Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract This study defined the genetic epidemiology of dengue viruses (DENV) in two pivotal phase III trials of the tetravalent dengue vaccine, CYD-TDV, and thereby enabled virus genotype-specific estimates of vaccine efficacy (VE). Envelope gene sequences (n = 661) from 11 DENV genotypes in 10 endemic countries provided a contemporaneous global snapshot of DENV population genetics and revealed high amino acid identity between the E genes of vaccine strains and wild-type viruses from trial participants, including at epitope sites targeted by virus neutralising human monoclonal antibodies. Post-hoc analysis of all CYD14/15 trial participants revealed a statistically significant genotype-level VE association within DENV-4, where efficacy was lowest against genotype I. In subgroup analysis of trial participants age 9–16 years, VE estimates appeared more balanced within each serotype, suggesting that genotype-level heterogeneity may be limited in older children. Post-licensure surveillance is needed to monitor vaccine performance against the backdrop of DENV sequence diversity and evolution. https://doi.org/10.7554/eLife.24196.001 eLife digest Each year, about 100 million people—mostly children in tropical parts of Asia and Latin America—are infected with the dengue virus. It has been difficult to produce a vaccine against the virus, because there are four different types of the virus, and people respond to infections with different types in an unusual way. Once a person is infected with one type of dengue, they are protected from future infections with that type. However, if that person later becomes infected with a different type, they are more likely to experience severe illness. As a result, a dengue vaccine must simultaneously protect against all four types of the virus to be safe and effective. The first dengue vaccine has recently become available. Clinical studies of the vaccine show that it can protect against all four virus types, but that the protection against certain types and in some age groups varies. Complicating matters, the four types of the dengue virus have continued to evolve since scientists first began developing the vaccine. Therefore, scientists are concerned that the vaccine may not be as effective against the newly evolved subtypes. To find out, scientists would have to carefully compare the genetics of the strains used to develop the vaccine with the strains currently circulating. They would also have to see how well the vaccine protects against current strains. Now, Rabaa et al. show that there is a high level of genetic similarity between the viruses used to create the vaccine, and dengue viruses that caused infections in people participating in clinical studies of the vaccines. The analyses also showed that in children between the ages of 2 and 16, the vaccine is more effective against one subtype of the dengue type-4, compared to the other circulating subtype. In children between the ages of 9 and 16, who are eligible to receive the vaccine in some countries, the vaccine was largely equally effective across the various subtypes. In addition to providing reassurance that the vaccine is working against currently circulating types, Rabaa et al. provide a valuable snapshot of the genetic diversity of dengue viruses. This snapshot will help scientists develop more effective dengue vaccines and treatments. More studies following vaccinated people are needed to ensure that the current vaccine remains effective as circulating strains of the virus evolve. https://doi.org/10.7554/eLife.24196.002 Introduction Dengue is the commonest arboviral disease of humans and has been a major public health problem in tropical Asia and Latin America for decades (Stanaway et al., 2016). Reducing the population of competent mosquito vectors of dengue viruses has been the central aim of disease control efforts, but these have had little success in eliminating or stopping the spread of dengue globally. Effective dengue vaccines will be essential tools to achieving dengue control. Accordingly, the licensure of the first tetravalent dengue vaccine (chimeric yellow fever–dengue virus tetravalent dengue vaccine (CYD-TDV), Sanofi Pasteur) together with recommendations from The World Health Organisation’s Strategic Advisory Group of Experts (SAGE) on Immunization on its use in highly endemic