Rotavirus Epidemiology and Monovalent Rotavirus Vaccine Effectiveness in Australia: 2010–2017

This study was set in New South Wales, Australia (7.48 million total population in 2016; annual birth cohort of ∼96 000), and involved all subjects residing in New South Wales over the study period.^14,15 

Notifications of all laboratory-confirmed rotavirus between January 1, 2010 (when notification commenced), and December 31, 2017, as reported under legislation to the New South Wales Notifiable Conditions Information Management System were included.^16,17 Laboratory-definitive evidence includes the detection of rotavirus antigen in stool samples by enzyme immunoassay, latex agglutination, or reverse transcriptase polymerase chain reaction (RT-PCR).¹⁷ 

Genotype analysis of samples positive for rotavirus from hospitalized patients at select sites around Australia is performed as part of the Australian Rotavirus Surveillance Program at the Murdoch Children’s Research Institute.¹⁸  All available samples positive for rotavirus with adequate volume are processed and genotyped by RT-PCR. Results from a deidentified sample of fecal specimens collected in New South Wales from January 1, 2017, through December 31, 2017, were analyzed. Information regarding contributing laboratories, specimen collection, and analysis have been described elsewhere by Roczo-Farkas et al.¹⁸ 

The Australian Immunisation Register (AIR), a population-level national register, was used to identify the rotavirus immunization status of children residing in New South Wales.¹⁹  The vaccination status of patients notified as having rotavirus was obtained from the case report in the New South Wales Notifiable Conditions Information Management System and the AIR, and the vaccination status of controls selected for the case-control VE analysis (see below) was obtained from the AIR. The AIR contains data on all children in Australia with a Medicare number (the universal health insurance scheme) and data on any child vaccinated in Australia, even if the child is a temporary visitor. Any child who did not have rotavirus vaccination status recorded on the AIR was classified as unvaccinated.

Notification rates were calculated by using the midyear estimated resident populations released by the Australian Bureau of Statistics for 2010–2016 and population projections for 2017.²⁰  Incidence rate ratios and 95% confidence intervals (CIs) were presented for comparison of notification rates by age, sex, and notification year. Genotyping data were analyzed by age, sex, and G and P type and defined as either vaccinelike or wild-type strain.

Case patients (rotavirus notifications from the New South Wales Notifiable Conditions Information Management System) were individually matched to controls (deidentified from the AIR) by date of birth ±1 day. For each identified case patient, up to 10 controls were randomly selected. This method has been used previously and allows for likely exclusion of the case patient from being randomly selected as a control by virtue of a similar, but not the same, date of birth.^6,21,–23 

Children were included in the VE analyses if they were eligible for at least 1 dose of the funded rotavirus vaccine (ie, children born between January 1, 2008, the first full calendar year of National Immunisation Program–funded rotavirus vaccine, and June 31, 2017) and were aged ≥6 months at the time of notification. This age cutoff for the analysis was applied to avoid inclusion of young infants notified as having rotavirus disease but who had vaccine virus shedding (see genotypic analysis and results below). Rotavirus became notifiable in December 2009; thus, notifications from January 1, 2010, were included as the first full calendar year of notifications. A dose was considered valid if the rotavirus infection notification date was >14 days after receipt of the rotavirus vaccine to allow for time to develop a protective immune response. A dose of rotavirus vaccine was considered invalid if dose 1 was administered at <5 weeks of age, if dose 2 was administered <4 weeks after dose 1, or when the child was ≥52 weeks old. These children, and any who received a dose of RV5, for example during residence in another jurisdiction, were excluded from the analyses. If a child had a vaccine dose recorded as dose 2 on the register, but a first dose was not recorded, the “prior dose” assumption was made, assuming 2-dose receipt.²⁴  The inclusion of case patients in the VE analyses is described in Fig 1.

VE estimates and 95% CIs were calculated by using a matched case-control analysis, using the formula VE = 1 − odds ratio (OR) × 100%. Conditional logistic regression models were used to determine the OR for the VE calculations. Analyses included stratification by age group at notification (aged 6–11 months, 1–3 years, and 4–9 years) and year of notification. ORs were adjusted for sex if the addition of this changed the point estimate by >5%. For the time since vaccination model, 5 defined time periods after dose 2 (<1, 1–<2, 2–<3, 3–<5, and 5–<10 years) were analyzed, with children who were unvaccinated set as the reference category. The analyses were conducted by using Stata version 14.0 (Stata Corp, College Station, TX).

