Introduction

The benchmark dose (BMD) approach [including the use of the 95% lower confidence limit of benchmark dose (BMDL)] has gained popularity as a tool for estimating no-effect level in hazard associated with non-cancer responses (Bailer et al. 1997; Crump 2002; Filipsson et al. 2003; Slob et al. 2005; Muri et al. 2009). The application of these methods to epidemiological data has also been recommended (Bailer et al. 1997; Budtz-Jǿrgensen et al. 2001; Morales and Ryan 2005; Sand et al. 2006).

The application has been expanded from metal toxicology in environmental health [e.g., organic mercury on the central nervous system development (van Wijngaarden et al. 2006), cadmium on renal tubular dysfunction (as to be reviewed in the “Discussion” section), on pancreatic function (Lei et al. 2007) and on osteoporosis (Suwazono et al. 2010a)] to occupational health issues, for example, effects of lead on haematopoiesis (Murata et al. 2003; Karita et al. 2005) and manganese on Parkinsonian syndrome (Park et al. 2006; Santamaria and Sulsky 2010).

The methods have been used also in studies on lifestyle-related factors, such as alcohol consumptions on liver function (Dakeishi et al. 2004, 2006) and working hours on mental and physical fatigue symptoms (Suwazono et al. 2007). Recently, the European Food Safety Authority (2009a, b) employed BMD approach to issue an opinion regarding Cd levels in foods. However, the number of groups of populations analyzed in each of these studies was generally limited, and the stability or reproducibility of the BMD and BMDL values was seldom examined, leading to questions regarding whether the application of the BMD and BMDL methods to similar populations would give consistent values.

For the present study, data were available on Cd and the tubular effect marker levels in urine from more than 17,000 adult women in non-polluted areas in 16 prefectures in Japan. The present study was initiated to investigate whether consistent BMD and BMDL values would be obtained when the same procedures for BMD and BMDL calculation were applied to multiple but similar groups of people who lived in Cd-non-polluted areas in Japan with no occupational exposure to Cd.

Materials and methods

Ethical issues

The study protocol for the BMD and BMDL analyses was approved by the Ethics Committee of the Kyoto Industrial Health Association, Kyoto, Japan. Each of the participating women provided informed consent in writing at the time of each survey.

Data sources and methods of urinalyses

Urine samples were collected from 17,468 adult women in Cd-non-polluted areas in 16 prefectures (Prefectures 1, 2, 4, 5, 6, 14, 15, 16, 17, 18, 20, 26, 34, 39, 40 and 47; see Fig. 1 for locations) all over Japan from the northern-most Prefecture 1 to the southern-most Prefecture 47. Japan is divided into 47 prefectures, primarily by watershed mountain ranges, rivers or seas (see Fig. 1).

Fig. 1
figure 1

The locations of the sixteen prefectures where urine samples were collected. The sixteen prefectures of urine sample collection are shown with shades. The numbers correspond to the numbers in Table 2. Note that Japan consists of 47 prefectures in total

Full size image

For the present study, the data were cited from the six previous publications of Ezaki et al. (2003a), Tsukahara et al. (2003) for Prefectures 1, 4, 14, 15, 20, 26, 34, 39, 40 and 47, Moriguchi et al. (2005a, 2010) for Prefectures 5, 6 and 17, Yamagami et al. (2006) for Prefecture 16 and Ikeda et al. (2011) for Prefectures 2 and 18.

Urine samples were analyzed for Cd, the three renal tubular dysfunction markers of α1-microglobulin (α1-MG), β2-microglobulin (β2-MG), and N-acetyl-β-D-glucosaminidase (NAG), and creatinine (CR) by graphite furnace atomic absorption spectrometry [0.5 μg/l as the material limit of determination (LOD) for Cd], latex methods (0.6 mg/l for α1-MG and 1.0 μg/l for β2-MG,), a NAG test kit from Shionogi Pharmaceuticals, Osaka, Japan (0.01 unit/l for NAG) and a clinical auto-analyzer method (0.05 mg/l for CR), respectively. The analyses were conducted using the same methods throughout all studies. The results were corrected for CR concentration (Jackson 1966). The quality of Cd determination was certified by Round Robin (German Society of Occupational Medicine and Environmental Medicine, 2001–2009). The qualities of other assays were approved by Japan Medical Association (2001–2009) (Ezaki et al. 2003a, b; Moriguchi et al. 2010).

