence of an association with amalgam
exposure, although numbers were small.
We then examined individual three-digit ICD-9 codes for
hospital admissions for the two main target organs of inorganic
mercury toxicitythe nervous system and the kidney. Table 5
shows results for psychiatric and neurological disorders. Two
models that did not satisfy the proportionality assumption were
excluded. These were acute reaction to stress (ICD-9: 308,
13 cases) and conduct disturbance not elsewhere identified
(ICD-9: 312, 7 cases). Four conditions had CI excluding the null
value: other paralytic syndromes (ICD-9: 344) and mononeuritis
of the upper limb and mononeuritis multiplex (ICD-9: 354) had
elevated HR; adjustment reaction (ICD-9: 309) and inflammatory
and toxic neuropathy (ICD-9: 357) had reduced HR.
Of the conditions that have been alleged to be associated with
dental amalgam, MS (ICD-9: 340) had an elevated relative risk
estimate (HR = 1.24, 95% CI: 0.99, 1.53, P = 0.06), but there
was no evidence of any association between amalgam expo-
sure and CFS (ICD-9: 780) (HR = 0.98, 95% CI: 0.94, 1.03).
There were no cases of Parkinsons disease (ICD-9: 332) in the
cohort, one of motor neurone disease (ICD-9: 335.2), and two
of Alzheimers disease (ICD-9: 331). These were considered
insufficient for useful calculation of HR.
There were no elevated HR for kidney diseases (Table 6). There
were HR less than unity for several of the conditions, parti-
cularly nephritis not otherwise specified (ICD-9: 583) and chronic
renal failure (ICD-9: 585).
In the models shown in Tables 46, tooth surfaces were
weighted to reflect the greater surface areas of occlusal fillings,
and therefore their greater potential for mercury release. We
investigated the effect of eliminating the differential weighting
so that all tooth surfaces were considered to contribute equally to
exposure. The results (not shown) were consistent. Removing
the weighting in almost all cases resulted in HR that were further
away from the null value than the corresponding weighted value,
although usually only by a few per cent.
Discussion
The debate about possible health effects induced by dental
amalgam is marked by an absence of adequate epidemiological
studies. Previous studies have mostly been small, had limited
exposure data, been particularly subject to selection bias in terms
Table 1 Outcome of the follow-up process, 19771997, for the initial
and final cohorts, New Zealand Defence Force dental amalgam study
No. of subjects (column %)
Outcome Initial cohort
a
Final cohort
b
Deceased 508 (1.7%) 189 (0.9%)
Lost to follow-up 2235 (7.5%) 1400 (7.0%)
Alive at end of follow-up 26 937 (90.8%) 18 411 (92.1%)
Total 29 680 (100%) 20 000 (100%)
a
With complete dental records available and before application of restriction
criteria for age and tooth retention.
b
In initial cohort and aged 25 at entry to NZDF, with no missing posterior
teeth (except wisdom teeth).