The study was approved by the Australian National University Human Research Ethics Committee. Approvals for access to data held by the New South Wales Ministry of Health and the Australian Government Department of Health were granted.

There were 9517 rotavirus notifications during the 8-year study period. The median age was 5 years, with the range being 0 to 103 years; 63.2% were <10 years old and 26.3% were <2 years old. There were 4758 (50.0%) boys and men, and of all with a recorded Aboriginal and Torres Strait Islander status; 176 (7.6%) identified as Aboriginal or Torres Strait Islander. There were no deaths recorded as attributed to rotavirus in this notification data set.

Fifty percent of case patients (n = 4752) were not eligible for vaccination because of their age. Of the vaccine-eligible individuals, 496 case patients (10.4%) reported receiving 1 dose of rotavirus vaccine, 2814 case patients (59.1%) reported receiving ≥2 doses of vaccine before notification, and 1455 case patients (30.5%) did not have the vaccine recorded in the AIR.

In 2017, 2319 cases of rotavirus were reported, representing a rate of 29.7 per 100 000 population (Fig 2). This was significantly higher than the 2016 notification rate of 9.7 per 100 000 (incidence rate ratio: 3.1; 95% CI: 2.8 to 3.3). Before this, annual rotavirus notification rates also fluctuated with increases in 2012 and 2015. The number of rotavirus notifications reported in 2017 was 1.3 times (95% CI: 1.2 to 1.3) the previous highest recorded annual number of notifications in 2012.

There was no significant difference in the rotavirus notification rate between males and females (P = .36). The highest notification rates in 2017 were among children aged 1 year, followed closely by infants aged <1 year (Fig 3). Rates decreased with age but rose again in the elderly aged ≥80 years. In 2017, the notification rate for all age groups was at least 1.4 times the rate in 2010–2016. The age of case patients increased over time, with a significantly higher proportion of case patients aged ≥10 than <10 years in 2017 compared to 2010 (P < .001) (Fig 4). The median age of case patients increased from 3.9 years in 2010 to 7.1 years in 2017.

In 2017, 430 of the 2319 notified specimens were sent to Murdoch Children’s Research Institute for genotypic testing. Of the 386 that were successfully genotyped, 186 (44.3%) were children aged <5 years and 200 (55.7%) were individuals aged ≥5 years. The RV1 vaccine strain G1P[8] was identified in 27 of 38 (71%) infants aged ≤6 months. In other age groups (except those aged ≥80 years), equinelike G3P[8] was the dominant strain, and G8P[8] was the second most common (Fig 5). The G8P[8] strain predominated in the elderly aged ≥80 years (41.5%).

There were 3587 children with notified rotavirus born on or after January 1, 2008 (37.7% of all notified cases), excluding children aged <6 months. As shown in Fig 1, of those included in the VE analysis, 214 children received 1 RV1 dose, 2407 received ≥2 RV1 doses, and 966 had no reported RV1 doses received. The median age at rotavirus notification was 25 months (range: 6 months–9 years). There is some evidence that boys and men were more likely to have a notification for rotavirus (P = .008); however, for those children included in the VE model, sex did not change the estimate by >5% and hence was not included in the final VE model.

The overall VE for the years 2010–2017 was estimated as 68.0% and 73.7% for 1 dose and 2 doses, respectively. In 2017 alone, the VE was significantly higher at 83.0% and 82.5% for 1 dose and 2 doses, respectively. The 2017 VE was higher than the VE in all previous individual years, with the exception of 2010, which had the highest 2-dose VE of 89.7% (95% CI: 86.5% to 92.0%) for the cohort of children aged 6 months to 2 years (Table 1, Supplemental Table 3).

The 2-dose VE estimates were highest among infants aged 6 to 11 months in all study years, except in 2011 when there were low case numbers. The VE point estimates for children aged 6 months to 3 years were higher than for children aged 4 to 9 years; however, there was substantial overlap in CIs.