BMD and BMDL calculations

The hybrid approach by Budtz-Jǿrgensen et al. (2001) was employed. BMR = 5%, and BMD05 and BMDL05 (rather than BMD10 and BMDL10) were used; hereafter, BMD05 and BMDL05 are described as BMD and BMDL, respectively, unless otherwise specified.

α1-MG, β2-MG and NAG were introduced after logarithmic conversion for quasi-normal distribution of the response parameters (van Wijngaarden et al. 2006). For the present study, 93 cases with β2-MG ≥ 1,000 μg/g cr (Ezaki et al. 2003b; Bernard 2008) were deleted, and α1-MG and β2-MG data were available for the 17,375 cases in 16 prefectures (to be called ‘the total prefecture database’). NAG was measured in 6,409 cases. Nevertheless, preliminary analyses for BMD and BMDL revealed that dose–response relationship was not significant (p > 0.05) in Pref. 16 (as to be detailed in the “Results” section), and 566 cases in the prefecture were excluded in further analyses so that 5,843 cases in 9 prefectures were employed as ‘the selected prefecture database’ with complete sets of three effect markers of α1-MG, β2-MG and NAG (Table 1).

Table 1 Basic parameters by prefectures
Full size table

Distribution patterns and statistical analysis

Age and CR were distributed normally. A log-normal distribution was assumed for Cd, α1-MG, β2-MG and NAG. Accordingly, the distributions of age and CR were presented as arithmetic means and arithmetic standard deviations (i.e., AM ± ASD), whereas the Cd, α1-MG, β2-MG and NAG data were presented as geometric means and geometric standard deviations [GM (GSD)]. In calculating GM and GSD, the value below the corresponding LOD was assumed to be half the LOD.

Wilcoxon test and Spearman rank correlation analysis were employed when an assumption of a normal distribution was not applicable to the parameter distribution. Multiple regression analysis was also used. In evaluating correlation coefficients, a value of 0.2 was selected as a cut-off to identify the significance when the number was >100 (Ezaki et al. 2003a).

Results

Populations surveyed

The basic parameters are presented in Table 1 for individual prefectures and the total study population (Table 1) in terms of means (AMs or GMs) and standard deviations (ASDs or GSDs) for age, CR (as a urine density marker), Cd (as an exposure marker), and α1-MG, β2-MG and NAG (as the three effect markers). As NAG was measured only in selected prefectures, the numbers of available cases for NAG are shown separately from the numbers of the parameters other than NAG. The parameters are presented for the total prefecture database and for the selected prefecture database (‘A’ and ‘B’ in Table 1).

When the total prefecture database was taken as an example, the number of cases by prefecture was not uniform, distributing in a broad range from 650 (Prefecture 16) to 3,081 cases (Prefecture 26). In examining NAG, the fewest cases were observed in Prefecture 4 (104 cases), and the most abundant cases (1,789 cases) were in Prefecture 26. The lowest and highest levels of the exposure marker, Cd, were 0.79 and 2.67 μg/g cr as GM, respectively, with >3-fold difference. The grand GM was 1.34 μg/g cr. Prefectural GM values for α1-MG, β2-MG and NAG were distributed in relatively narrow ranges with the minimum and the maximum of 2.06–3.12 mg/g cr for α1-MG, 94–133 μg/g cr for β2-MG and 2.89–4.17 units/g cr for NAG. The GSDs by prefecture were mostly <2, but the GSD for NAG was up to 3 in four prefectures out of ten studied.

BMD and BMDL by prefectures

The calculation for BMD and BMDL values was possible for α1-MG and β2-MG in all prefectures in the total prefecture database (Table 2). In cases of NAG BMD and BMDL, calculation was not possible in 6 prefectures as no data were available (Table 1). In addition, the calculation results were not reliable in Pref. 16 due to poor dose–response relationship with p > 0.05 for the regression coefficient. Thus, the values were available for 9 prefectures with 5,843 cases (the selected prefecture database). BMD and BMDL for α1-MG and β2-MG were recalculated for the selected prefectures in parallel with NAG (Table 3).

Table 2 BMD and BMDL valuesa for α1-MG and β2-MG calculated for 16 prefectures
Full size table
Table 3 BMD and BMDL valuesa for α1-MG, β2-MG and NAG calculated for selected nine prefecturesb
Full size table

The BMD and BMDL for the total study population of 17,375 cases were 1.92 and 1.83 μg Cd/g cr, respectively, for α1-MG, 2.46 and 2.32 μg Cd/g cr for β2-MG (Table 2). There were variations in the calculated BMD and BMDL values for α1-MG and β2-MG among 16 prefectures. For example, the minimum and the maximum BMD were 0.78 and 3.18 μg Cd/g cr, respectively, for α1-MG, and 1.05 and 4.82 μg Cd/g cr, respectively, for β2-MG. Similar variations were observed also in BMDL values for α1-MG and β2-MG.

With regard to BMD and BMDL for NAG (Table 3), the values for a total of 5,843 cases in 9 prefectures were 2.32 and 2.09 μg Cd/g cr, respectively. Although the number of prefectures available for NAG BMD and BMDL calculation was limited to nine, the BMD and BMDL values for each of the nine prefectures distributed in a range of 0.70–4.98 μg Cd/g cr for BMD, and 0.53–3.50 μg Cd/g cr for BMDL.

Comparison on BMD and BMDL values for effect markers

Comparison of 16 BMDs for α1-MG and β2-MG (Table 2) by Wilcoxon test indicated a significant difference (p = 0.013) among the BMDs of the two effect markers. A similar comparison of BMDL values also showed a significant difference (p = 0.013). A further comparison of the three pairs of α1-MG and β2-MG, α1-MG and NAG, and β2-MG and NAG (Table 3), however, detected no significant differences (p > 0.05) in any pairs. The absence of the significant differences may be due to limited number (n = 9) of available prefectures (Table 3).

Factors possibly influencing BMD or BMDL

In a preliminary analysis, correlation of age (AM in years), CR (AM in g/l) and Cd (GM in μg/g cr) with BMD and BMDL for α1-MG, β2-MG and NAG was examined by calculating Spearman rank correlation coefficient. Cd correlated significantly (p < 0.01–0.05) with all of BMD and BMDL for α1-MG, β2-MG and NAG. In contrast, age and CR did not show significant correlation (p > 0.05) with any of BMD or BMDL for the three markers.

Subsequently, multiple regression analyses were conducted taking one of the BMD or BMDL for α1-MG, β2-MG or NAG as a dependent variable, and age, CR and Cd as independent variables (thus calculation of 6 cases). In all of the 6 cases examined (Table 4), the regression coefficient (r) was significant (p < 0.01) with R 2 of 0.74–0.96. Among the three independent variables, Cd correlated significantly (p < 0.01–0.05) with BMD and BMDL of both α1-MG and β2-MG. In addition, significant correlations of age with BMD and BMDL of α1-MG were observed. In case of NAG, CR showed significant (p < 0.01) correlations with both BMD and BMDL.

Table 4 Multiple regression analyses
Full size table

Discussion

The analysis of 17,375 urine samples from non-exposed adult women for α1-MG and β2-MG and 5,843 cases for NAG followed by application of the hybrid approach resulted in calculated BMD values of 1.92, 2.46 and 2.32 μg Cd/g cr for α1-MG, β2-MG and NAG, respectively, with the corresponding BMDL values of 1.83, 2.32 and 2.09 μg Cd/g cr (Tables 2, 3). Substantial variations were observed in BMD and BMDL values among prefectures. There was 4.1-fold to 4.6-fold difference in the BMD and BMDL values between the minimum and the maximum for prefectures when they were calculated for α1-MG and β2-MG in 16 prefectures, and the variation was even greater (i.e., 6.6-fold to 7.1-fold) for NAG in nine prefectures (Tables 2, 3). Attempts to identify BMD- and BMDL-influencing factors indicated that Cd levels in urine were most influential to BMD and BMDL of α1-MG and β2-MG among age, average urine density (in terms of the CR level in the urine) and the Cd concentration (Table 4), but none of them affected both BMD an BMDL of all of the three effect markers. Previously, Kobayashi et al. (2008) observed age-dependent decrease in BMD and BMDL values for β2-MG. It was not possible in the present analyses to observe such effects (Table 4).

To compare the present study results with the values in literature, studies were sorted for BMD and BMDL values of effect markers among the populations exposed to Cd either occupationally or environmentally, or among the populations with no anthropogenic exposures to Cd. Table 5 summarizes the results of the literature review along with the relevant results of the present study at the top of the table.

Table 5 BMD and BMDL values reported in literature
Full size table

Five studies (Kobayashi et al. 2006a, b, 2008; Shimizu et al. 2006; Uno et al. 2005) out of ten (the studies mentioned above plus Hong et al. 2004; Jin et al. 2004; Chen et al. 2006; Shao et al. 2007) studied Japanese populations. α1-MG was examined in one study (Suwazono et al. 2006), whereas β2-MG and NAG were measured in 15 and 12 groups of people, respectively. An examination of the reported BMD and BMDL values clearly indicates that the use of the 84% cut-off always gives smaller values than the 97.5% cut-off, as expected. Of particular interest is the range of reported BMD and BMDL values for each of the three effect markers. When studies on Japanese women (with 84% cut-off) were selected to ensure similarity in lifestyles with cases of the present study, the literature BMD and BMDL values for β2-MG were in ranges of 0.9–3.8 and 0.7–3.2 μg Cd/g cr, respectively, and the values were 0.8–4.7 and 0.6–3.7 μg Cd/g cr for NAG (Table 5). Thus, the minimum and the maximum values reported differ by four to sixfold for BMD and BMDL of β2-MG and NAG, respectively (the bottom of Table 5). The observation suggests that the variations reported in literature were apparently similar to what was observed in the present study (Tables 2, 3). A similar within-Japan variation was reported in a recent review of four studies in Japan, and one study each in China and Sweden (Suwazono et al. 2010b). As for the potential modifier for BMD and BMDL values, the present analyses suggest that Cd in urine affects BMD and BMDL of both α1-MG and β2-MG, age influences BMD and BMDL of α1-MG and β2-MG and CR those of NAG (Table 4).

It is of practical interest to identify an effect marker that gives the lowest BMD and BMDL among the three effect markers studied. The present analysis indicated that BMD and BMDL for α1-MG appeared to be lower than the values for and β2-MG when the total prefecture database was employed (Table 2). However, comparison of NAG with α1-MG and β2-MG by use of the selected prefecture database failed to identify any superior marker. Nevertheless, higher sensitivity for α1-MG in comparison with β2-MG is consistent with the previous observations by this study group (Moriguchi et al. 2004, 2005b) on better sensitivity of α1-MG than β2-MG as an indicator of Cd-induced effects on the renal tubules. Overall, published data on α1-MG-based BMD and BMDL values are still limited (Table 5) to examine whether α1-MG is the best marker of Cd-induced health effects.

An overall evaluation of the results of the present analysis in combination with findings in the literature suggests that both BMD and BMDL values vary substantially depending on the populations studied. Such was the case even when all data are from non-exposed adult women in a single nation.