b
y

g
u
e
s
t

o
n

M
a
r
c
h

2
4
,

2
0
1
3
h
t
t
p
:
/
/
i
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

HEALTH EFFECTS OF DENTAL AMALGAM EXPOSURE 897
of dental treatment access, and focused on a narrow range of
outcomes. By contrast, our study had a large sample size, detailed
exposure data, a consistent level of dental treatment across the
cohort, and investigation of a wide range of possible health
outcomes. The success of the follow-up process, accounting for
93% of the theoretically possible maximum person-time,
suggests that selection bias from loss to follow-up is not likely to
be a problem. Also, as noted in the Results, there were 8920
incomplete and 1716 missing records for people potentially in
the cohort. We think it highly unlikely that the distribution of
Table 2 Demographic characteristics of the initial and final cohorts, New Zealand Defence Force dental amalgam study
Initial cohort
a
Final cohort
b
Characteristic No. (%) No. (%) % of initial cohort
Service
Air Force 7884 (26.6%) 5264 (26.3%) 67
Army 18 772 (63.2%) 12 348 (61.7%) 66
Navy 3024 (10.2%) 2388 (11.9%) 79
Years of entry into cohort
19771981 12 781 (43.1%) 6541 (32.7%) 51
19821986 4410 (14.9%) 3331 (16.7%) 76
19871991 4742 (16.0%) 3850 (19.3%) 81
19921997 7747 (26.1%) 6278 (31.4%) 81
Gender
Male 25 178 (84.8%) 16 752 (83.8%) 67
Female 4502 (15.2%) 3248 (16.2%) 72
Age at entry to the cohort
1625 21 685 (73.1%) 17 075 (85.4%) 79
2635 5221 (17.6%) 2540 (12.7%) 49
3645 2067 (7.0%) 355 (1.8%) 17
4655 666 (2.2%) 30 (0.2%) 4.5
5665 39 (0.1%) 0 0
65 2 (0.01%) 0 0
Age at end of follow-up
1625 6020 (20.3%) 5189 (25.9%) 86
2635 10 963 (36.9%) 8690 (43.5%) 79
3645 7880 (26.5%) 5064 (25.3%) 64
4655 2849 (9.6%) 884 (4.4%) 31
5665 1472 (5.0%) 155 (0.8%) 11
6675 469 (1.6%) 18 (0.1%) 3.8
75 27 (0.1%) 0 0
Years of follow-up
15 5693 (19.2%) 4453 (22.35%) 78
610 7396 (24.9%) 5828 (29.1%) 79
1115 4609 (15.5%) 3447 (17.2%) 75
1621 11 982 (40.4%) 6272 (31.4%) 52
Total 29 680 20 000 67
Mean years of follow-up 12.5 11.2
Total person-years of follow-up 369 736 223 698 61
Males 323 339 192 280 59
Females 46 397 31 418 68
Total weighted amalgam-filled surface-years 22 442.7 13 687.3 61
of exposure (1,000s)
a
With complete dental records available and before application of restriction criteria for age and tooth retention.
b
In initial cohort and aged 25 at entry to New Zealand Defence Force, with no missing posterior teeth (except wisdom teeth).

b
y

g
u
e
s
t

o
n

M
a
r
c
h

2
4
,

2
0
1
3
h
t
t
p
:
/
/
i
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

898 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
these records was in any way associated with morbidity or
mortality outcomes.
A strength of the present study was the detailed, tooth
surface-specific amalgam exposure data, by year of treatment,
for the entire cohort across the period of military service. It is
particularly unusual to have longitudinal amalgam placement
data, and unique to have it in a cohort of this size. A recent
cross-sectional study of neuropsychological function in relation
to amalgam exposure highlighted this as a particular limitation.
15