As assessed in 2017, 2-dose VE estimates declined with time since vaccination from 89.5% within 1 year of vaccination to a significantly lower VE 5 to <10 years after vaccination of 77.0% (Table 2).

The 2017 rotavirus outbreak in New South Wales, Australia, resulted in an increase in notification rates across all age groups. The majority of cases were in children aged <4 years, particularly those aged 1 to 2 years. As seen in this outbreak, young children are well recognized as the primary population group affected by rotavirus.²⁵  However, our study also revealed that the median age of rotavirus cases has increased in Australia over the last 8 years from 3.9 years in 2010 to 7.1 years in 2017. Adults and older children born before the availability of vaccination in Australia are unimmunized and may have been less likely to have repeated subclinical infections because of reductions in virus circulation overall, resulting in less immune boosting.²⁶  The increasing proportion of cases among adults, as well as increased notifications in 2017, may be influenced by vaccine introduction and also by a number of other factors such as increased recognition of rotavirus disease in older age groups and the increasing availability and use of enteric multiplex polymerase chain reaction panels in Australia, which has increased the likelihood of primary care physicians ordering fecal testing.²⁷ 

In our genotypic subanalysis of 386 specimens collected in 2017 in New South Wales, the equinelike G3P[8] strain was dominant, representing almost half of all rotavirus strains identified. This genotype was previously reported as a dominant rotavirus strain among Australian children in 2013 and 2014.^3,28 Equinelike G3P[8] has similarly been observed in international settings where RV1 is primarily used, albeit on the private market with lower national coverage, including in Thailand (2013) and Hungary (2015), but has also been observed in regions where RV5 is primarily used (Spain; 2015).²⁹  G8P[8] was identified as the second most dominant genotype in the 2017 New South Wales outbreak. G8 strains are rarely detected outside of Africa and had previously only been identified in 12 specimens positive for rotavirus between 1995 and 2015 in Australia.³⁰  This outbreak is the first time a G8 rotavirus strain has been the major genotype in the 20 years of rotavirus surveillance in Australia.³  G8P[8] is also emerging in several Asian countries including Japan, Vietnam, and Thailand.³¹  It is possible that vaccine-related selective pressure, in the context of high vaccine coverage, may have contributed to the emergence of G8 and other unusual genotypes.³  However, a systematic review of RV1 (and RV5) revealing similar effectiveness against homotypic and heterotypic rotavirus strains, as well as a lack of persistence of specific strains, has been suggested as evidence of absence of vaccine-induced selective pressure.³²  Unfortunately, we were not able to conduct a genotype-specific VE analysis in this study because of the anonymization of the subsample of specimens genotyped. However, the high VE seen in infants in 2017 suggests good short-term protection against these novel strains by RV1, albeit in the context of higher notification rates. Future studies in New South Wales using data linkage to merge information on genotyping, laboratory notification, as well as hospitalization data could provide more insight into vaccine performance.

Notably, the RV1 vaccine strain was identified in 27 of 38 infants aged ≤6 months who had genotypic analysis conducted but not in any older case patients. This was not unexpected and may represent incidental detection of excreted vaccine virus in stools postvaccination due to the highly sensitive RT-PCR testing, with another cause for diarrhea being present given that rates of diarrhea in randomized controlled trials of RV1 vaccine did not differ greatly between vaccinated and unvaccinated infants.³³  These notifications would artificially reduce VE estimates; therefore, case patients aged <6 months were excluded from our VE analyses. From July 1, 2018, the national notifiable case definition for rotavirus changed to reflect that recently vaccinated case patients should be excluded, unless confirmed to have infection with wild-type rotavirus. However, because most routinely used diagnostic tests do not differentiate between wild-type and vaccine virus and most notified case patients do not have samples genotyped, many notified cases in infants aged <6 months would appear to be related to vaccine virus shedding.¹⁷ 

