Hviid et al. 2019 Vaccine-Autism Study: Much Ado About Nothing?

4.1. Aims

A preceding retrospective cohort study (children born in Denmark 1991 – 1998) by Madsen and colleagues, including Hviid and Melbye, was published in 2002 in The New England Journal of Medicine (NEJM) [[74]], and titled, “A population-based study of measles, mumps, and rubella vaccination and autism.” Using analogous methodologies and data sources, Madsen et al. 2002 concluded as Hviid et al. that “This study provides strong evidence against the hypothesis that MMR vaccination causes autism.” Hviid et al. referred to this previous study and stated, “In this study, we aimed to evaluate the association again in a more recent and nonoverlapping cohort of Danish children that has greater statistical power owing to more children, more cases, and longer follow-up”, and “To evaluate whether the MMR vaccine increases the risk for autism in children, subgroups of children, or periods after vaccination.” The authors note the criticism that the earlier study “did not address the concern that MMR vaccination could trigger autism in specific groups of presumably susceptible children.” They claimed that their new study “addresses this concern in detail” by evaluating “the risk for autism after MMR vaccination in subgroups of children defined according to environmental and familial autism risk factors.”

4.2. Study Design, Methodology & Demographic, and Conclusions

The authors analyzed data for 663,236 children born in Denmark to Danish-born mothers from January 1, 1999, through December 31, 2010. Of these children, 5,775 were excluded, resulting in a cohort of 657,461 children. The observation period was from age one until August 31, 2013, so the earliest-born children had reached the age of fourteen by the end of follow-up, whereas the latest-born were still as young as two years. The total number of children who were followed until the end of the study was 650,943, and among these children, 6,517 (1%) had received a diagnosis of autism as of the follow-up end date. The average age of autism diagnosis for their study population was 7.22 years for children born in Denmark from January 1994 – 1999. Parner et al. by contrast, reported in 2008 that the average age of autism diagnosis in Denmark was 5 – 6 years, observing a decrease in the age of diagnosis over the study period [[75]]. Overall, the Hviid et al. study population was about 95% “vaccinated”, with an average vaccination age of 1.34 years (≈ 16 months). Among children with autism, 5,992 (92%), were “vaccinated” and 525 (8%) were “unvaccinated.” Hviid et al. summarize their methodology as follows; “Survival analysis of the time to autism diagnosis with Cox proportional hazards regression was used to estimate hazard ratios of autism according to MMR vaccination status, with adjustment for age, birth year, sex, other childhood vaccines, sibling history of autism, and autism risk factors (based on a disease risk score).”

The Cox regression model has been primarily applied by researchers for time-to-event analysis and allows one to estimate the hazard ratio (HR) of a given endpoint associated with a specific risk factor. Analysis by survival or time-to-event is frequently used in epidemiological and clinical studies [[76]]. As discussed by Anderson et al., time-to-event data typically feature challenges related to, among other things, censored observations and changes over time in the absolute or relative risks, as well as in the values of the predictors. In the context of “rare events” like autism, such approaches can suffer from erratic behavior [[77]]. A fundamental assumption underlying the application of the Cox model is proportional hazards; in other words, the effects of different variables on survival are constant over time and additive over a particular scale [[78]]. The chosen methodology also involved comparing the cumulative incidence of autism for the “vaccinated” and “unvaccinated” cohorts, calculated as the number of new events or cases of a disease divided by the total number of individuals in the population at risk for a specific time interval. Thereby, a child remained “at risk” of developing autism until they received an autism diagnosis or were otherwise “censored” from the study, meaning that they ceased to be included in the population of children under observation and hence ceased contributing to “person-years” at risk.

Children in the cohort contributed person-time to follow-up from 1 year of age to the end of the study on 31 August 2013, until a first diagnosis of autism, or censorship. So, for example, a child born in 1999 who was uncensored from the study until its end without having received an autism diagnosis would have contributed 14 years of “person-time” (or “person-years”) at risk, whereas a child born the same year who received an autism diagnosis in 2004 would have contributed five person-years at risk. The incidence rate among the study population was 129.7 cases of diagnosed autism per 100,000 person-years. Children were excluded due to diagnosis of any of several genetic disorders or conditions or censored due to death, emigration, or disappearance.

