How Mercury Was Absolved: Creativity, Collusion and Censorship

by Jeffrey Allen Trelka, M.Ed.
Algona, Washington
(download pdf)


Thimerosal and the Occurrence of Autism, Negative Ecological Evidence from Danish Population-Based Data” by Madsen et al. was recommended by the CDC, published in the September 2003 issue of PEDIATRICS and considered a "well-designed epidemiological study" by the IOM Vaccine Safety Review Committee in February, 2004.

The study appears to be a foul purport of scientific labor because of changes in participant selection methods, mid-study expansion of cohort populations and changes in diagnostic grouping. 

These concerns were formulated in a letter to the editor of PEDIATRICS.  They were not addressed and the letter was never published. 

“The greatest evil is…conceived and ordered in clean, carpeted, warmed, and well-lighted offices, by quiet men with white collars and cut fingernails and smooth-shaven cheeks who do not need to raise their voice.” —Clive S. Lewis [1] 

The 5/18/04 published findings from the Institute of Medicine (IOM) Vaccine Safety Review (VSR) Committee report on “Autism and Vaccinations” concludes, in part, with the following:

“Given the lack of direct evidence for a biological mechanism and the fact that all well-designed epidemiological studies provide evidence of no association between thimerosal and autism, the committee recommends that cost-benefit assessments regarding the use of thimerosal-containing versus thimerosal-free vaccines and other biological or pharmaceutical products, whether in the United States or other countries, should not include autism as a potential risk [2].”

Given the IOM’s conclusions, my concerns are two-fold.  First, there are at least three points that, if left without clarification, invalidate the Danish, Madsen et al. 2003 study, “Thimerosal and the Occurrence of Autism, Negative Ecological Evidence from Danish Population-Based Data” which was published in the September 2003 issue of Pediatrics [3].  This is of primary concern because the Madsen et al. study is one of the works that the IOM Vaccine Safety Review (VSR) Committee has deemed “well-designed.”  My concern related to the Madsen et al. publication is not that the study may be disconfirmatory evidence central to the thimerosal-autism controversy, but that it is a foul purport of scientific labor.  This is demonstrated among at least three points:  changes in participant selection methods, expansion of cohort populations completed mid-study, and changes in diagnostic grouping markedly skewed findings. 


The collection of inpatient treatment data since 1971 was polluted by adding outpatient activities post 1994 (Figure 1, below).  Madsen et al. attempts to reconcile this disparity by writing:  “In additional analyses we examined data using inpatients only…to elucidate the contribution of the outpatient registration to the change in incidence.  The same trend with an increase in the incidence rates from 1990 until the end of the study period was seen (data not shown) [3].”  This is an incredible claim in light of their previous declarations.  First, within the same study, Madsen et al. claim that “The proportion of outpatient to inpatient activities was about 4 to 6 times as many outpatients as inpatients…[3]”  Perhaps more importantly, in an earlier publication, Madsen et al. use the same data to purport:  “In our cohort, 93.1 percent of the children were treated only as outpatients…[4]”  This creates a huge disparity between the study published in Pediatrics and the previous one published in The New England Journal of Medicine which reportedly used the same data. The former study claims “4 to 6 times as many outpatients as inpatients”, yet the latter study maintains that 93.1% of all cohorts were outpatients and 6.9% were inpatients, yielding a proportion in excess of 13.5:1.  Despite the authors’ remarks claiming that the outpatient registration did not substantively change the incidence in autism, we may easily see, by making an adjustment for the more valid claim of 13.5:1 (Figure 2), that the original assertion of a statistically valid upward trend is due only to the difference in accounting of cohorts before and after 1995. 

Figure 1      -      Uncorrected data from Madsen et al. 2003 [3]
Reproduced with permission from PEDIATRICS Vol. 112 No. 3 September 2003, pp. 604-606  by the AAP

 Two additional corrections need be made in order to accurately reflect the autism incidence rates in Denmark thus contributing to our Figure 3.  First, it is necessary to address Madsen’s changes in diagnostic grouping. According to the authors:  “The date of onset was defined as the first day of the first admission leading to a diagnosis of psychosis proto-infantilis (International Classification of Diseases, Eighth Revision [ICD-8]: 299.00) or psychosis infantilis posterior (ICD-8: 299.01) or from 1994 onward, infantile autism (International Classification of Diseases, 10th Revision [ICD-10]: F84.0) or atypical autism (ICD-10: F84.1) [3].”  One might wonder why the authors feel comfortable changing the diagnostic grouping.  Although the authors do not fully address the effect of this change, data presented in Figure 1 of the Madsen et al. 2003 study show that the autism incidence prior to 1994 is on average approximately 6/10000 patients (derived by summing the incidence data for each of the three age groups presented and dividing that sum by the inpatient to total patient ratio).  In contrast, the earlier Madsen et al. 2002 publication reports autism incidence from anywhere between 1.2/10000 (DSM IV criteria) to 30.8/10000 (ICD-10 criteria).  The earliest available study on the incidence of autism in Denmark [5] reports an intermediate value of 4/10000.  Thus, using the pre-1992 ICD-8 psychosis infantilis incidence derived from the 2003 Madsen study (6/10000), the ICD-10/ICD-8 diagnosis ratio is at least 5:1 (i.e.,30.8 divided by 6) and may be as high as 25:1 (30.8 divided by 1.2).  To remain conservative in the analysis, a correction factor of 5:1 is applied to the data in Figure 3.

Second, one last correction needs be made in order to accurately reflect the autism incidence rates in Denmark thus resulting in our Figure 3.  In an article published in the American Journal of Preventive Medicine, Stehr-Green et al. write:  “Prior to 1992, the data in the [Danish] national register did not include cases diagnosed in one large clinic in Copenhagen (which accounts for approximately 20% of cases occurring nationwide).”[6]  This presumably means that all data points up to 1992 should be multiplied by 1.2 to correct for this discrepancy.

    Figure 2.     Madsen et al. 2003 data corrected for inpatient to outpatient ratios.         
Reproduced with permission from PEDIATRICS Vol. 112 No. 3 September 2003, pp. 604-606  by the AAP

Figure 3 reflects the aforementioned correction factors which are applied to the Madsen et al. 2003 data (previously revised for a correct inpatient to outpatient ratio, as in Figure 2).  In contrast to Madsen’s published claim, our correction for the disparity between the two studies (i.e., Madsen et al. 2002 versus Madsen et al. 2003), which purport to utilize the same data, reveals a dramatic decrease in autism incidence after 1993—particularly among the 2-4 year-old cohort. 

Figure 3.

  Madsen et al. 2003 data corrected for the difference between outpatient to inpatient ratio (13.5:1), diagnoses codes (conservatively at 5:1 based on Madsen et al. 2002) and missing data prior to 1992 from the Copenhagen clinic.
Reproduced with permission from PEDIATRICS Vol. 112 No. 3 September 2003, pp. 604-606  by the AAP

Further examining the 2-4 year-old cohort, it is apparent that if the discrepancy in the diagnosis codes before and after 1993 is not corrected for, the Madsen et al. 2003 study is essentially statistically indeterminate.  This is demonstrated in Figure 4 which shows P-values and Odds Ratios (OR) for average autism incidence calculated before and after 1993 for the 2-4 year-old cohort when applying corrections for diagnosis code differences before and after 1993 of between 0.5 and 25.  At a correction of 1.0 (i.e., no factor applied), autism incidence is nearly identical before and after the removal of thimerosal at an OR of 1.17 (P=0.39).  In fact, a statistically significant increase in autism incidence after the removal of thimerosal is seen only if the correction is lowered to 0.6 (OR=1.95, P<0.05).  Conversely if a modest upward correction of 2.0 is applied, a statistically significant decrease in autism incidence for this cohort is seen (OR=0.59, P<0.05).  At a correction of 5.0 used to calculate autism incidence in Figure 3, the odds ratio becomes 0.23 (P<0.0005).  Thus, at most (i.e., at a correction of 1.0), the Madsen et al. 2003 study is indeterminate.  When the data are treated properly, however (at a correction of 2.0 or great), it is easily shown that autism incidence decreased in the 2-4 year-old cohort as a direct consequence of the removal of thimerosal.

Figure 4.

  Calculated odds ratios (OR) and P-values for autism incidence before and after removal of thimerosal from vaccines for different diagnosis code correction factors.


Madsen attempted to publish the study in two respected journals and was turned down. This is when his supporters at the Centers for Disease Control and Prevention sprang into action.

On December 10, 2002, Assistant Surgeon General Jose F. Cordero MD, MPH, Director of the National Center on Birth Defects and Developmental Disabilities sent a “personal” letter to Jerold F. Lucey, MD, the editor of Pediatrics that started: 

“I am writing in support of an expedited review and consideration of the enclosed manuscript that examines the association between thimerosal, an ethyl mercury containing preservative and autism.”  

The letter ended:

“I feel this is a very important study that deserves thoughtful consideration by the Journal. Its findings provide one strong piece of evidence that thimerosal is not causally linked to autism.

Thank you for your consideration,..."

By so stating, Cordero not only validated the study’s contents but indeed anointed them strong and reliable enough to rule out a causal link between Thimerosal and regressive autism.

On February 9, 2004 the study was pronounced a "well-designed epidemiological study “ by the Vaccine Safety Review Committee of the Institute of Medicine. The committee rejected outright the “hypothesis” that thimerosal could trigger autism because it was theoretical and lacked supporting evidence. The committee also opinioned that further research to find the cause of autism should be directed toward other lines of inquiry “that are supported by current knowledge and evidence and offer more promise for providing an answer.”  


Another disturbing remnant related to the Madsen study is the absence of concern for nullifying arguments in frontline publications.  That is, when confronted with evidence that invalidates an already published study, the medical community has ignored substantive criticism by appealing to the IOM’s findings.  On 4/13/04, my concerns regarding the Madsen et al. 2003 publication, previously detailed herein, were submitted as a letter to the editor of the journal, Pediatrics.  This letter was in fact solicited by one editor who noted that the rebuttal would be most appropriately submitted as a letter to the editor rather than as a Post-publication Peer Review (P3R) submission.  Soon after the submittal of this letter, the authors of the Madsen et al. 2003 publication, “Thimerosal and the Occurrence of Autism, Negative Ecological Evidence from Danish Population-Based Data,” were notified of the letter and given a chance to reply to the issues elucidated therein.  These researchers chose NOT to reply, and on June 9, 2004, the editor of Pediatrics, Dr. Jerold F. Lucey, rejected the submitted letter without addressing the substance of my concerns. 

“A major review carried out by the US Institute of Medicine published last month (May 2004) found no evidence that thimerosal was linked to autism. Similarly, investigations by the UK Committee on Safety of Medicine, Europe's Agency for the Evaluation of Medicinal Products and the World Health Organization concluded the preservative was safe.  I consider this issue closed.”


Jerold F. Lucey MD
Editor-in-Chief, Pediatrics
University of Vermont College of Medicine
Pediatrics Editorial Office
89 Beaumont Ave, Given D201
Burlington, VT 05405-0068

The journal Pediatrics, like the members of the Institute of Medicine Vaccine Safety Review Committee, chose politics over science, thus squelching true debate over the issues surrounding vaccine safety.  It is all too often misunderstood that conventional publications do not customarily circulate studies and censure subsequent, substantive materials which challenge the validity of such works.  The previous example demonstrates a perhaps desired effect of the IOM VSR’s May 18th conclusions.  Additionally, it is completely clear that these findings were based upon profoundly flawed science, in contrast to what most professionals expect from solid, “well-designed” epidemiological studies.

My first concern detailing three criticisms of the Madsen et al. findings invalidates the Danish study.  The three criticisms together implicate a greater mischievousness occurring than simple carelessness.  The Madsen et al. publication takes on the form of a study, but leaves the aftertaste of propaganda through the apparent use of incongruent language, suspicious data collection, and data omission all of which are mechanisms of confusion rather than clarity.  Indeed, such mechanisms suggest that this team of researchers was trying to create truth; in contrast, science seeks to discover it.  In light of their declared outcome, one might be left wondering why they did not recommend a continued use of thimerosal as an autism preventative.  Given the authors’ trivializations, omissions, inconsistent data interpretations between studies, and complete lack of self criticism, it seems prudent to reflect on Madsen’s conclusions with extreme skepticism. 


I wish to thank the friends who have helped with the preparation of this manuscript and its mathematical calculations.

I have no financial conflicts of interest. I am the father of two children with medical diagnoses of autism.


[1]    Lewis, C.S. 1961. The Screwtape Letters. New York: Macmillan Publishing Company

[2]    Institute of Medicine of the National Academies. Immunization Safety Review: Vaccines and Autism. Washington, DC: National Academy Press

[3]    Madsen et al. 2003. Pediatrics 112(3):604-6

[4]    Madsen et al. 2002. New England Journal of Medicine 347(19):1477-82

[5]    Brask, 1972. A prevalence investigation of childhood psychoses. In Nordic Symposium on the Comprehensive Care of the Psychotic Children, 1972:145-153, Oslo: Barnpsykiatrist Forening

[6]    Stehr-Green et al. 2003. Autism and Thimerosal-Containing Vaccines: Lack of Consistent Evidence for an Association. Am J Prev Med 25(2):101-106