In our study, 2 doses of RV1 was 73.7% effective in protecting children aged 6 months to 9 years against laboratory-confirmed rotavirus over our 8-year study period. Somewhat surprisingly in the 2017 outbreak year, a high 2-dose VE of 88.4% in those aged 6 to 11 months was also observed. Our VE estimates are similar to findings from a Queensland study (2007–2008) of RV5 but are much higher than RV1 estimates during a rotavirus outbreak among indigenous children in Central Australia (2009–2010).^6,11 Rotavirus vaccines have been shown to have reduced efficacy in resource-poor settings such as Bolivia, Malawi, and Brazil; Indigenous Australians experience similar patterns of diarrheal disease morbidity seen in less developed countries.^6,34 An observational study in the United States found a similarly high 2-dose RV1 VE of 83% against laboratory-confirmed rotavirus and a European randomized controlled trial showed the same vaccine prevented 90% of severe rotavirus gastroenteritis episodes.^35,36 

We have shown that RV1 effectiveness decreased with age in all calendar years in which adequate case numbers were reported, indicating potential waning of vaccine protection against laboratory-confirmed rotavirus infection. RV1 clinical trials of 20 000 participants determined 85% VE against severe rotavirus gastroenteritis, which reduced in the second year of life to 79%.^34,37 This is consistent with our results revealing the highest VE in children aged 6 to 11 months. Authors of a Brazilian observational study (2006–2008) indicated possible waning of RV1 protection against hospitalized rotavirus, with a lower VE in children aged 12 to 36 months compared with infants aged 6 to 11 months.³⁸  Authors of an observational study in the United States concluded RV1 to be most effective in the second year of life, and effectiveness decreases in the third year of life.³⁵  Our analysis by time since vaccination provided further evidence of waning vaccine protection. VE was generally highest in the year postvaccination and significantly decreased by 12.5% in the 5 to <10 years postvaccination. There are limited data on the duration of protection of rotavirus vaccines; in the majority of randomized controlled trials, researchers do not look past the first 2 years of life.^34,36 Waning rotavirus vaccine protection could occur because of the absence of natural exposure to the virus over time, whereby sequential infections (usually minimally or asymptomatic) boost immunity to both homotypic and heterotypic viruses. Vaccination of older children or high-risk adults has not been considered to date.³⁹  The reduced likelihood of developing significant symptomatic disease with increasing age, notwithstanding increased recognition of disease outbreaks in the elderly, is thought to reduce the beneficial effectiveness and potential disease prevention from vaccination.

There were a number of limitations to our study. Because we sourced deidentified controls from the AIR, there is the potential for misclassification bias because of the inability to definitively exclude case patients from controls selected from the register. If controls were included, the OR would be truly equivalent to the relative risk rather than an approximation, making the results more robust and accurate. Other limitations of the case-control VE analysis include the previously noted low completion of data fields such as indigenous status and the lack of socioeconomic data or other variables available for adjustment in this model. Additionally, absence of data on severity or hospitalization should be noted and may have resulted in lower VE estimates given that vaccination may be better at preventing severe disease. A strength of the case-control VE analysis was that the AIR was used as the source of vaccination data for both case patients and controls, and vaccination status was classified in the same way for both groups. Because of the variations in disease severity, health care seeking behavior, and diagnostic testing, the rotavirus notifications reported here underrepresent the total number of rotavirus cases that have occurred in the population. Notifications and genotyped specimens are limited to those who sought care during infection and were more likely to have suffered moderate to severe disease.

This analysis has revealed high effectiveness of RV1 vaccine over an 8-year period in an Australian population with high vaccine uptake. Although there is evidence of waning immunity with age, effectiveness remains relatively high over time, even in the context of a large outbreak from 2 novel strains, equinelike G3P[8] and G8P[8]. VE was not reduced in 2017 and thus does not appear to have contributed to the unexpectedly high rotavirus notification rate. However, more analysis is required to investigate how novel or unusual strains (such as the equinelike G3, G8, and G12) interact with rotavirus vaccines and whether antigenic changes affect VE and challenge vaccination programs.^39,40  Investigation of population-level VE in relation to rotavirus genotype data should continue in a range of settings to improve our understanding of rotavirus vaccines and the impact they have on disease across the age spectrum over time.

AIR

Australian Immunisation Register

CI

confidence interval

OR

odds ratio

RT-PCR

reverse transcriptase polymerase chain reaction

RV1

Rotarix

RV5

RotaTeq

VE

vaccine effectiveness