Health Reports
Cadmium levels and sources of exposure among Canadian adults

Warning View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

by Rochelle Garner and Patrick Levallois

Release date: February 17, 2016 Correction date: (if required)

Cadmium is a heavy metal found in the environment largely as a result of industrial and agricultural processes, but it can also occur naturally.Note 1 Cadmium can have negative health consequences, including increased risk of cancer, kidney dysfunction, skeletal damage, and possible cardiovascular effects.Note 1Note 2 Non-occupational exposure is generally through cigarette smokingNote 3 and consumption of foods high in cadmium.Note 1Note 4Note 5

Owing to a biologic half-life of 10 to 35 years, cadmium accumulates in the body, mainly in the liver and kidneys.Note 6 The level of cadmium in the blood is a good indicator of recent exposure, and urinary cadmium is indicative of long-term exposure, although overlap is substantial.Note 7

This analysis uses data from cycles 1 (2007 to 2009) and 2 (2009 to 2011) of the Canadian Health Measures Survey (CHMS) to examine blood (BCd) and urinary (UCd) cadmium levels among Canadians aged 20 to 79, and associations with sources of exposure.


The CHMS is an ongoing, biennial survey that collects health information through a household interview and direct physical measures at a mobile examination centre (MEC). The CHMS covers the population aged 3 to 79 (minimum age 6 in cycle 1) in the 10 provinces. It excludes  full-time members of the Canadian Forces and residents of reserves and other Aboriginal settlements in the provinces, institutions, and certain remote regions. Together, these exclusions represent less than 4% of the target population.Note 8Note 9

The CHMS had overall response rates of 57.1% in cycle 1 and 55.5% in cycle 2. Among households that provided a roster of members, response rates were much higher at 88.3% and 84.9% for the household and MEC components, respectively, in cycle 1, and 90.5% and 81.7% in cycle 2.Note 8Note 9 The analytic sample for the present study was limited to respondents aged 20 to 79 (7,095) excluding pregnant women.


Biological sample collection and analyses

To obtain nationally representative information on a variety of biomarkers (such as nutrition, chronic and infectious diseases, and environmental exposure), blood and urine were collected from all eligible respondents at the MEC. Blood samples were collected from respondents aged 20 to 79 by a phlebotomist using a standardized venipuncture technique. Respondent-provided urine samples were collected using the first-catch urine in cycle 2, and mid-stream urine in cycle 1. Cycle 2 respondents were asked not to urinate for two hours before their appointment. This may have resulted in a shift in creatinine levels between cycles 1 and 2, which could affect creatinine-adjusted levels of some chemicals.

BCd and UCd were measured by inductively coupled plasma mass spectrometry (ICP-MS), Perkin Elmer Sciez, Elan DRC II. Urine creatinine was measured using the colorimetric Jaffe method. Absorbance was read at 505 nm on a Hitachi 917 chemical analyzer. The level  of detection for blood cadmium was 0.045 µg/L in both cycles. The detection level for urinary cadmium was 0.09 µg/L for cycle 1 and 0.07 µg/L for cycle 2. Urine samples with creatinine  below the level of detection (0.049 µg/L) were excluded from analyses (n = 11).Note 10 For the remaining samples, observations below the level of detection for the particular measure (BCd, UCd or urinary cotinine) were imputed with values at half the level of detection. When the level of detection for a test differed between cycles, the highest was used (UCd, 0.09 µg/L; urinary cotinine, 1.1 µg/L). All values were rounded to two significant digits,Note 11 and were converted from Système International units to conventional units.

For descriptive analyses, UCd was divided by the respondents’ urinary creatinine to yield UCd adjusted for creatinine (µg/g). Thirteen respondents were missing both the BCd and UCd measures.


Cigarettes and the smoke they generate contain relatively high levels of cadmium in the particulate phase, ranging from 10 to 250 ng per cigarette.Note 3 Smoking status was based on self-report and was categorized as never-smoker, former smoker, and current smoker. Self-reported never-smokers (n = 41) and former smokers (n = 147) with urinary cotinine levels greater than 50 ng/mL were re-classified as current smokers.Note 12Note 13 An indicator variable was used to identify individuals reporting daily or almost daily exposure to to second-hand smoke (SHS) at home, in a private vehicle, at work, or in a public place.