countries, has provided the first prospects of an integrated public health approach to disease control (WHO, 2016). Dengue vaccine development has been challenging, in part because dengue viruses (DENV) exist as four phylogenetically and antigenically distinct serotypes (DENV-1 to −4). Within each virus serotype exists considerable genetic diversity at local, national and continental scales (Holmes, 2008). Subtle antigenic differences can also be measured amongst members of the same virus serotype and are speculated to be of epidemiological and clinical importance (Katzelnick et al., 2015). The virus population dynamics of DENV in hyperendemic areas are complex, often involving the emergence and extinction of viral lineages against a backdrop of multiple virus types co-circulating and oscillating in their relative prevalence. Human population immunity and intrinsic virus fitness in mosquitoes and humans are potential drivers of DENV evolution in these settings (Holmes and Burch, 2000). Acting to balance high mutation rates of DENV within individual hosts, the vector-human transmission cycle subjects viral populations to strong purifying selection, whereby emergent virus variants that are less fit for disseminated infection of both humans and mosquitoes are lost from the viral population (Holmes, 2003). CYD-TDV was found to be safe and efficacious for use in children 9 years of age and older, with efficacy varying according to age, baseline serostatus and virus serotype (Capeding et al., 2014; Villar et al., 2015). Furthermore, a trend toward reduced efficacy against DENV-2 was observed in the Asian phase III trial compared to the Latin American trial (Hadinegoro et al., 2015). This finding suggested that the efficacy of CYD-TDV might be affected by sub-serotype (i.e. genotype) level diversity in DENV populations, often associated with geographical boundaries. Beyond the epidemiological factors identified in previous studies of CYD-TDV efficacy, the performance of dengue vaccines could also be influenced by the evolving nature of DENV populations in endemic settings. For example, the possibility that circulating DENV populations could ‘escape’ vaccine-elicited immune responses was nominated as one of several possible explanations for the relatively low efficacy of CYD-TDV against DENV-2 in a phase IIb trial in Thailand (Sabchareon et al., 2012). Two phase III efficacy trials of CYD-TDV, involving more than 31,000 children between the ages of 2–14 years in the Asia–Pacific region (CYD14 trial) and between the ages of 9–16 years in Latin America (CYD15 trial) (Hadinegoro et al., 2015) enable, for the first time, a post hoc investigation of vaccine efficacy versus DENV population diversity. Thus, the aims of the present study were threefold. First, to document the genetic distance between the components of the CYD-TDV formulation and the DENV strains detected amongst cases in the CYD14 and CYD15 trials. Second, to perform focused analysis of the level of sequence conservation between CYD-TDV vaccine strains and wild-type DENV at epitope locations targeted by potent virus neutralising human monoclonal antibodies (mAbs). Lastly, we aimed to explore if a more complex genotype-specific efficacy pattern existed in the CYD14 and CYD15 trials, notwithstanding the limitations inherent to post hoc analysis. Collectively, these data provide insights into the characteristics of the CYD-TDV product relative to contemporary DENV populations and provide preliminary insight into genotype-level vaccine efficacy that can serve as a baseline for future research. Results Acquisition of DENV envelope gene sequences 433 acute serum samples from 595 virologically-confirmed dengue (VCD) cases in CYD14 and 512 samples from 662 VCD cases in CYD15 were selected for investigation on the basis of subject consent, viremia level and sample volume considerations (Figure 1A and B, respectively). From CYD14, 314 complete DENV envelope (E) gene nucleotide sequences (1485 nt for DENV-1,–2, −4; 1479 nt for DENV-3) were acquired directly from 433 serum samples (72.5%, including three mixed infections), with a subset of 299/433 (69.1%) samples also having a complete premembrane (prM) nucleotide sequence. From CYD15, 333 complete DENV E gene nucleotide sequences were acquired directly from 512 serum samples (65.0%, including eight mixed infections), with a subset of 313/512 (61.1%) samples