Mark Richardson corrects the American Dental Associations inaccurate critiques of his dental amalgam risk assessment presented to FDA

adaThe American Dental Association sent a critique of Mark Richrdson’s dental amalgam risk assessment to the FDA and described it as “seriously flawed” and urged the FDA not to use the risk assessment as the basis for any recommendations. Yet there are so many inaccuracies and falsehoods in the ADA critique that one must wonder if the ADA even bothered to read the risk assessment as their critique is so offbase as to render their opinion uninformed and obsolete.

In light of their intent to misinform ther FDA, Dr. Mark Richardson of SNC-Lavalin (and previously Health Canada) responds to the ADA critique with this rebuttal.

Supplemental Comments of the American Dental Association Before The Dental Products Panel of the Medical

Devices Advisory Committee

[Docket No. FDA–2010–N–0268]1

December 2010

It is the understanding of the ADA, that some seeking to ban dental amalgam are relying upon a certain report (Mercury Exposure and Risks from Dental Amalgam, November 8, 2010, prepared by SNC-Lavalin Environment, Ottawa, Canada (G. Mark Richardson, lead author)(the “Richardson Paper”). The report states that it was prepared in order to estimate current levels of mercury (Hg0) exposure from dental amalgam in the U.S. general population. Additionally, the report discusses the methodology and results of the Children’s Amalgam Trials, and compares Hg0 exposure from amalgam fillings to proposed Reference Exposure Levels (RELs).

The Richardson Paper is seriously flawed and certainly cannot form the basis for any well-founded recommendation to the FDA. Assumptions underlying portions of the paper are not explained and other data and extrapolations of data are not supported by the underlying source for that data.

Finally, the critique of the Children’s Amalgam Trials found in the Richardson Paper ignores the universally accepted, gold standard of randomized clinical trials in favor of less scientifically reliable evidence in determining true cause and effect. It also relies upon a discredited study in doing so.

Unfortunately, the Richardson Paper appears to be crafted to reach a particular result, rather than to properly focus on the best available scientific data.

The ADA offers the following specific comments regarding this report.

The data used to estimate the current levels of mercury exposure from dental amalgam is drawn from the CDC National Health and Nutrition Examination Survey (NHANES).

Data presented in Table 1, “Summary of NHANES data of 2001-02 and 2003-04”, on page 23 of the Richardson paper is misleading.

  • The age ranges (months) presented may not be appropriate for the defined age groups. The group defined as “Toddlers”, for example, range in age from 24 to 59 months. While there may not be a nationally defined range for “toddlers”, including 4 and 5 year old children in this group may increase the number of individuals with restored teeth in a group that is considered a sensitive population. One does not normally consider a toddler to be as old as 59 months. This may be why the “Toddlers” group as presented by Richardson et al. has more mean filled surfaces than children and adolescents.

mark-richardson-01Mark Richardson:
The age groups/ranges were defined through consultation with the FDA in advance of the work, in order to match age groups/ranges routinely employed by the FDA in the assessment of medical devices, drugs, etc.

  • The sixth column “Average number of restored surfaces (amalgam bearers only)” is illogical. NHANES data does distinguish between the restorative material used (e.g. amalgam versus composite). And, it would be assumed that the values in this column would be lower than the previous one which represents the “average number of restored tooth surfaces (all participants)”, assuming that this encompasses any type of restoration. The data in the sixth column also seem extraordinarily high, especially for toddlers. If this table is using Richardson’s Scenario 1, then it will also drastically over estimate the number of amalgam surfaces.

mark-richardson-01Mark Richardson:

The ADA is correct in the context that “amalgam bearers” was an error on my part. The numbers entered into the column are correct, but the column should have been labeled “persons with restored teeth only” or “dental restoration bearers only”. The intent was to indicate that the average values presented in column six excluded that proportion of the population with no (zero) dental restorations (unlike the data presented in Column 5 which included everyone (with and without restorations)). The numbers in column six apply only to that proportion of the population that has at least one restored dental surface, irrespective of the material used.