The only other published study with exposure data of comparable
quality to our own was a small study by Saxe et al.
10
Another strength was consistency of dental treatment across
the cohort. All NZDF personnel have received compulsory and
equivalent treatment, irrespective of rank. However, among
civilians, dental treatment is not equally accessible. People of
higher socioeconomic status (SES) are more likely to obtain
dental treatment, and hence to have more amalgam fillings.
Since higher SES is associated with better health status, there is
Table 4 Hazard ratios (HR) for broad disease categories in the final cohort, New Zealand Defence Force amalgam study
Disease category (ICD-9
a
codes) No. of cases HR
b
95% CI P-value
Infectious and parasitic diseases (001139) 439 1.00 0.97, 1.04 0.83
Neoplastic diseases (140239) 264 1.00 0.97, 1.03 0.83
Endocrine, nutritional, and metabolic diseases, and 126 0.96 0.92, 0.99 0.02
immunity disorders (240279)
Diseases of the blood (280289) 82 0.97 0.91, 1.04 0.37
Psychiatric disorders (290319) 572 0.98 0.95, 1.02 0.31
Nervous system diseases (320389) 341 1.02 0.98, 1.05 0.34
Circulatory system diseases (390459) 359 1.02 0.99, 1.05 0.14
Respiratory system diseases (460519) 659 0.99 0.97, 1.02 0.56
Digestive system diseases (520579) 1142 1.00 0.98, 1.02 0.73
Genitourinary system diseases (580629) 599 1.00 0.97, 1.03 0.98
Complications of pregnancy and childbirth (630677) 1062 0.99 0.97, 1.02 0.66
Skin and subcutaneous tissue diseases (680709) 331 1.02 0.98, 1.06 0.24
Musculoskeletal and connective tissue diseases (710739) 696 1.00 0.98, 1.03 0.95
Symptoms, signs, and ill-defined conditions (780799). 859 1.01 0.99, 1.03 0.59
Injuries and poisonings (800999) 2628 1.01 1.00, 1.03 0.10
a
International Classification of Diseases, Ninth Revision.
b
Hazard ratio for 100 amalgam surface-years relative to no amalgam exposure. All HR stratified by sex, year of birth (5-calendar year blocks), and equivalent
rank (four levels) at either discharge, death, or the end of follow-up.
Table 3 Weighted-average annual numbers of tooth surfaces with amalgam and corresponding numbers of person-years in the final cohort
during the period of follow-up, 19771997, New Zealand Defence Force dental amalgam study
Years of follow-up of the cohort
Age-
19771981 19821986 19871991 19921997 All years
groups Amalgam Person- Amalgam Person- Amalgam Person- Amalgam Person- Amalgam Person-
(years) surfaces years surfaces years surfaces years surfaces years surfaces years
1620 41.0 7945 28.4 6656 18.6 8356 14.9 6530 26.0 29 487
2125 50.9 10 894 40.7 14 133 27.4 15 653 19.7 23 034 31.6 63 714
2630 59.1 4243 52.0 10 888 41.2 14 503 27.5 24 920 38.5 54 554
3135 62.5 1922 60.0 4144 52.5 10 749 40.8 21 580 47.3 38 395
3640 62.5 908 63.4 1934 60.5 4109 51.6 15 373 54.7 22 324
4145 61.7 298 62.9 913 63.7 1881 60.1 5834 61.2 8926
4650 54.7 142 62.8 295 63.0 898 63.8 2513 63.2 3848
5155 48.2 28 55.0 137 63.0 287 62.8 1172 61.9 1624
5660 39.7 5 48.3 28 54.8 134 62.6 398 59.8 565
6165 39.7 5 48.3 28 56.7 174 55.1 207
6670 39.7 5 48.5 40 47.5 45
7175 43.9 9 43.9 9
Total 50.6 26 385 45.7 39 133 38.8 56 603 34.8 101 577 39.6 223 698