also having a complete prM nucleotide sequence. The proportion of serum samples that yielded an E gene sequence was similar between control and dengue vaccine recipients within each study (Supplementary file 1a). The probability of acquiring an E gene sequence from serum samples was positively associated with the DENV viremia level (Figure 1—figure supplement 1). Figure 1 with 1 supplement see all Download asset Open asset Sequencing flow chart for samples obtained in CYD-TDV trials. (A) CYD14, (B) CYD15. https://doi.org/10.7554/eLife.24196.003 Figure 1—source data 1 Sequence alignment of DENV-1 prM and E genes from CYD-TDV trials. https://doi.org/10.7554/eLife.24196.005 Download elife-24196-fig1-data1-v1.fasta Figure 1—source data 2 Sequence alignment of DENV-2 prM and E genes from CYD-TDV trials. https://doi.org/10.7554/eLife.24196.006 Download elife-24196-fig1-data2-v1.fasta Figure 1—source data 3 Sequence alignment of DENV-3 prM and E genes from CYD-TDV trials. https://doi.org/10.7554/eLife.24196.007 Download elife-24196-fig1-data3-v1.fasta Figure 1—source data 4 Sequence alignment of DENV-4 prM and E genes from CYD-TDV trials. https://doi.org/10.7554/eLife.24196.008 Download elife-24196-fig1-data4-v1.fasta Phylogenetic profile of CYD-TDV vaccine strains and DENV detected in CYD14 and CYD15 trials Full and partial E gene sequences determined directly from serum samples collected in CYD14 and CYD15 trials (253 DENV-1, 191 DENV-2, 107 DENV-3 and 110 DENV-4) were aligned with E gene sequences corresponding to the CYD-TDV vaccine strains and sequences from GenBank for which the year and country of sampling were known. Maximum likelihood trees representing subsampled E gene sequence datasets allowed the classification of CYD14/15 viruses to the major intra-serotype lineages (genotypes) previously described for DENV (Figure 2—figure supplements 1–4). At the country level, CYD14/15 viruses were closely related to publicly available DENV sequences acquired from the same country, an indicator of ongoing local evolution. Figure 2 shows the genotypes detected in the CYD14/15 virus populations according to their country of sampling. Collectively, these data define the population genetics of viruses responsible for dengue cases in the CYD14/15 trials and provide a unique contemporaneous snapshot of DENV diversity in ten endemic countries. Figure 2 with 4 supplements see all Download asset Open asset Distribution of DENV serotypes and genotypes sequenced in CYD14 and CYD15 vaccine trials by country. Numbers in parentheses indicate the total number of samples of each genotype for which complete or partial E gene sequences were obtained. https://doi.org/10.7554/eLife.24196.009 Sequence differences between CYD-TDV vaccine strains and circulating wild-type viruses We quantified the differences between the E gene amino acid sequences in the components of the tetravalent CYD-TDV formulation and viruses from VCD cases in the CYD14 and CYD15 trials. The mean level of E gene amino acid sequence difference between vaccine strains and viruses from VCD cases in CYD14 and CYD15 was <3% for all serotypes (Figure 3 and Supplementary file 1b). To define the nature of these sequence differences, the amino acid positions that varied between CYD-TDV vaccine strains and the E gene sequences sampled in CYD14/15 trials and in the subsampled GenBank sequences were annotated adjacent to the subsampled maximum likelihood phylogenetic trees for each serotype. The DENV-2 E gene phylogeny (incorporating the vaccine strain) of relevance to the CYD14 trial is shown in Figure 4A and for CYD15 in Figure 4B. The equivalent annotated phylogenies for DENV-1,–3 and −4 are shown in Figure 4—figure supplements 1–6. These data reveal that positions of amino acid non-identity between CYD-TDV vaccine strains and wild-type viruses were dispersed across the E protein and do not cluster to any particular structural domain. Figure 3 with 1 supplement see all Download asset Open asset Average genotype-specific amino acid identity of DENV isolated in CYD-TDV trials compared to the vaccine strain of the corresponding DENV serotype. Black bars indicate the IQR of the full sample set. Coloured dots show the geographic regions from which each genotype was collected – red: CYD14, maritime SE Asia; blue: CYD14, mainland SE Asia; grey: CYD15, Americas. Black dots indicate the genotype of the