In 2007, Canadians were estimated to have an average daily dietary intake of 0.22 µg of cadmium per kg of body weight.Note 5 During the household interview, respondents reported the frequency with which they consumed specific foods and food categories. Based on the literatureNote 14Note 15Note 16Note 17and Health Canada’s Total Diet Study,Note 18 yearly consumption of the following foods was examined as a possible source of cadmium:  liver (excluding liverwurst and liver pâté) and/or other organ meats (for example, kidneys, heart, giblets);  shellfish (lobster, shrimp, mussels, scallops, oysters, clams, squid, crab and other shellfish);  cereals (hot or cold); white or brown breads (including bagels, rolls, pita bread, tortillas); fries and hash brown potatoes; other potatoes (excluding sweet potatoes); green salads; dark leafy greens (spinach, mustard greens or collards, excluding kale); and nuts and seeds (peanuts, walnuts, seeds, or other nuts, excluding nut butters such as peanut butter). Annual consumption values had a minimum of 0, but no prescribed maximum, as individuals could consume items from a particular food category many times a day. A dichotomous variable was included to indicate whether respondents had consumed red meat in the past year.

Shellfish consumption was determined differently in cycle 1 and cycle 2. In cycle 1, shellfish consumption was assessed with one general item during the household interview. In cycle 2, shellfish consumption was assessed during the MEC interview with nine separate items; annual consumption was asked only of respondents who reported consumption of the specific shellfish in the past month. Responses to the nine items were summed for cycle 2 respondents to yield annual consumption. To improve comparability between cycles, consumption less than 12 times a year from cycle 1 was converted to zero consumption (n = 742). In each cycle, about the same percentages of respondents were non-consumers of shellfish (44.6% in cycle 1, 41.0% in cycle 2, chi-square p = 0.30), although the frequency of consumption was greater among cycle 2 consumers (mean 60.1 times a year versus 37.1 times a year in cycle 1, t-test p-value < 0.0001).

Based on earlier studies,Note 16Note 17 several other covariates were included in the models.  Hemoglobin levels below 120 g/L in women and below 130 g/L in men were used as indicators of anemia.Note 19 Body mass index (BMI), defined as weight in kilograms divided by height in metres squared, was derived from respondents’ measured height and weight. Because parity is associated with cadmium levels in women,Note 16Note 20 the reported number of live births was included in female-specific models: zero (nulliparous), one, two, or three or more. An indicator of CHMS cycle was included to account for other unmeasured differences between cycles.

Statistical analysis

For descriptive analyses, the geometric means of cadmium levels were produced, along with their  standard errors. Dietary consumption items were usually categorized as non-consumption (0 times a year), less than weekly (1 to 51 times a year), and at least weekly (52 or more times a year). The exceptions were shellfish, for which the lowest consumption category was fewer than 12 times a year, and cereal and bread, with categories of non-consumption, less than daily (fewer than 365 times a year), and daily (365 or more times a year).

For regression analyses, the outcome variable was the natural logarithm of the BCd or UCd concentration. As recommended by Barr et al.,Note 21 the urinary creatinine level was included as a covariate in the UCd model; this value was also transformed using the natural logarithm. Dietary consumption measures were centered at their mode values, and divided by 12 so that a unit change in the consumption variables reflected consuming the particular product an additional time per month. Mode values were: 0 for organ meats (63.7% were non-consumers), shellfish (42.6%), and leafy greens (31.5%); 52 for french fries (22.6%), other potatoes (22.6%), and nuts and seeds (15.4%); 104 for salad (16.2%); and 365 for bread (19.8%). The mode value for cereal was 365, but because so few respondents had higher consumption values (0.6%), this variable was centered at its median of 104.

Continuous BMI was centered at 21.75, and age was centered at 45. Both variables were subsequently divided by 5 so that a one-unit change in the covariates related to a 5-unit change in BMI and age.

To assess the impact of the selected predictors of cadmium levels, multivariate models were built in a sequential block-wise fashion, showing the change in model R2 associated with each block of covariates. Blocks were added in the following order: linear and quadratic effects of age; smoking status (never-smokers as reference) and exposure to SHS; consumption of certain foods; and other factors, including anemia, BMI, CHMS cycle indicator, and parity (women). Urinary creatinine (ln-transformed) was included in the model for UCd before adding age. Separate models were developed for men and women. Sex-specific models were also run for never-smokers.