With respect to the suggestion of ‘numbers being high’ (in column seven), the data are what the data are. These data were drawn directly from the NHANES surveys. These data do relate to all types of restorations, up to and including 5 surface fillings (Scenario 1). There is no error here, and the likely non-applicability of scenario 1 is described and discussed in the text and is the reason why we developed multiple scenarios. However, it must also be kept in mind that many people have all of their fillings as amalgam, and it cannot be ruled out that persons with the maximum number of restored surfaces might have all of those surfaces restored with amalgam.

  • ADA in-house analysis of 2003-2004 NHANES data does not confirm the values presented in Richardson’s Table 1. Average number of restored tooth surfaces for “all participants” and “amalgam bearers only” for both toddler and children age groups were significantly lower by ADA calculations. ADA analysis found “all participants” and “amalgam bearers only” to be 0.53 and 9.62 respectively for toddlers and 1.83 and 5.5 for children. Compared to the numbers reported by Richardson, (1.2 and 18.2) and (3.4 and 8.8), respectively. This may indicate an overestimate by Richardson of the mean numbers for restored tooth surfaces that were subsequently used to calculate Hg0 exposure from amalgam fillings. This discrepancy would be expected to affect the findings presented in Table ES-2 where estimates for the “Proportion and numbers of US citizens with amalgam fillings that exceed doses associated with published reference exposure levels for Hg0” are presented.

mark-richardson-01Mark Richardson:

I suspect that the ADA attempted to calculate averages for “amalgam bearers only” (based on incorrect heading to column 6) and not for “persons with at least one restored surface”. Beyond that, we would have to sit down and work through the data together. However, I am confident in the data as we summarized it.

Data presented in table ES-01, “Summary of Hg doses estimated for the US population with amalgam fillings is also flawed. It appears that each of the following issues led to an over estimate of the dose of Hg associated with amalgam surfaces. These numbers were then used in Table ES-2 “Proportion and numbers of US citizens with amalgam fillings that exceed doses associated with published reference exposure levels for Hg0”.

  • “Number of filled surfaces” column does not correspond to “Average number of restored tooth surfaces” column in table 1 (mentioned above).

mark-richardson-01Mark Richardson::

The values presented in Table 1 relate to the data for the NHANES survey participants only (total N=17,267) whereas Table ES-01 relates to numbers for the US population as a whole (181.1 million with restored teeth), after application of the statistical weighting factors prescribed by NHANES for this data set. Therefore, the values in Tables 1 and ES-01 should not match.

NHANES specifically warns that the NHANES survey data does not apply to the total US population without application of those statistical weighting factors. As such, average values derived for survey participants and averages based on extension to the US population as a whole will be different, noting that NHANES surveys over sample certain sub-populations. This was discussed and described in Section 4.2 of the report.

  • In Scenario 4, 30% of the population is excluded because it is assumed that their restorations are 100% amalgam free. It is not explained how this 30% was chosen. Improper methods would bias the US population estimates because they are based on specific weights that NHANES2 has provided for each individual.

mark-richardson-01Mark Richardson:

The basis for this 30% as amalgam-free is clearly described in section 4.6 and 4.7 of the report. On page 32 of the report it states: “Finally, it was further assumed, in addition to the assumptions outlined for scenarios 2 and 3 above, that 30% of persons with restored tooth surfaces had all of those surfaces restored with a dental material other than amalgam. This assumption was made recognizing that approximately 30% of dentists in the US (Haj-Ali et al., 2005) reported being amalgam-free, and the possibility that all of their patients might have all existing fillings placed/replaced with materials other than amalgam. This assumes that dental patients are distributed equally across all dentists in the US.”

  • The mean numbers of filled surfaces are well above expected numbers. The authors state that these were derived as weighted US population mean, not the mean of NHANES participants. No other information on how the means were calculated is presented.

mark-richardson-01Mark Richardson:

Firstly, what are the “expected numbers”? When conducting research on this issue, I could find no good quantitative information in this regard, from ADA web sources or others. Therefore, we have relied on data presented by NHANES, which we consider the best such data available worldwide.