serotype-specific CYD-TDV vaccine component. https://doi.org/10.7554/eLife.24196.014 Figure 4 with 6 supplements see all Download asset Open asset Amino acid differences between the DENV-2 E gene vaccine sequence, DENV-2 viruses isolated in CYD14 and CYD15 vaccine trials, and representative subsets of publically available DENV-2 sequences from the vaccine trial sites. (A) CYD14 DENV-2 phylogeny, (B) CYD15 DENV-2 phylogeny. Coloured tips on the trees show sequences isolated in the CYD-TDV trials (country of origin coloured as indicated in the key) and the vaccine sequence (purple star); grey tips indicate publicly available sequences isolated from other studies in the countries of interest. Columns to the right indicate amino acid sites at which variation was observed in two or more CYD14/CYD15 sequences. Numbers at the top of columns indicate the amino acid site within the E gene. Bars at the top of the figures indicate the E gene domain of the site. Amino acids at variable sites in the E gene sequence of the vaccine component are shown in colour. For all other sequences, a lack of colour indicates an amino acid identical to that of the vaccine component at that site. https://doi.org/10.7554/eLife.24196.016 Human mAb epitope sequences in vaccine and wild-type viruses We examined amino acid sequence identity between vaccine strains and wild-type CYD14/15 virus sequences at twelve B cell epitopes. The twelve epitopes represent some of the best structurally defined epitopes in DENV that are targeted by potent virus neutralising human mAbs and are thus of particular interest in vaccine development and immune correlate assays (Fibriansah et al., 2014; Cockburn et al., 2012a; Smith et al., 2013; Fibriansah et al., 2015a, 2015b; Teoh et al., 2012; Rouvinski et al., 2015; Costin et al., 2013; Cockburn et al., 2012b). Sequence analyses indicated limited variation at these epitope regions in the CYD14/15 sequences, as well as in a global database of wild-type virus sequences (Figure 5 and Figure 5—figure supplement 1). The conservation of these epitope sequences between the decades-old ‘donor’ viruses from which the CYD-TDV product was derived and contemporary virus populations suggests that these amino acid sites are not highly prone to evolutionary drift. Figure 5 with 1 supplement see all Download asset Open asset Sequence conservation between the DENV-2 vaccine component and wild-type DENV-2 viruses at epitope locations targeted by virus neutralising human mAbs. Amino acid targets for five neutralising human mAbs (Fibriansah et al., 2015b; Rouvinski et al., 2015) are coloured as indicated in the key (top) and compared to the vaccine sequence and wild-type sequences obtained within the CYD14 and CYD15 trials (middle), as well as complete E gene sequences from wild-type DENV-2 available on GenBank (bottom). Sites are indicated at the top of columns. For wild-type virus populations, the darker the block, the greater the proportion of sequences with an amino acid differing from the target amino acid at that site. When disagreement between amino acids was observed between epitope targets (as at E67 and E71), wild-type sequences were compared to 2D22 as a reference, denoted by an asterisk. https://doi.org/10.7554/eLife.24196.023 Vaccine efficacy by DENV serotype and genotype Given the high degree of overall amino acid sequence identity, including at key epitope positions, between the E protein found in CYD-TDV vaccine strains and contemporary wild-type CYD14/15 viruses, we postulated that vaccine efficacy would be largely independent of virus genotype. We report two levels of intention to treat genotype-level efficacy from the CYD14 and CYD15 trials: the observed estimates and the observed+imputed estimates. The observed estimate refers to vaccine efficacy in the population of VCD cases who had serum samples yielding an E gene sequence that was empirically assigned a genotype. The observed+imputed estimates used the observed genotype data plus imputation to give genotype assignments to VCD cases where the serotype was known but genotype information was absent. Imputation was likely to be accurate because data from this study (Figure 2) indicated eight out of the ten study countries had only a single genotype of each serotype in circulation during the study period. Publicly available sequence data largely mirror the genotype distributions observed in this