All analyses were conducted in SAS-callable SUDAAN (version 11.0.1) using 24 denominator degrees of freedom. Analyses were weighted using the cycle 1-cycle 2 combined sample weights and bootstrapped to account for the sample design.Note 11


The mean age of the 7,082 respondents was 46.3 (SE 0.15) for women and 45.3 (SE 0.18) for men. Nearly a quarter of respondents (23.2%) were current smokers, 28.1% were former smokers, and 48.8% were never-smokers. Close to a quarter (23.6%) were exposed to SHS daily or almost daily.

Cadmium levels were significantly higher in women than men, in older relative to younger people, in former and current smokers relative to never-smokers, and among people exposed to SHS (Table 1). Figure 1 shows the cumulative distribution of BCd levels by sex and smoking status. Figure 2 shows the same distribution for UCd, adjusted for creatinine. Regardless of smoking status, women had higher average cadmium levels than men, a difference that was most pronounced among never-smokers and current smokers.

Frequent consumption of organ meats was associated with higher cadmium levels; frequent  consumption of leafy greens and nuts and seeds was associated with lower levels (Table 1). Frequent consumption of cereal, salad, leafy greens, and nuts and seeds was associated with lower BCd. Frequent consumption of bread, french fries and hash browns, and nuts and seeds was associated with lower creatinine-adjusted UCd. Consumption of red meat was not related to cadmium levels (BCd or creatinine-adjusted UCd). Anaemia (low hemoglobin) was significantly associated with higher creatinine-adjusted UCd,  but not with BCd. BMI was significantly related to cadmium levels, as was parity among women.

Multivariate results

The combined association of these characteristics with cadmium levels was examined in multivariate regression models. Smoking status and SHS exposure explained the greatest amount of variation in BCd levels (Table 2). Current and former smokers had significantly higher BCd than did never-smokers. Among men who reported daily or near-daily exposure to SHS, BCd was 22% greater than among men exposed less frequently; the relationship was not significant for women (p = 0.11).

Age explained the next greatest amount of variation, with older respondents having higher BCd than did younger respondents. As a block, dietary factors explained about 1% of the variation in BCd levels. BCd was lower among people with frequent cereal consumption. Among women, frequent shellfish consumption was positively associated with BCd, while frequent consumption of potatoes other than french fries was negatively associated with BCd; these were not significant for men (p = 0.07 and p = 0.08, respectively). For both sexes, higher BMI was related to lower BCd levels. Anaemia was associated with higher BCd among men, but not among women.

After urinary creatinine, age was the next largest contributor to variations in UCd, followed by smoking status and SHS exposure (Table 3). As with BCd, former and current smokers had higher UCd than did never-smokers, although the impact of smoking status (compared with never-smokers) on UCd was less than on BCd. As well, men and women who reported daily or near-daily SHS exposure had higher UCd than those exposed less frequently. Diet generally explained 1% or less of variation in UCd. Men who were anaemic had significantly higher UCd than those who were not. Parity was significant for women—those who had given birth had higher UCd than nulliparous women.


When analyses were limited to never-smokers, the explanatory impact of diet on cadmium levels increased, and the strong association with age persisted.  Cadmium levels were significantly higher among never-smokers who did not consume red meat (except for UCd in women). Among men, frequent organ meat consumption was associated with higher BCd, and frequent bread and cereal consumption was associated with lower BCd (Table 2). Among women, frequent consumption of potatoes (excluding french fries and hash browns) was associated with lower BCd (Table 2) and UCd (Table 3). Higher BMI was negatively associated with BCd and UCd among never-smokers. UCd was 30% to 50% higher among female never-smokers who had children, compared with those who were nulliparous (Table 3).


The present study found that cadmium levels were strongly associated with smoking, while the impact of diet was modest to small. Consistent with the literature,Note 7Note 16Note 22 current and former smokers had higher BCd and UCd than did never-smokers. Exposure to SHS was associated with significantly higher UCd among both sexes, and with higher BCd among men overall. However, SHS exposure was not associated with cadmium levels among never-smokers, which is similar to other research.Note 22 In the present study, exposure to SHS was less frequent among never-smokers than among current smokers (15% versus 52%, p < 0.0001), which may explain why it was significant overall but not among never-smokers. The association with SHS may also depend on time since exposure.