b
y

g
u
e
s
t

o
n

M
a
r
c
h

2
4
,

2
0
1
3
h
t
t
p
:
/
/
i
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

HEALTH EFFECTS OF DENTAL AMALGAM EXPOSURE 899
potential for confounding by SES. This may be why the only other
cohort study to examine this issue, found apparent protective
associations with amalgam across a wide range of outcomes.
16,17
As a measure of SES we had data on military rank at end of
follow-up, death, or discharge from the NZDF. We investigated
changes in amalgam associations with this variable in and out of
the models. Its inclusion had some small effects on some asso-
ciations, but these were always 5% of the HR and not in a
consistent direction.
The possibility remains of differential dental treatment access
in periods before joining and after discharge from the NZDF.
However, by restricting our cohort to subjects 26 years at entry
Table 5 Psychiatric and neurological disorders and symptoms at the three-digit International Classification of Diseases, Ninth Revision (ICD-9)
code level, for which there were at least five cases in the final cohort, New Zealand Defence Force dental amalgam study
Condition (ICD-9 code)
a
No. of cases HR
b
95% CI P-value
Transient organic psychoses (293) 6 1.05 0.81, 1.35 0.73
Schizophrenic disorders (295) 45 1.01 0.87, 1.16 0.91
Affective psychoses (296) 71 1.00 0.92, 1.08 0.93
Paranoid states (297) 6 1.03 0.67, 1.58 0.91
Other non-organic psychoses (298) 18 0.96 0.78, 1.17 0.67
Neurotic disorders (300) 75 0.98 0.90, 1.05 0.52
Personality disorders (301) 38 1.03 0.92, 1.15 0.65
Special symptoms or syndromes not found elsewhere (307) 13 0.68 0.44, 1.03 0.07
Adjustment reaction (309) 58 0.90 0.82, 0.98 0.02
Specific nonpsychotic mental disorders due to organic causes (310) 18 1.10 0.92, 1.32 0.31
Depressive disorder not elsewhere identified (311) 52 0.99 0.90, 1.08 0.75
Meningitis, unspecified cause (322) 5 1.03 0.64, 1.66 0.94
Multiple sclerosis (340) 7 1.24 0.99, 1.53 0.06
Hemiplegia and hemiparesis (342) 10 0.89 0.77, 1.04 0.16
Other paralytic syndromes (344) 14 1.23 1.02, 1.48 0.03
Epilepsy (345) 30 0.94 0.84, 1.06 0.33
Migraine (346) 34 0.97 0.87, 1.08 0.59
Other conditions of the brain (348) 10 1.07 0.83, 1.37 0.60
Other and unspecified neurological disorders (349) 7 1.14 0.90, 1.45 0.29
Facial nerve disorders (351) 6 0.92 0.70, 1.22 0.58
Nerve root and plexus disorders (353) 7 1.03 0.81, 1.32 0.80
Mononeuritis of upper limb and mononeuritis multiplex (354) 41 1.11 1.02, 1.21 0.02
Mononeuritis of lower limb (355) 13 0.92 0.78, 1.08 0.29
Inflammatory and toxic neuropathy (357) 5 0.79 0.63, 0.99 0.04
Chronic fatigue syndrome (780) 132 0.98 0.94, 1.03 0.48
Symptoms involving the nervous and musculoskeletal systems(781) 7 0.99 0.74, 1.31 0.94
a
Excludes conditions involving infectious agents and conditions involving alcohol or drug abuse.
b
HR: Hazard ratio for 100 amalgam surface-years relative to no amalgam exposure. All HR stratified by sex, year of birth (5-calendar year blocks), and
equivalent rank (four levels) at either discharge, death, or the end of follow-up.
Table 6 Kidney disorders at the three-digit International Classification of Diseases, Ninth Revision (ICD-9) code level, for which there were
at least 5 cases in the final cohort, New Zealand Defence Force dental amalgam study
Condition (ICD-9 code)
a
No. of cases HR
b
95% CI P-value
Nephritis not otherwise specified (583) 14 0.82 0.69, 0.97 0.02
Acute renal failure (584) 9 0.91 0.74, 1.12 0.37
Chronic renal failure (585) 6 0.80 0.66, 0.97 0.02
Renal failure unspecified (586) 6 0.83 0.69, 1.01 0.07
Calculus of kidney and ureter (592) 30 0.98 0.89, 1.07 0.65
Other disorders of kidney and ureter (593) 12 0.95 0.83, 1.10 0.51
a
Excludes conditions involving infectious agents.
b
HR: Hazard ratio for 100 amalgam surface-years relative to no amalgam exposure. All HR stratified by sex, year of birth (5-calendar year blocks), and
equivalent rank (four levels) at either discharge, death, or the end of follow-up.

b
y

g
u
e
s
t

o
n

M
a
r
c
h

2
4
,

2
0
1
3
h
t
t
p
:
/
/
i
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

900 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
to the NZDF, we would have substantially reduced that problem,
as most amalgam fillings are put in place before age 26.
Despite this studys strength, it had some potential limitations.
Firstly, characteristic of retrospective cohort studies, is the absence
of data on potential confounding factors, such as smoking and
diet, both of which affect dental caries (and, therefore, amalgam
placement) and other disease rates.
Another possible limitation was the reliance on routinely
collected outcome data. Linkage of hospital discharge records to
NHI numbers was incomplete in the earlier part of our follow-
up period. But to bias the results of our study, linkage to NHI
numbers would have had to have been differential by amalgam
exposure status. We know of no reason why this would have
been likely. Reporting of hospital discharge data by private
hospitals in New Zealand may be less complete. If patients with
higher amalgam exposure are more likely to have been treated
in private hospitals and these data were not reported, then that
would likely lead to underestimation of the relative risk esti-
mates for any diseases truly associated with amalgam exposure.
However, private hospitals account for only a small proportion
of non-surgical admissions in New Zealand.
Some of the cases of conditions of interest in this study may
not have involved hospital admission. For such conditions there
would have been undercounts of the actual numbers of cases
occurring in the cohort. However, again, provided that this under-
counting was not differential by amalgam exposure it should
not bias the results. The main impact would be a loss of statistical
power and wider CI. Also, the cohort is relatively young and,
therefore, has limited statistical power to investigate diseases of
old age (such as Parkinsons diseases and Alzheimers disease)
and causes of death.
Another area of uncertainty is dental treatment in the period
following discharge from the military. A constant amalgam
burden has been assumed from that point, although some
people will have had additional tooth surfaces filled and
some teeth will have been removed. The impact may not be
great, as most people receive the bulk of their amalgam fill-
ings in their younger years. Nonetheless it introduces some
uncertainty into risk estimates, particularly for diseases pre-
dominating at older ages, and the direction of any bias that
might result is unclear.
This study was mainly useful for investigating medical con-
ditions with clearly defined diagnoses. It had limited ability to
investigate symptom complexes not specifically diagnosed. We
examined individual ICD codes within the subcategory of symp-
toms and signs (ICD-9: 780789). There were no individual three-
digit ICD-9 codes associated with amalgam exposure. There is, in
fact, little evidence that ill-defined symptom complexes are asso-
ciated with amalgam fillings. Epidemiological studies have
examined people holding the belief that their symptoms were
associated with amalgam, and have generally found no associa-
tion, or, in some cases, a negative association with amalgam
exposure.
1824
One poorly understood condition, involving symptoms but
no objective medical changes, that we could investigate was
CFS (ICD-9: 780). This condition has been alleged to be
associated with dental amalgam, although there is only limited
scientific literature on this.
12
We had a sufficient number of
cases CFS in our cohort to investigate this hypothesis. Our HR
of 0.98 (95% CI: 0.94, 1.03) provides no evidence of any asso-
ciation between amalgam exposure and CFS (Table 5). A small
number of other cases (n = 20) fell into the category other
ill-defined and unknown causes of morbidity and mortality
(ICD-9: 799) (not shown in Tables), which had an elevated
relative risk estimate (HR = 1.18, 95% CI: 1.03, 1.36, P = 0.02).
There is the hypothetical possibility of diagnostic misclassifica-
tion of CFS cases into this category. However, it seems implausible
that people with CFS caused by amalgam would be differen-
tially allocated to code 799, rather than the code for CFS.
Therefore, we think it unlikely that this affects the validity of
the null association for CFS.
Two neurological conditions with elevated risk estimates were
other paralytic syndromes (ICD-9: 344) and mononeuritis of the
upper limb and mononeuritis multiplex (ICD-9: 354) (Table 5).
We know of no other data bearing on the plausibility of these
results. Therefore, in the absence of supporting evidence, we
regard these results as hypothesis-generating. It is likely they
have arisen as a result of the number of statistical tests that were
carried outthe well-known multiple comparisons issue.
25
This
is also likely to be the case for the apparently protective associations
shown in Tables 5 and 6.
We found no positive association between kidney disorders
and amalgam exposure (Table 6)an important result, as the
kidney is a primary target of inorganic mercury toxicity.
1,2
An interesting result was the suggestion of an association
between amalgam exposure and MS (Table 5). While the number
of cases was small, the HR of 1.24 (95% CI: 0.99, 1.53) for one
unit of amalgam exposure was relatively strong in this study.
A link between MS and dental amalgam was first suggested by
Craelius, who noted a strong correlation in the geographical
distributions of MS and dental caries.
9
Later, Ingalls suggested
that this was confounded by the prevalence of dental amalgam
fillings in high caries areas.
26
MS prevalence increases with
latitude, and recent studies have examined a possible role of
ultraviolet light.
2729
Three MS case-control studies have investigated an
association with amalgam exposure.
3032
However, these
studies had various limitations, such as prevalent cases, limited
exposure measures (including lack of dental treatment records),
and small numbers of subjects. The one study that used incident
cases and actual dental records found higher relative risk
estimates for a larger number of amalgam fillings.
30
The odds
ratio for having more than 15 fillings (relative to having none)
was 2.6 (95% CI: 0.8, 8.5).
For reasons in large part to do with our exclusion criteria and
the high prevalence in NZDF entrants of prior tooth extractions,
the cohort is relatively young. This can be presumed to account
for the small numbers of cases of diseases that are predominant
in the elderly, such as Alzheimers disease and Parkinsons disease,
which have been hypothesized to be caused by amalgam.
10,11
Also, for some outcomes the number of cases occurring in the
cohort was low relative to the number of variables, with the
possibility of some resulting bias in the effect estimates.
33
Despite
these limitations, this study is the most comprehensive so far
to investigate the amalgam safety issue, and generally provides
reassurance. In particular, the study has shown no association
between dental amalgam exposure and either kidney diseases
or CFS. Some important questions remain, however. Therefore,
given the uniqueness of the cohort, it is essential that it be
followed into the future. Continuing follow-up would collect
updated data on dental treatments and health outcomes, would
establish whether there were any associations with rarer diseases

b
y

g
u
e
s
t

o
n

M
a
r
c
h

2
4
,

2
0
1
3
h
t
t
p
:
/
/
i
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

HEALTH EFFECTS OF DENTAL AMALGAM EXPOSURE 901
and diseases more common in the elderly, and confirm whether
associations found in the cohort continue to be present.
Acknowledgements
We gratefully acknowledge provision of accommodation and
access to personnel files and dental treatment data by the
New Zealand Defence Force, with particular thanks to Colonel
Simon Holdgate. For dental data identification and data entry
we thank Tanya Budart, Helena Easter, Eileen Haveler,
Rosemary Healy, Katherine Jordain, Melanie Pendergast, Louise
Te Araki, Shelly Wanden, and Kathryn Weaver. This study
was funded by a grant from the Health Research Council of
New Zealand.
KEY MESSAGES
There is a need for epidemiological studies using longitudinal dental treatment records to investigate whether
amalgam fillings are associated with adverse health events.
In this cohort study there was no evidence of an association between amalgam exposure and adverse kidney effects.
There was no evidence that chronic fatigue syndrome is associated with dental amalgams.
The possibility that multiple sclerosis could be associated with dental amalgams deserves further investigation.
Further follow-up of this cohort will permit investigation of disease outcomes more prevalent in the elderly.
References
1
Clarkson TW. The toxicology of mercury. Crit Rev Clin Lab Sci 1997;
34:369403.
2
Satoh H. Occupational and environmental toxicology of mercury and
its compounds. Ind Health 2000;38:15364.
3
Abraham JE, Svare CW, Frank CW. The effect of dental amalgam
restorations on blood mercury levels. J Dent Res 1984;63:7173.
4
Snapp KR, Boyer DB, Peterson LC, Svare CW. The contributions of
dental amalgam to mercury in blood. J Dent Res 1989;68:78085.
5
Vimy MJ, Lorscheider FL. Dental amalgam mercury daily dose
estimated from intra-oral vapor measurements: a predictor of mercury
accumulations in human tissues. J Trace Elem Exp Med 1990;3:11123.
6
Weiner JA, Nylander M, Berglund F. Does mercury from amalgam
restorations constitute a health hazard? Sci Total Environ
1990;99:122.
7
DHHS (Department of Health and Human Services). Public Health
Service. Dental amalgam: a scientific review and recommended Public
Health Service strategy for research, education and regulation. Final
report of the Subcommittee on Risk management of the Committee
to Co-ordinate Environmental Health and Related Programs. 1993.
8
Cutress TW. Dental amalgam and human health, current situation
a New Zealand review. In: Mjor IA, Pakomov GN (eds). Dental Amalgam
and Alternative Direct Restorative Materials. Geneva: World Health
Organization, 1997, pp. 6371.
9
Craelius W. Comparative epidemiology of multiple sclerosis and
dental caries. J Epidemiol Community Health 1978;32:15565.
10
Saxe SR, Wekstein MW, Kryscio RJ et al. Alzheimers disease, dental
amalgam and mercury. J Am Dent Assoc 1999;130:19199.
11
Seidler A, Hellenbrand W, Robra B-P et al. Possible environmental,
occupational, and other etiologic factors for Parkinsons disease: a
case-control study in Germany. Neurology 1996;46:127584.
12
Stejskal VD, Danersund A, Lindvall A et al. Metal-specific
lymphocytes:biomarkers of sensitivity in man. Neuroendocrinol Lett
1999;20:28998.
13
Fawcett J, Garrett N, Bates MN. Follow-up methods for retrospective
cohort studies in New Zealand. A N Z J Public Health 2002;26:25661.
14
Cox DR. Regression Models and Life-Tables. J R Statist Soc B
1972;34:187220.
15
Factor-Litvak P, Hasselgren G, Jacobs D et al. Mercury derived from
dental amalgams and neuropsychologic function. Environ Health
Perspect 2003;111:71923.
16
Ahlqwist M, Bengtsson C, Lapidus L. Number of amalgam fillings in
relation to cardiovascular disease, diabetes, cancer and early death in
Swedish women. Community Dent Oral Epidemiol 1993;21:4044.
17
Ahlqwist M, Bengtsson C, Lapidus L, Bergdahl IA, Schtz A. Serum
mercury concentration in relation to survival, symptoms, and
diseases: results from the prospective population study of women in
Gothenburg, Sweden. Acta Odontol Scand 1999;57:16874.
18
Ahlqwist M, Bengtsson C, Furunes B, Hollender L, Lapidus L. Number
of amalgam fillings in relation to subjectively experienced symptoms
in a study of Swedish women. Community Dent Oral Epidemiol 1988;
16:22731.
19
Malt UF, Nerdrum P, Oppedal B, Gundersen R, Holte M, Lne J.
Physical and mental problems attributed to dental amalgam fillings:
a descriptive study of 99 self-referred patients compared with
272 controls. Psychosomatic Med 1997;59:3241.
20
Bratel J, Haraldson T, Meding B, Yontchev E, Ohman SC, Ottosson JO.
Potential side effects of dental amalgam restorations: (1) An oral and
medical investigation. Eur J Oral Sci 1997;105:23443.
21
Bratel J, Haraldson T, Ottoson J-O. Potential side effects of dental
amalgam restorations: (2) No relation between mercury levels in the
body and mental disorders. Eur J Oral Sci 1997;105:24450.
22
Melchart D, Whr E, Weidenhammer W, Kremers L. A multi-center
survey of amalgam fillings and subjective complaints in non-selected
patients in the dental practice. Eur J Oral Sci 1998;106:77077.
23
Ganss C, Gottwald B, Traencker I et al. Relation between mercury
concentrations in saliva, blood, and urine in subjects with amalgam
restorations. Clin Oral Investig 2000;4:20611.
24
Bailer J, Rist F, Rudolf A et al. Adverse health effects related to
mercury exposure from dental amalgam fillings: toxicological or
psychological causes? Psychol Med 2001;31:25563.
25
Rothman KJ. No adjustments are needed for multiple comparisons.
Epidemiology 1990;1:4346.
26
Ingalls TH. Epidemiology, etiology, and prevention of multiple
sclerosis. Hypothesis and fact. Am J Forensic Med Pathol 1983;4:5560.
27
Kurtzke JF. Geography in multiple sclerosis. J Neurol 1977;215:126.

