Hepatitis B vaccines and the risk of multiple sclerosis: a comment on the study by Langer-Gould et al. (2014)

In accordance with my previous criticism regarding the methodological reliability of most studies presented as confirming the safety of vaccines, this investigation [1] raises a number of serious concerns.

Case ascertainment – Whereas the study title makes special emphasis on multiple sclerosis (MS: ICD code 340), case identification includes no less than nine ICD codes, some of which (optic neuritis or acute disseminated encephalomyelitis being sometimes difficult to differentiate from genuine MS, whereas others [transverse myelitis] are generally considered as distinct). The most expected result of such a diagnosis blending is to weaken statistical power and to blur epidemiological evidence.

Vaccination assessment – Only 4.0% of the 3885 controls were exposed to hepatitis B vaccine in the 3 years before the index rate; this may be compared with the study by Hernan et al. [2] (the design of which was fairly similar), where 2,4% of the 1565 controls were exposed to a recombinant hepatitis B vaccine. The trouble is that this immunization was highly selective in the latter population (UK), whereas it was massive in the former (USA). In spite of this major discrepancy in the vaccine policy between the two countries, the surprisingly small difference between these two percentages raises the hypothesis that, for one reason or another, vaccination recording was incomplete in the American sample. Although duly pointed out as remarkable by Langer-Gould et al., low vaccine exposure in their sample was not seriously discussed by the authors.

Control selection – Although a black ethnicity was the most prominent risk factor identified by the authors in their previous study on the incidence of demyelinating syndromes (quoted as reference 17 in their current paper), one may wonder why their control selection did not include race in their matching method. As it happens, imbalance in the distribution of black race between cases and controls was the most striking feature of the baseline samples characteristics.

Index date – Although the timing of symptoms appearance is generally a crucial argument for causality in drug monitoring (there may be exceptions to this rule), this parameter is never properly considered in investigations devoted to post-vaccine MS. Actually, as the disease may remain clinically silent for years, the relevant parameter is neither the date of diagnosis nor that of the late symptoms which lead to the investigations leading to positive diagnosis. In spite of this, what investigators mean by “symptoms onset date” is never clearly defined: which symptoms? For example, in their abovementioned reference 17 (Table 1), Langer-Gould et al. estimated at 0.9 month the median time from symptom onset to diagnosis, after having stipulated that, defining MS required two or more episodes of MS “separated in time”: is unlikely that 0.9 month is a sufficient time interval to separate two distinct MS episodes… At the opposite side of the clinical spectrum, the very first symptoms of a MS are often an unexplained fatigue, mild paresthesia, etc. the onset of which may be quite close to the time of vaccine injection (a few days or weeks), but which may last for years before onset of more significant symptoms: thus, if one focus on the late significant symptoms, this very long time lag is almost always interpreted as speaking against a vaccine role whereas, when considering the whole of symptoms sequence from its very beginning (i.e. from the time of quite discrete symptoms just after injection), it is on the contrary highly suggestive of a vaccine causality. I have never seen this crucial problem properly taken into account in any database, so that most investigations about the time between vaccination and the onset of MS symptoms are essentially misleading.

Regarding MS and in spite of their denials, the authors ended up to a result very close to that of Hernan et al.’s., namely an overrepresentation of cases (4.2%) as compared to the controls (3.1%) within a time windows of 3 years. Of course, this difference just failed to reach statistical significance but: i) as documented above, the methodological tendency of the authors contributed to decrease the power of their results; ii) amongst the published case/control studies supposed to exclude a post-vaccine risk of MS (by means of like strategies of dilution of the cases or of insufficient observation period), the number of those suggesting (even in a nonsignificant way) an overrepresentation of cases in vaccinated subjects is clearly higher than those suggesting an underrepresentation, and the difference between the two groups of studies is clearly significant from a statistical point of view.

Finally and as with most papers devoted to the safety of hepatitis B vaccines, the authors cannot refrain from concluding that no “change in vaccine policy” is warranted: yet, their investigation is totally devoid of the slightest element likely to validate any vaccine policy, whose potential shortcomings (included issues of cost, of resources allocation, of individual and collective efficacy, of nonneurological risks, etc.) go far beyond the sole issue of MS. In psychoanalysis, such optimism (going far beyond the available evidence from a given investigation) is called “the return of the repressed”…