Although consumption of certain foods was associated with cadmium levels, the variation explained by diet tended to be small in multivariate models. The explanatory contribution of diet increased dramatically when the analysis was limited to never-smokers, thereby eliminating the potential contribution of tobacco smoke. Some earlier research found consumption of shellfish and organ meats to be related to higher cadmium levels, but in the present study, these associations were not consistent in all models.Note 1Note 14Note 15 Moreover, other studies also failed to find an association between shellfish consumption and cadmium levels when controlling for other factors.Note 14Note 15Note 16Note 23

Consumption of other high-cadmium foods, such as potatoes, nuts and seeds, and leafy greens, was either not significant, or frequent consumption was associated with lower cadmium levels. Such findings illustrate the complex interplay between dietary composition and the absorption of metals and other toxins. For example, levels of iron, fiber and zinc may affect the degree to which cadmium is absorbed by the body.Note 24Note 25Note 26

Adjusting for urinary creatinine in the UCd models explained the greatest proportion of variation in cadmium levels. Creatinine adjustment corrects for inter-sample differences in the concentration of urine, which can affect levels of other chemicals. Cadmium and creatinine levels in urine are highly correlated (Pearson correlation rho=0.43, p < 0.0001 in the present study). Furthermore, urinary creatinine is associated with other covariates, such as age and BMI.Note 27 Therefore, some of the variance explained by urinary creatinine could also be attributed to these factors.

When urinary creatinine was taken into account, the remaining covariates explained about 10% of the variation in UCd. Given that UCd is considered to be an indicator of long-term exposure,Note 7 current behaviours (for example, diet, smoking) reported to a cross-sectional survey are limited in their ability to explain differences in exposure levels. Accounting for longer-term behaviour and exposure may explain more of the differences in UCd at the population level.

Consistent with other reports,Note 7Note 22 women in the present study had higher cadmium levels than men. Choudhury et al. have proposed that, because of their lower iron stores, women absorb a larger proportion of ingested cadmium.Note 28 An association between iron stores and cadmium levels has been demonstrated by others.Note 24Note 29 Such an explanation is consistent with this study’s findings that non-consumption of iron-rich red meat was related to higher cadmium levels, and that frequent consumption of cereals and breads, which are often fortified with iron, was related to lower cadmium levels. Anaemia (low hemoglobin) was significant only among men (although a positive association with BCd  emerged for never-smoking women). When whole blood ferritin levels were used as a measure of iron store rather than hemoglobin levels (cycle 2 respondents only), low (less than 30 µg/L) ferritin was significantly predictive of higher BCd for women (beta = 0.20, p = 0.004) and men (beta = 0.44, p = 0.002), but was not associated with UCd for either sex.

Factors beyond smoking and diet were also related to cadmium levels.  BCd and UCd were lower among those with higher BMI, overall and when analyses were limited to never-smokers. Some research has found higher BMI to be associated with lower cadmium in bloodNote 23 and urine,Note 30 although others failed to find an association.Note 14 Consistent with previous analyses,Note 16Note 23 the current study found that women with more children had significantly higher cadmium levels, particularly UCd, compared with nulliparous women. Åkesson et al. surmised that decreased iron stores among parous women explained their higher cadmium levels.Note 20


This analysis has a number of limitations. Because the CHMS is cross-sectional, the present study can describe associations between cadmium levels and other factors, but it is not possible to assign causality. In addition, although geographic variations in exposure may explain some of the differences in cadmium levels,Note 17 the CHMS was designed to produce only national estimates. Further, differences between cycles 1 and 2 in the way data were collected, particularly urine samples and shellfish consumption information, may affect results. As well, it was not always possible to isolate the consumption of specific foods in a category, and foods such as sweet potatoes and tofu, found to be associated with cadmium levels in other studies,Note 14 were not included in the CHMS. Consumption frequency was based on recollection and extrapolated to yearly values, which may not reflect actual consumption. Lastly, the response rate of the combined sample from cycles 1 and 2 was 53.5%. This largely reflects a low response rate among households initially contacted for participation; households that agreed to provide a household roster had high response rates (more than 80%) to the interview and MEC components of the survey. It is not known if households that initially declined to participate differed in important ways from participating households.


According to the present study, smoking behaviour was the greatest contributor to cadmium levels among Canadians aged 20 to 79, with modest or small contributions from diet. However, among non-smokers, diet may be a significant source of cadmium.

Date modified: