Norway vaccinated a generation against rotavirus. Type 1 diabetes did not fall.

The idea that a routine infant vaccine might prevent an autoimmune disease is attractive enough that it has survived a decade of inconsistent evidence. Rotavirus infection has long been suspected as an environmental trigger for type 1 diabetes, and studies of rotavirus vaccination and type 1 diabetes risk have produced conflicting results [s1]. A national analysis published in Diabetologia on 28 August tests it in about as clean a setting as observational epidemiology offers, and the answer is no [s1].

The natural experiment

Norway introduced rotavirus vaccination for infants in September 2014, and uptake was both rapid and high — the feature that makes the country usable as a natural experiment [s1]. Researchers followed every child born in Norway between 2007 and 2019 from six months to five years of age for a diagnosis of type 1 diabetes [s1]. Of 740,744 children included, 846 were diagnosed in that age window [s1].

The primary analysis was an interrupted time series comparing type 1 diabetes incidence in birth cohorts before and after the programme began [s1]. This is a stronger design than a simple vaccinated-versus-unvaccinated comparison, because it does not depend on the characteristics of families who accept or decline a vaccine — the confounding that has muddied most of the previous literature.

What it found

The primary interrupted time series showed an increasing trend after the programme was implemented: a relative hazard ratio of 1.11 (95% CI, 1.03–1.20) [s1]. That direction is the opposite of the protective effect the hypothesis predicts, and it did not survive the study's own sensitivity analysis. After excluding the 2013–2015 birth cohorts to avoid potential herd immunity effects around the transition, the estimate fell to 1.05 (95% CI, 0.93–1.18) [s1] — no longer distinguishable from no effect.

A secondary analysis comparing fully vaccinated with non-vaccinated children gave an adjusted hazard ratio of 1.17 (95% CI, 0.78–1.77) [s1], a confidence interval wide enough to include a meaningful reduction and a meaningful increase.

The authors' conclusion is that the study found no evidence of a reduced risk of type 1 diabetes after the introduction of infant rotavirus vaccination [s1].

Consistent with the UK

A 2021 cohort study using UK primary care records reached a similar place from a different direction. Britain introduced infant rotavirus vaccination in 2013; researchers followed 880,629 children born between 2010 and 2015 from six months to seven years, of whom 343,113 (39.0%) received at least one dose — 93.4% of those born after the programme started [s2].

Across 4,388,355 person-years with a median follow-up of 5.66 person-years, 733 participants had recorded type 1 diabetes, an incidence of 17.1 cases per 100,000 person-years [s2]. In adjusted analysis, rotavirus vaccination was not associated with type 1 diabetes risk (HR 0.89; 95% CI, 0.68–1.19) [s2]. The same study found no association with coeliac disease either (adjusted HR 1.05; 95% CI, 0.86–1.28 across 1,657 cases, an incidence of 38.0 per 100,000 person-years) [s2].

Two national programmes, two designs, two null results.

What the finding is worth

Null results on off-target vaccine effects tend to be reported quietly and read as a disappointment. They should be read as two things at once.

The first is the deflation of a hypothesis. The rotavirus–type 1 diabetes link had generated real scientific interest and some public expectation that a vaccine already in the schedule might turn out to prevent a serious chronic disease. On this evidence it does not.

The second is safety information, and it runs in the useful direction. Type 1 diabetes incidence did not rise after national introduction once the transition cohorts were handled [s1], and vaccinated children in the UK cohort were not at higher risk of type 1 diabetes or coeliac disease [s2]. The authors of the UK study framed their result explicitly as further evidence of rotavirus vaccine safety [s2].

Both studies carry the standard limits of their design. Neither can rule out a small effect in either direction, and the Norwegian primary estimate's initial positive signal is a reminder that interrupted time series are sensitive to how the transition period is handled — the two published estimates, 1.11 and 1.05, differ only in which birth cohorts were excluded [s1]. Reporting both is the honest presentation, and the authors did.