The key finding from the main analysis of the study was that “vaccinated” children were not at a higher risk of autism than “unvaccinated” children. As stated in the abstract, “Comparing MMR-vaccinated with MMR-unvaccinated children yielded a fully adjusted autism hazard ratio of 0.93 (95% CI, 0.85 to 1.02).” Figure 3 from Hviid et al. further summarizes HRs, confidence intervals, and p-values of correlation. Except for the reduced association of autism and the MMR vaccination for females (HR = 0.79, 95% CI = 0.64 – 0.97) all p-values were > 0.05, indicating no significant association. On this basis, the authors made bold conclusions that “The study strongly supports that MMR vaccination does not increase the risk for autism, does not trigger autism in susceptible children, and is not associated with clustering of autism cases after vaccination. It adds to previous studies through significant additional statistical power and by addressing hypotheses of susceptible subgroups and clustering of cases.” However, there are numerous reasons why their findings can and should not support these bold conclusions, including major study flaws, numerous discrepancies, and unexplained analysis; salient features, yet all of which remained oblivious to regulators, associations, and mainstream media in the USA.

5.1. Misleading Definition of “Genetic Susceptibility”, Exclusion of Children with High Susceptibility & Inadequate Sample Size

One of the criticisms of the studies cited by the CDC to support its claim that “vaccines do not cause autism” is that they do not consider the possibility of “susceptible subpopulations.” Hviid et al. acknowledge this and state, “Specific definitions of susceptible subgroups have been lacking.” The authors reference and follow the lead of Jain et al. [[79]] in defining “genetic susceptibility” merely as having “a sibling history of autism” at the time of study entrance. Therefore, if a child had an autistic sibling, but the sibling was not diagnosed until after the child had entered the study, then the child would have been misclassified as not “genetically susceptible.” Likewise, if a child had a genetic or environmental susceptibility but had no siblings, the child would have been wrongly classified. 49% of the study population was defined as having no “genetic susceptibility” simply by being an only child, as there were 319,936 children with no siblings out of a study population of 657,461.

The central dogma that autism is a highly heritable genetic disease is under debate. In 2011, Hallmayer et al. [[80]], in the largest twin study to date, reported moderate genetic heritability of 37-38% [[81]]. Later in 2014, using an epidemiological sample from Sweden, Gaugler et al. [[82]] concluded that autism's genetic architecture has a narrow-sense heritability of ≈52.4%, with most due to the common variation and rare de novo mutations. Considering current evidence of moderate autism genetic heritability, if a child had a genetic susceptibility and one or more siblings, but no siblings sharing the genetic trait or environmental trigger, the child would be wrongly classified as not “genetically susceptible.”

Moreover, and counterintuitive to the stated aims of the study, while touting their study as being designed to “address the concern that MMR vaccination could trigger autism in specific groups of presumably susceptible children”, Hviid et al. excluded 620 children who received a diagnosis during the first year of life of any of the following genetic disorders: neurofibromatosis, tuberous sclerosis, Angelman syndrome, Fragile X syndrome, Prader-Willi syndrome, Down syndrome, and DiGeorge syndrome. All these disorders demonstrate a higher rate of comorbidity with autism [[83],[84],[85],[86],[87],[88],[89]]. While they did not explain their rationale for these exclusionary criteria, their working assumption was presumably as follows: if these children were later diagnosed with autism, then it was due to their underlying condition and not vaccination. However, that syllogism is a non sequitur fallacy; the conclusion doesn’t follow from the premise. In effect, Hviid et al. treated all these conditions as competing hypotheses. One would contend they should have treated them as potential risk factors or indicators of epigenetic susceptibilities that might predispose these children to vaccine injury manifesting as symptoms of autism. Therefore, by excluding those children, the authors acted directly contrary to their stated purpose to investigate if “vaccination could trigger autism in specific groups of presumably susceptible children.”

As we have seen, the media touted this study as “large”, quoting a study population of 657,461. However, what the media consistently failed to point out is that only a small number met the authors’ definition of being “genetically susceptible”, with only 838 (0.13%) children meeting the criterion of having a sibling with autism. Following on, the reported HR for “siblings with autism” indicated an HR of 2.69 for autism (95% CI 0.58 – 12.63) among those who received the MMR vaccine compared to those who didn’t. Although this correlation was not statistically significant, one can speculate the result may have been significant if the authors had not excluded the 620 children with genetic disorders. We will never know since the authors have refused to release their underlying data to other scientists to be able to reproduce the authors’ findings.

Apart from genetic factors, Hviid et al. developed an “autism risk score” based on several “environmental autism risk factors”, but these were limited to “maternal age, paternal age, smoking during pregnancy, method of delivery, preterm birth, 5-minute Apgar score, low birth weight, and head circumference.” Although a low 5-minute Apgar score, low birth weight, and large head circumference may be indicative of a developing autism phenotype, it would be incorrect to label them as “risk factors” in the etiology or pathogenesis of ASD. In addition, apart from “smoking during pregnancy”, there is no consideration for assessment of the risk from xenobiotic environmental insults. This includes a long legacy of scientific literature spanning many decades implicating exposure to pharmaceuticals, industrial chemicals, and toxic and heavy metals in the etiology of ASD [[90]]. Future studies should apply a more rigorous risk assessment of exposure to environmental toxins [[91],[92],[93]] and other socioeconomic factors. For example, scientists could develop a risk score for exposure to environmental toxins based on factors including but not restricted to geographical location. Recent work by Palmer et al. [[94]] provides state-of-the-art approaches for rigorous assessment of chemical risk factors and intolerance in children and parents of developing autism and ADHD. This includes a complete evaluation of symptoms, intolerances, and life impacts of chemical, food, and drug exposures.

The second entry for the definition of the verb “lie” in Merriam-Webster’s dictionary is “to create a false or misleading impression” [[95]]. We would infer this is precisely what Hviid et al did when they delivered the public message that their study proved that the MMR vaccine “doesn’t cause autism even in children who are at greater risk of autism” or “genetically susceptible.” Indeed, Hviid et al. defined “genetic susceptibility” and did include children who met their definition. However, as outlined above, such an “ad hoc” definition of “genetic susceptibility” lacks scientific rigor and is inadequate and misleading.

5.2. Failure to control for “healthy user bias”

5.2.1. Jain et al.

A “healthy user bias” has been highlighted in previous studies for vaccination uptake [[96],[97],[98]]. In this scenario, parents of children who show symptoms at an early age or who have an older MMR-vaccinated sibling with autism, developmental delays, or other chronic disease, are more likely to skip the MMR vaccine, thereby biasing correlations in favor of finding no association. This “healthy user bias” was acknowledged by Hviid et al., who reference the study of Jain et al., [[99]] which “identified lower MMR uptake rates in children with affected siblings.” Published in 2015, Jain et al. investigated autism occurrence by MMR vaccine status among US children with older siblings with and without autism. The conclusion drawn by Jain et al. was that “In this large sample of privately insured children with older siblings, receipt of the MMR vaccine was not associated with increased risk of ASD, regardless of whether older siblings had ASD.” Based on the lower vaccine uptake among children who were considered at higher risk of autism due to “genetic factors”, such conclusions would need reevaluation.

To illustrate, as detailed by Jain et al., whereas the MMR vaccination rate for children with unaffected siblings was 84% at age 2 years and 92% at age 5, by contrast, the vaccination rate for children with autistic older siblings was 73% at age 2 and 86% at age 5. This would indicate that parents whose first child is diagnosed with autism after receipt of the MMR vaccine are less likely to get the shot for their second child for fear that it might contribute to the development of autism in the younger sibling. Similarly, parents who notice early developmental delays might skip the MMR vaccine for fear of it contributing to the development of autism. In the authors’ own words, considering that lower relative risk (RR) estimates were observed among children with autistic older siblings versus children with unaffected siblings, “It is possible, for example, that this pattern is driven by selective parental decision-making around MMR immunization, i.e., parents who notice social or communication delays in their children decide to forestall vaccination. Because as a group children with recognized delays are likely to be at higher risk of ASD, such selectivity could result in a tendency for some higher-risk children to be unexposed . . . . It is also plausible that parents of affected older siblings would be especially attentive to developmental delays in their younger children and decide to forestall immunization.” Thus, Jain et al. reasonably hypothesized that families with one child already affected by autism might be particularly concerned about this for any younger siblings, resulting in a lower vaccination rate among “genetically susceptible children.” In addition, although not statistically significant, Jain et al. found a negative correlation between the rate of autism in children with an autistic older sibling and receipt of the MMR vaccine. Rather than indicating some protective effect of the vaccine, we would speculate this would further indicate confounding by “healthy user bias.” This is an inherent risk of confounding in all observational studies, which needs to be accounted for and controlled for.

5.2.2. Hviid et al.

We know that Hviid et al. were aware of “health user selection bias” because they cited Jain et al. and acknowledged their finding of lower vaccine uptake among “susceptible children.” Yet they failed to account for it. Indeed, Hviid et al. affirmed that children who had siblings with autism had 7.32 times greater HR (95% CI 5.29 - 10.12) of autism relative to children who had siblings without autism. They also acknowledged the finding of Jain et al. that children with an autistic older sibling were less likely to receive the MMR vaccine. However, based on their analysis of the Danish study population, they claimed to have observed a vaccination rate of 95.19% and “no appreciable differences in vaccine uptake according to . . . . autism history in siblings.” On closer examination of the data, alternative interpretations are needed. For girls, the authors leaped baselessly to the conclusion that the vaccine is protective, asserting that MMR vaccination “reduced the risk for autism in girls.” We would conjecture that the overall negative (HR = 0.79, 95% CI 0.64 – 0.97) association could instead be due to girls at “higher risk” being less likely to receive the vaccine. The study’s table of population characteristics shows that 838 of the children in the study population had a sibling with autism, among whom 759 (90.6%) were MMR-vaccinated and 79 (9.4%) were not. Thus, whereas in the general study population only 4.8% were “unvaccinated”, the proportion who were unvaccinated among “genetically susceptible” children was nearly double that. Figure 3 from Hviid et al. further shows that among these 838 “genetically susceptible children”, 37 (4.4%, or 1 in 23) were diagnosed with autism. As discussed earlier, the HR shown for this cohort indicates a 2.69 times greater risk of autism among “vaccinated” children compared to “unvaccinated”. Among the 37 children diagnosed with autism, 32 were “vaccinated”, and 5 were “unvaccinated.” Therefore, 4.2% of the susceptible vaccinated children had an autism diagnosis compared to 6.3% of the susceptible unvaccinated children, indicating a possible pooling of children at “higher risk” into the “unvaccinated” group.

To approach the question from yet another angle, 759 of the 625,842 “vaccinated” children had an autistic sibling compared to 79 of the 31,619 “unvaccinated” children. Therefore, 0.12% of “vaccinated” compared to 0.25% of “unvaccinated” children were “genetically susceptible.” The “unvaccinated” were, thus, twice as likely to be “genetically susceptible” according to the author’s definition. In addition, the table shows that 319,936 children in the study had no siblings, among whom 4.2% were “unvaccinated”, and 331,994 had siblings without autism, among whom 5.3% were “unvaccinated.” These proportions contrast with the 9.4% of children with autistic siblings being “unvaccinated.” Children considered “genetically susceptible” were thus 1.8 times more likely than children with non-autistic siblings and 2.3 times more likely than single children to remain MMR-unvaccinated. Looking again at environmental risk factors for autism, Table 1 of the study shows that 3.97% of “very-low risk”, 4.35% of “low risk”, 5.44% of “moderate risk”, and 6.79% of “high-risk” children remained “unvaccinated.” Therefore, “high-risk” children were 1.25 times more likely than “moderate risk”, 1.56 times more likely than “low-risk” and 1.71 times more likely than “very low-risk” children to remain MMR-unvaccinated. This would again indicate potential confounding of “healthy user bias” with environmental and genetic risk factors.

Hviid et al. describe their study as “by far the largest single study to date” and state that it “allows us to conclude from one study that even minute increases in autism risk after MMR vaccination are unlikely, assuming unbiased results.” If their findings instead reflect the same “healthy user bias” identified by Jain et al., then the study by Hviid et al. does not allow us to draw such a conclusion. Strikingly, Hviid et al. acknowledge this limitation: “If the onset of symptoms results in avoidance of vaccination”, they admitted, bias in favor of no association “is possible.” Should they have said, “likely?”

5.3. Failure to Consider All Vaccines Routinely Recommended for Children in Denmark

The study of Hviid et al. focused only on the effect of the MMR vaccine on autism rates and not the complete Danish childhood vaccine schedule. In a secondary analysis, they also considered other routinely administered vaccines as covariables. From this secondary analysis, they concluded that “MMR vaccination did not increase the risk for autism in children characterized by other early childhood vaccinations. . . .” However, this analysis was limited to include only “other childhood vaccinations administered in the first year of life”, with no consideration of vaccinations given to children after twelve months other than the first MMR dose typically given at fifteen months. Other vaccines considered by Hviid et al. included five of the six vaccines recommended by Danish health authorities for routine use in infants under one year of age [[100],[101]]. Hviid et al. described the “mainstays” of the early Danish vaccine schedule as consisting of “MMR and diphtheria, tetanus, acellular pertussis, inactivated polio, and Haemophilus influenzae type b (DTaP-IPV/Hib) combination.” In Denmark, the combination of DTaP, IPV, and Hib vaccines has been recommended for infants, each with a three-dose course. However, Hviid et al. failed to consider that Danish authorities had since October 2007 additionally recommended three doses of the pneumococcal conjugate vaccine (PCV) during the first year of life [[102]]. The introduced formulation was the 7-valent PCV7, replaced with the 13-valent PCV13 starting in April 2010 [[103]]. An additional fourth “booster” dose of DTaP-IPV combination vaccine is also recommended at age 5 [[104]]. A four-dose course of hepatitis B (HepB) vaccine was recommended in 2005 for children born to mothers who are a carrier at birth and then 1, 2, and 3 months of age [[105]], which Hviid et al. further failed to consider. The authors also failed to consider the human papillomavirus vaccine (HPV) [[106],[107]]. According to the World Health Organization (WHO) [[108]], “The primary target group in most of the countries recommending HPV vaccination is young adolescent girls, aged 9-14.” In Denmark, Merck’s [[109]] quadrivalent HPV vaccine Gardasil was introduced in October 2008 as a catch-up program targeting 12-year-old girls, with routine vaccination for girls aged 12 years starting in January 2009. The study’s follow-up period was from January 1, 2000, through August 31, 2013, and girls of this initial birth cohort would have reached the age of thirteen or fourteen. While girls born in subsequent cohorts would have been too young to receive the HPV vaccine (unless administered earlier than the age of 12), girls in this 1999 – 2001 cohort may have received the HPV vaccine starting in 2011.

Thus, while this is not a flaw in the study per se, the choice by Hviid et al. to narrow their focus fails to meaningfully address parents’ concerns about the long-term effects on health outcomes of the complete and extended Danish vaccine schedule [[110]]. However, even if they had done such a study, its findings would not have been generalizable to the US childhood population since Denmark has a different schedule than that recommended in the USA.

5.4. Failure to Account for MMR Formulation Change

According to Hviid et al., the MMR vaccine used in Denmark from 2000 through 2007 contained the Schwarz strain of measles virus, which would have been GlaxoSmithKline’s (GSK’s) “Priorix” vaccine, and a different formulation was used from 2008 through 2013 that contained the Enders’ Edmonton strain, which would have been Merck’s “MMR-II.” This would indicate that children using the Merck formulation were much too young to receive an autism diagnosis as the oldest they would be at the time of study is 6 years of age or younger. On further evaluation, however, the information provided by Hviid et al. is incorrect; they mistakenly reversed the order in which MMR vaccines were used during those periods. According to a 2018 study on the use of the MMR vaccine in Denmark by Sørup et al., [[111]] until 2008, the Danish vaccination program used Merck’s MMR-II, which was marketed in Europe as “Virivac” and contained the Enders Edmonston B strain of measles virus [[112],[113],[114]]. From mid-October 2008, “Virivac” was replaced by GSK’s “Priorix”, which contained the Schwarz strain of measles virus [[115]]. Since mid-June 2013, a new version of Merck’s MMR-II has been used, which is manufactured by Sanofi Pasteur and marketed as “MMRvaxPro”, and which likewise contains the Edmonston strain of measles virus [[116],[117],[118],[119]]. Coauthor Christine Stabell Benn (personal communication, August 19, 2024) [120] corresponded with lead author Signe Sørup to confirm that the information in their paper was correct. GSK’s Priorix was used from 2008 until 2013, not Merck’s MMR-II, as mistakenly reported by Hviid et al. [[121]] (email correspondence is provided in supplementary material, Appendix 1). Following on, the average age of autism diagnosis for their study population was 7.22 years, and the typical age of first MMR vaccination in Denmark is 15 months. Since the study’s follow-up period ended on August 31, 2013, children who received “Priorix” would have been under 5 years of age and thus, on average, too young to receive an autism diagnosis [[122]]. Again, this could bias the study in favor of finding no association between the vaccine administered to the 2008 – 2010 birth cohort and the risk of autism.

5.5. Children Too Young for Autism Diagnosis

Hviid et al., report the average age of the sample as 8.64 years with a standard deviation (SD) of 3.48 years. The average age of autism diagnosis is reported as 7.22 years, with an SD of 2.86 years. If the age of diagnosis follows a normal distribution, 34.2% of the sample (z = -0.408) would be too young to get an autism diagnosis. This could account for as many as 3,387 additional cases not included in the analysis, which would further bias the outcomes to favor acceptance of the null hypothesis and no association between MMR vaccine and autism.

5.6. Failure to Consider a Change of Recommended Age for 2nd MMR Dose

Hviid et al. did not consider the second dose of the MMR vaccine in their primary analysis. In a secondary analysis, they reported “no evidence of a dose-response.” However, they failed to account for a change in the recommended age at which the second dose was administered. The first dose of MMR vaccine is recommended for children in Denmark at 15 months, followed by a second dose at the age of 4 years. However, before April 2008, the second dose was routinely administered at age 12 [[123]]. Therefore, children in the birth cohorts of 1999 – 2001 and 2002 – 2004 would not have received the second dose until years after the average age (7.22 years) of an autism diagnosis for the overall study population. Receiving the second dose of the MMR vaccine earlier in childhood development rather than in early adolescence may be associated with an increased risk of autism. The inclusion in the secondary analysis of those cohorts of older children who did not get both doses during early childhood would again bias the results erroneously in favor of acceptance of the null hypothesis.

5.7. Failure to Consider Maternal Vaccination

Maternal vaccination is another factor that Hviid et al. failed to account for in their study. It is recognized within the scientific community that maternal inflammation is associated with the development of autism in the offspring [[124]]. Vaccines intended to produce an immune response involving inflammation mechanisms could infiltrate the placenta, compromise fetal development, and increase the risk of ASD in offspring [[125],[126],[127]]. In alignment with recommendations by the WHO [[128]], the Danish Health Authority, since 2010, has recommended seasonal trivalent inactivated split influenza virus vaccination for pregnant women with selected high-risk chronic diseases in any trimester; and vaccination is additionally recommended for all pregnant women in the second and third trimesters [[129]]. Mølgaard-Nielsen et al. report up to 10% vaccine uptake by Danish women between 2010 – 2016. This means children born in the last cohort and reaching the age of 3 years by the end of the follow-up period may have been born to mothers vaccinated during pregnancy. Future studies should account for prenatal risk factors, including vaccination and the use of other pharmaceuticals during pregnancy.

5.8. Exclusion of Immigrants

Hviid et al. included only children “born to Danish-born mothers from 1 January 1999 through 31 December 2010” and registered in the Danish Medical Birth Registry, with the exclusion of 1,498 children. Asylum seekers to Europe may come from war-torn countries where health systems have broken down. There is evidence that asylum-seeking children have low coverage of childhood vaccinations in their home countries, as well as high uptake of immunizations in host countries [[130],[131]]. Therefore, immigrant children might receive multiple vaccines, doses, and boosters at once or spaced closer together to “catch up” on ones they may have missed in their home country. This might place immigrant children at higher risk of vaccine injury and developmental disorders such as autism. In addition, children of non-Danish ethnicity may have a higher risk of autism due to one or more epigenetic traits. Their exclusion could further bias the study’s findings.

5.9. Potential Misclassification of Study Subjects

As the authors acknowledge, “A limitation of our study is that we used the date of first diagnosis of autism, which is probably delayed compared with the age at onset of symptoms.” They then suggest this might bias their findings in favor of an association between MMR vaccination and autism by citing a hypothetical example in which “symptoms precede vaccination and diagnosis occurs after vaccination”, resulting in “misclassification of autism cases as vaccinated.” While they focus on the hypothetical scenario of bias favoring an association, they do not account for the misclassification of “vaccinated” children as “unvaccinated.” A study published in 2017 by Holt et al. [[132]], using data from the Danish National Health Service Register, addressed known concerns about the reliability of vaccination coverage data. To that end, the study authors compared MMR vaccination coverage according to medical records from general practitioners with that reported by the national registration database. Researchers report that the national database showed significantly lower vaccine coverage than medical records. Among the practices included in the study, the national database showed vaccine coverage of 86%, whereas the medical records showed coverage of 94%. The study authors state, “More than half of the children who were unvaccinated according to the register-based data (55%) had been vaccinated according to the medical records.” Vaccinated children being misclassified as “unvaccinated” in the study by Hviid et al. would, of course, bias their findings in favor of the null hypothesis and no association.

10.1. Hviid et al.

Hviid and his three coauthors (Hansen, Frisch, and Melbye), at the time of the study’s publication, were affiliated with the Statens Serum Institut (SSI), which develops vaccines and is “responsible for the purchase and supply of vaccines to the Danish national vaccination programs” [[155]]. Like the CDC, the SSI is a government agency and research institute; its purpose is to “ensure preparedness against infectious diseases and biological threats as well as control of congenital disorders.” For vaccine research, the SSI is “devoted to vaccines against tuberculosis, chlamydia, HIV and novel adjuvants to direct and potentiate the immune responses.” Upon the study’s publication, the SSI issued a press release proclaiming that it “once again invalidates the claim that the MMR vaccine increases children’s risk of developing autism” [[156]].

Financial support was provided by the Novo Nordisk Foundation [[157],[158]] and the Danish Ministry of Health [[159]]. The Novo Nordisk Foundation is a charitable foundation that issues funding grants for scientific research while owning the holding company Novo Holdings A/S [[160]], the majority voting shareholder in the Danish pharmaceutical corporation Novo Nordisk [[161]]. Novo Nordisk is a large multinational pharmaceutical company in Denmark with a market capitalization greater than US$497 billion [[162]]. According to their annual report, they anticipate an effective 2019 tax rate of 20-22%. The government of Denmark receives significant tax revenue from Novo Nordisk. Therefore, the Danish Ministry of Health and Novo Nordisk have a vested interest in a study that might influence the demand for the MMR vaccine. Also of note, Novo Holdings A/S investments include vaccine companies [[163]]. For example, in April 2019, the group invested tens of millions of dollars into Oxford Biomedica, which was involved in a consortium to develop and manufacture the AstraZeneca COVID-19 vaccine [[164],[165]]. In December 2023, the Novo Nordisk Foundation Initiative for Vaccines and Immunity (NIVI) was announced, [[166]] which is a partnership between the University of Copenhagen, and the SSI [[167],[168]]. The stated goal of NIVI is “to revolutionize and accelerate vaccine development in Denmark by bridging the gap between academic research and industry innovation.” Simultaneously, the foundation established a limited liability company, the Novo Nordisk Foundation Vaccine Accelerator (NVAC), to “facilitate the translational efforts of NIVI by providing industry-level expertise in vaccine development and conducting the early clinical testing of our vaccine candidates” [[169]]. Predating its vaccine initiative, the Novo Nordisk Foundation had funded numerous researchers in SSI and other institutions active in vaccine-related research [[170],[171],[172],[173],[174],[175]]. The foundation also funded the SSI “Danish National Biobank”, which aims to grant scientists access to data on residents in Denmark from national health registries along with information about biological samples [[176]]. The Danish Ministry of Health and the SSI unquestionably have a stake in preserving their own credibility and existing policies, not unlike the United States Department of Health and Human Services (HHS) and CDC. However, as summarized by Fig. 1 and as is the documented case with American regulatory agencies and professional bodies [[177]], an inherent conflict of interest in researching, marketing, and supplying childhood vaccines becomes apparent, if not explicit.

10.2. Annals of Internal Medicine

Two editors of the Annals of Internal Medicine, Jaya K. Rao (Deputy Editor) and Catharine B. Stack (Deputy Editor for Statistics), disclosed holding stocks in pharmaceutical companies active in vaccine research and manufacture. This includes Eli Lilly, Pfizer, and Johnson and Johnson. Eli Lilly is a former manufacturer of Dr. Jonas Salk’s inactivated polio vaccine and the developer of the mercury-based preservative thimerosal [[178]].