As for the methods to derive the weighted US population means, the methodology is described in Section 4.2 of the report. The statistical weighting is a variable that is unique to each NHANES survey participant and is included in the dataset itself. NHANES provides guidance and other information with respect to these weighting factors and their application when analyzing NHANES data. I can provide URLs for these websites should they be needed.

  • A background urine Hg concentration of 0.50 μg Hg/g creatinine was assigned for all age groups. This number is an estimate of the mean background urine Hg concentration for women of childbearing age with no restorations by Dye et al.3 It would be more appropriate to use the geometric mean of 0.31 μg Hg/g. Also, background exposure in toddlers, children and adolescents should be lower.

mark-richardson-01Mark Richardson:

There is much literature on the average background urinary Hg concentration; the value I selected was very reasonable and representative, based on the extant data. Irrespective of the background (amalgam-free) concentration of Hg in urine, the risk is presented by the incremental increase in urine Hg concentration due to the presence of amalgam, and not influenced by the assumed background concentration.

The ADA is incorrect with respect to use of the geometric mean. The geometric mean is a statistic that reflects the central tendency of data that is normally distributed (bell-shaped curve). For Hg in urine, the data are not normally distributed and must be statistically transformed to achieve a normal distribution. This is most frequently achieved by calculating the log of each datum, and then conducting statistical tests, etc on those log-transformed data. The geometric mean is the back-transformed (anti-log) value for the mean of the log values.

However, a geometric mean in no way reflects the true population average. The data transformation process is undertaken only so that statistical tests can be carried out on data that fit a normal distribution, a requirement of many statistical tests. However, people live in a linear world, not a log-transformed world. Therefore, the arithmetic mean is the most representative statistic to reflect population average values.

To provide a suitable analogy to make my point: consider a house with 10 rooms, with mercury vapour in each room, at the following concentrations (in ug/m3): 1, 1, 1, 1, 1, 1, 1, 7, 10, and 10 ug/m3.

If you spend equal time in each room then your average mercury vapour concentration that you would be exposed to is 3.4 ug/g. However, the geometric mean of these data is 1.93 ug/m3. 1.93 ug/m3 represents the central tendency of the levels of mercury vapour for all of the rooms, but the actual average concentration you would be exposed to is 3.4 ug/m3.

  • The last column, “Hg concentration”, which is estimated by the authors, does not correspond with actual laboratory measurements provided in the “Fourth National Exposure Report.” This report from the CDC summarizes urinary mercury (creatinine corrected) concentrations in μg/g of creatinine. Urine was collected and analyzed from 2537 of the same 2003-04 NHANES subjects >= 6 years of age. Never does the mean concentration for any of the group stratifications reach the minimum estimated value provided by Richardson in all scenarios and age groups (see CDC table below).

mark-richardson-01Mark Richardson:

ADA is reporting compiled statistics on the NHANES urine Hg data rather than looking at the data itself. I presented the following information within the report, based on direct examination of the NHANES data:” For an adult, each amalgam-filled surface results in an increase of Hg in urine of 0.1 μg Hg/L or 0.06 to 0.07 μg Hg/ g creatinine (see studies summarized in Table 2). Therefore, a hypothetical average adult with 100 amalgam-filled tooth surfaces would have a predicted incremental increase of Hg in urine of 10 μg/L, or 6 to 7 μg/g creatinine. 10 μg Hg/L urine or 6 to 7 μg Hg/g creatinine in urine falls well within the range observed for the general US population. Urinary Hg concentrations measured as part of NHANES (2003-04) ranged up to 75.75 μg/L and 36.1 μg Hg/g creatinine (N=2538).”

Please keep in mind that the data presented in the summary cited by the ADA relates to all those surveyed, those with and without amalgam fillings. Therefore, the average values will be depressed in the cited summary (relative to my analysis) due to the inclusion of a large proportion with no amalgam fillings.

Therefore, is it quite apparent that my estimated urine Hg concentrations do, in fact, fall well within the range of actual NHANES measurements.

More transparent methods and results are necessary in order to accurately interpret the results of this study. Further analysis of NHANES data regarding surface restorations and urinary mercury concentrations would be useful for assessing the accuracy of the population estimates provided in the report.

mark-richardson-01Mark Richardson:

It is apparent by many of my responses above that the ADA clearly did not read the report in its entirety. Many of their criticisms are answered by direct reference back to the report itself. The report, in its entirety, is clear, transparent, and provides adequate description such that a scientist interested in duplicating this work would be able to do so.

Richardson’s arguments against the findings of the Casa Pia Children’s Amalgam Trial,4 because urinary Hg at baseline and year seven was not different between the amalgam and composite groups, is likewise without sound basis. This randomized controlled trial allowed examination of the effect of amalgam placement by comparing two statistically similar populations over seven years. 

mark-richardson-01Mark Richardson:

I agree that randomized clinical trials are the ‘gold standard’; however, only if conducted correctly. There was one major problem with respect to the Children’s Amalgam Trials (CATs). The active agent for potential effects due to amalgam is exposure to mercury. The failure of CAT researchers to ensure that the control group (those receiving composite resin fillings) had significantly lower Hg exposure than the exposed group (those receiving amalgams) means that no difference in type, frequency, severity, or speed of on-set of effects would be expected. It is, therefore, totally unsurprising (i.e., totally expected) that the CAT studies in fact reported no significant differences between the exposed and control cohorts. You would not expect anything else given their equivalent or near-equivalent Hg exposure, as determined by urine analysis. Reliable conclusions regarding the ‘safety’ of amalgam cannot be drawn from studies in which mercury exposure was similar in the dosed and control groups.

As an analogy, if a drug were being tested in a clinical trial, and confirmatory testing for exposure demonstrated that the control group had the same levels of drug metabolites in their urine as did the dosed group, these results would be rejected (by the FDA and others) as being unusable for evaluation of the drug’s efficacy or side effects given that the researchers failed to ensure that the control group had no drug exposure. You cannot draw conclusions regarding the types of effects, relative incidence of effects, relative severity of effects, or relative speed of on-set of effects, if the dosed and control groups are demonstrated to have similar levels of exposure as determined by confirmatory testing (in this analogy, of drug metabolites in urine). This same logic applies to the Children’s Amalgam Trials.

The study clearly showed that the amalgam group had an increase in urine Hg after amalgam placement that peaked at two years with no effect on neurobehavioral function at year seven.

mark-richardson-01Mark Richardson:

That the amalgam groups within the CATs ‘clearly’ had Hg exposure greater than the control group is questionable. In the Casa Pia study, the amalgam group had urinary Hg levels only 1 to 1.5 μg/g creatinine greater than the control group. For the New England CAT, at year 5, the treatment group had a mean urinary concentration that was only 0.3 μg/g creatinine greater than the control group. These differences are not biologically or toxicologically significant, nor is this sufficient differentiation (WRT dose) to adequately quantify differences in toxicological response between exposed and control groups, particularly given sample size and the sub-chronic nature of exposure (5 to 8 years, depending on publication date relative to study start date).

It should also be mentioned that the recent CAT in China (Ye et al. 2009) had similar lack of exposure discrimination between the amalgam and control groups. In that study, the treatment group had a mean urinary concentration of mercury that was only 0.2 μg/g creatinine greater than the control group.

It should also be noted that the dose-response analysis from these CAT studies was inadequate. This is best demonstrated by the recent publication of Geier et al (2011) that demonstrated a clear dose-response effect in the Casa Pia study, of urine mercury levels on porphyrin residues in urine (a marker for impacts on the heme synthesis pathway), once all confounding dose-related factors were properly controlled. This dose-response effect was previously reported as insignificant due to an inadequate analysis by Woods et al (2009).

The dose-response analysis of neurobehavioral effects was equally inadequate. This dose-response analysis should have employed individual participant data, rather than the dose range grouping approach that was used. It is surprising that one of the CAT authors (Bellinger) did not employ a more rigorous dose-response analytical technique, given that he was a co-author on studies of the effects of lead (Pb) exposure on children’s IQ (see Lanphear et al. 2005, and others); in that Pb study, dose and effect data were plotted and assessed at the individual study participant level, and care was taken to control for confounders to the extent that data would allow. This is type of analysis that is required for the CAT studies (as proven by Geier et al 2011). I would encourage the CAT authors to pursue these lines of analysis.

Rather than assessing exposure in children using the gold standard – randomized controlled studies, or RCTs, Richardson uses Lettmeier’s5 derived REL, among others, to assess risk in the general population Lettmeier’s calculations are based on a study conducted by Ngim et al.6

mark-richardson-01Mark Richardson:

Lettmeier et al (2010) in no way relates to the Ngim et al study. Lettmeier et al reported on an independent, recent investigation of the neurobehavioral effects of mercury vapour exposure in cohorts exposed as a result of using Hg for gold amalgamation in Zimbabwe and Tanzania. Ngim et al investigated similar effects, but in a cohort of dentists in Singapore.

The Ngim study has a number of flaws quoted below.

According to Mackert and Berglund,7

mark-richardson-01Mark Richardson:

This referenced study is now 14 years old and is now unreliable as a reference given the significantly increased body of literature on the topic that has arisen over this time period. This is the reason why a new exposure assessment was undertaken rather than relying on the results of my 1995 assessment for Health Canada. Perhaps state-of-the-art when it was published, it is now similarly out of date and the significant body of new relevant studies and literature made it essential to undertake a new study.

“Two studies that reported subclinical neurobehavioral effects resulting from low-dose occupational exposure to mercury among dentists have attracted interest because of their potential for use in the establishment of exposure limits for mercury (Ngim et al., 1992; Echeverria et al., 1995). The Ngim et al. study was “rejected for use in deriving an MRL [Minimal Risk Level] because of methodological and reporting deficiencies.” by the ATSDR (1994). Among the deficiencies cited by the ATSDR were the following: “The exposure status of the subjects was known to the investigator during testing, mercury levels were not reported for controls, and methods used to correct for confounders (especially the common use in this population of traditional medicine containing mercury) were not reported.” Other deficiencies and inconsistencies in the Ngim et al. study involve the method in which exposure to mercury was estimated. The authors measured the dentists’ exposure on a single work day and used this measurement as an estimate of exposure, despite the fact that additional monitoring of four of the ten dentists showed large variations in the air mercury levels. Three of the four dentists’ offices exhibited high and low air mercury values that differed by a factor of two or more during the one-week monitoring period. It is not surprising, therefore, that the reported average (arithmetic mean) blood levels for the dentists in the Ngim et al. (1992) study (12.3 μg/L) are much higher than would be expected from the known Hg-air:Hg-blood ratios (Roels et al., 1987). Using the regression equation of Roels et al. for relating Hg-blood to Hg-air, a blood level of 12.3 μg/L would correspond to an air level of 28.5 μg/m3, not 16.7 μg/m3 (the arithmetic mean reported by Ngim et al., 1992).

Other factors besides mercury can reasonably be cited as producing the measured performance differences on the neurobehavioral tests. For example, the dentists in the Ngim et al. study typically worked 10 hours per day, six days per week. It is unlikely that the staff members of the National University of Singapore who served as controls had comparably demanding work schedules. It must be borne in mind, when considering the relatively slight differences in the aggression scores (Figure 3 of Ngim et al., 1992) exhibited by dentists, that they are a self-selected group that likely would have garnered higher aggression scores even prior to entering dental school. Echeverria et al. (1995) criticized the Ngim et al. (1992) study for the lack of an appropriate referent group and other shortcomings.”

mark-richardson-01Mark Richardson:

All studies have weaknesses. My primary perspective is that the weaknesses of the Ngim et al (1992) study are less of a concern than the concomitant exposure to chlorine gas that occurs in chloralkali worker studies, and which will confound (reduce) the exposure and effects of Hg vapour exposure. The confounding by chlorine gas exposure for chloralkali worker studies is detailed in Richardson et al (2009). It should be noted that EPA, in their RfC development, did not reject the Ngim et al (1992) study. Rather, they interpreted it as one of the two ‘bounding studies’; the bounding studies basically represented, in EPA’s analysis, the minimum and maximum exposure levels associated with measureable neurotoxic effects in the key studies selected by EPA.

The hypothetical issue of other substances that can cause similar effects to those associated with the toxicant under study is also common to most occupational studies. However, where data are not specifically presented to demonstrate that those other substances are a real factor in the observed health effects, such hypothetical exposure(s) cannot be used as a basis for study rejection. The role of other substances would be relevant to consider if the Ngim et al (1992) study was the only study to demonstrate the effects of Hg vapour exposure. However, there are hundreds of studies of Hg vapour exposure that repeatedly demonstrate and confirm the neurotoxic effects observed by Ngim et al (1992). Therefore, the hypothetical exposure to those other substances, and their hypothetical role in the observed effects, can be effectively dismissed.

{slide=Reference}:

Ngim, C-H., Foo, S.C., Boey, K.W. et al. 1992. Chronic neurobehavioral effects of elemental mercury in dentists. Br. J. Ind. Med., 49(11): 782-790.

Richardson, GM, R Brecher, H Scobie, J Hamblen, K Phillips, J Samuelian and C Smith. 2009. Mercury vapour (Hg0): Continuing toxicological uncertainties, and establishing a Canadian reference exposure level. Regulatory Toxicology and Pharmacology, 53: 32-38{/slide}

Based on these initial observations, the ADA believes that the findings of the report, “Mercury Exposure and Risks from Dental Amalgam, November 8, 2010, prepared by SNC-Lavalin Environment” are flawed and significantly over estimate Hg0 exposure from amalgam fillings.

{slide=References}

1 Questions regarding these supplemental comments may be directed to Jerome Bowman, ADA Public Affairs Counsel at bowmanj@ada.org.

2 Centers for Disease Control and Prevention: National Health and Nutrition Examination Survey (http://www.cdc.gov/nchs/nhanes.htm accessed 12-01-10).

3 Dye BA, Schober SE, Dillon CF, Jones RL, Fryar C, McDowell M, Sinks TH. Urinary mercury concentrations associated with dental restorations in adult women aged 16-49 years: United States, 1999-2000. Occup Environ Med. 2005;62(6):368-75.

4 Rather than assessing exposure in children using the gold standard – randomized controlled studies, or RCTs, Richardson uses Lettmeier’s5 derived REL, among others, to assess risk in the general population Lettmeier’s calculations are based on a study conducted by Ngim et al.6 The Ngim study has a number of flaws quoted below.

5 Lettmeier B, Boese-O’Reilly S, Drasch G. 2010. Proposal for a revised reference concentration (RfC) for mercury vapour in adults. Sci Total Environ. 2010;408:3530-3535.

6 Ngim CH,Foo SC, Boey KW, Jeyaratnam J. Chronic neurobehavioural effects of elemental mercury in dentists. British Journal of Industrial Medicine 1992;49:782-790.

7 Mackert JR Jr, Berglund A. Mercury exposure from dental amalgam fillings: absorbed dose and the potential for adverse health effects. Crit Rev Oral Biol Med. 1997;8(4):410-36.

{/slide}

[iframe width=”560″ height=”349″ src=”http://www.youtube.com/embed/EKgRiCam_3I” frameborder=”0″ allowfullscreen]

About The Author

Leave a Reply