b
y

g
u
e
s
t

o
n

M
a
r
c
h

2
4
,

2
0
1
3
h
t
t
p
:
/
/
i
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

902 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
28
McMichael AJ, Hall AJ. Multiple sclerosis and ultraviolet radiation:
time to shed more light. Neuroepidemiology 2001;20:16567.
29
Van der Mei IAF, Ponsonby AL, Blizzard L, Dwyer T. Regional
variation in multiple sclerosis prevalence in Australia and its
association with ambient ultraviolet radiation. Neuroepidemiology
2001;20:16874.
30
Bangsi D, Ghadirian P, Dudic S et al. Dental amalgam and multiple
sclerosis: a case-control study in Montreal, Canada. Int J Epidemiol
1998;27:66771.
31
McGrother CW, Dugmore C, Phillips MJ, Raymond NT, Garrick P,
Baird WO. Multiple sclerosis, dental caries and fillings: a case-control
study. Br Dent J 1999;187:26164.
32
Casetta I, Ivernizzi M, Granieri E. Multiple sclerosis and dental amalgam:
case-control study in Ferrara, Italy. Neuroepidemiology 2001;20:13437.
33
Peduzzi P, Concato J, Feinstein AR, Holford TR. Importance of events
per independent variable in proportional hazards regression analysis.
II. Accuracy and precision of regression estimates. J Clin Epidemiol
1995;48:150310.

b
y

g
u
e
s
t

o
n

M
a
r
c
h

2
4
,

2
0
1
3
h
t
t
p
:
/
/
i
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m