study; greater diversity is found in some Asian countries relative to those detected in this study, likely because the publicly available sequences are collated at the country level, whereas the CYD14/15 sequences represent those circulating only within the geographically limited trial populations (Supplementary file 1c). The count of observed and imputed genotypes is summarised in Supplementary file 1d. Estimates of genotype-level vaccine efficacy amongst the observed and observed+imputed case populations are described in Table 1 (all ages) and Table 2 (participants 9–16 years of age). For completeness, we also show the observed genotype-level vaccine efficacy for participants < 9 years of age in Supplementary file 2 but do not consider it in the main analyses because this age-class was only present in the CYD14 trial and is below the age for which the licensed vaccine is now indicated (i.e. ≥ 9 years) (WHO, 2016). For each serotype, a Cox proportional hazards regression model (expressing the hazard function) was used to estimate vaccine efficacy (derived as 100* [1- Hazard Ratio]) with vaccine group, genotype and the interaction between vaccine group and genotype included as covariates. The parameter estimates and the 95% confidence intervals of the interactions are given in Table 3 (all ages) and Table 4 (participants 9–16 years of age). Table 1 Observed and imputed efficacy of CYD-TDV in all participants who received ≥1 injection (intention to treat) by serotype and genotype. https://doi.org/10.7554/eLife.24196.025 CYD dengue vaccine groupControl groupVaccine efficacy ObservedVaccine Efficacy with imputation for missing genotype dataCasesPerson-years at riskDensity incidence (95% CI)CasesPerson-years at riskDensity incidence (95% CI)%(95% CI)%(95% CI)Serotype 163.1(52.7; 71.2)54.7(45.4; 62.3)Genotype I CYD14CYD15137420.1 (0.1; 0.2)1867960.3 (0.2; 0.4)58.8(18.3; 79.5)57.4(29.7; 74.2)Genotype IV CYD1440137420.3 (0.2; 0.4)5167960.8 (0.6; 1.0)61.3(41.5; 74.5)53.3(37.2; 65.3)Genotype V CYD1553270160.2 (0.1; 0.3)76134340.6 (0.4; 0.7)65.3(50.9; 75.7)54.9(40.7; 65.6)p-value*0.86140.9912Serotype 239.1(18.9; 54.3)43.0(29.4; 53.9)American/Asian CYD1548270350.2 (0.1; 0.2)50134610.4 (0.3; 0.5)52.2(28.9; 67.9)50.2(32.6; 63.2)Asian I CYD14CYD28137660.2 (0.1; 0.3)1468560.2 (0.1; 0.3)0.3(−94.9; 46.6)19.8(−30.0; 49.6)Cosmopolitan CYD1428137660.2 (0.1; 0.3)2168560.3 (0.2; 0.5)33.8(−18.0; 62.2)43.8(16.1; 62.2)p-value*0.14690.2493Serotype 375.1(62.9; 83.3)71.6(63.0; 78.3)Genotype I CYD14913835<0.1 (0.0; 0.1)1468950.2 (0.1; 0.3)67.9(26.9; 86.6)58.1(25.2; 76.8)Genotype II CYD14CYD0138350.0 (0.0; 0.0)46895<0.1 (0.0; 0.1)100.0(69.3; 100.0)85.8(41.1; 97.9)Genotype III CYD14413835<0.1 (0.0; 0.1)768950.1 (0.0; 0.2)71.6(6.1; 92.6)68.4(19.8; 88.4)Genotype III CYD152327060<0.1 (0.1; 0.1)47134590.3 (0.3; 0.5)75.7(60.5; 85.5)74.2(64.3; 81.4)Genotype III CYD14 + CYD152740896<0.1 (0.0; 0.1)54203540.3 (0.2; 0.3)75.2(61.0; 84.6)73.7(64.3; 80.8)p-value*0.37510.2561Serotype 474.1(61.7; 82.5)76.9(69.5; 82.6)Genotype I CYD1419138260.1 (0.1; 0.2)1868740.3 (0.2; 0.4)47.4(−0.9; 72.5)58.3(29.9; 75.2)Genotype II CYD14CYD813826<0.1 (0.0; 0.1)2468740.3 (0.2; 0.5)83.5(64.8; 93.1)83.8(69.3; 91.5)Genotype II CYD15CYD1127063<0.1 (0.0; 0.1)31134420.2 (0.2; 0.3)82.4(66.0; 91.5)80.8(71.2; 87.3)Genotype II CYD14 + CYD15CYD1940890<0.1 (0.0; 0.1)55203160.3 (0.2; 0.4)82.9(71.7; 90.1)81.8(74.3; 87.1)p-value*0.00720.0086 Cases: number of subjects with at least one sequenced symptomatic virologically-confirmed dengue episode during the active phase of follow-up. Density incidence: data indicate cases per 100 person-years at risk. *The p-value was obtained by testing the heterogeneity of genotype distribution between groups (within each serotype) using a Chi2 (or Fisher’s exact test). CYD Genotype of the serotype-specific CYD-TDV vaccine component. Table 2 Observed and imputed efficacy of CYD-TDV for subjects 9 years and older who received ≥1 injection (intention to treat) by serotype and genotype https://doi.org/10.7554/eLife.24196.026 CYD dengue vaccine groupControl groupVaccine efficacy ObservedVaccine Efficacy with imputation for missing genotype dataCasesPerson-years at riskDensity incidence (95% CI)CasesPerson-years at riskDensity incidence (95% CI)%(95% CI)%(95% CI)Serotype 167.7(56.1; 76.3)58.4(47.7; 66.9)Genotype I CYD14CYD66683<0.1 (0.0; 0.2)833060.2 (0.1; 0.5)62.8(−6.8; 87.8)69.0(33.8; 85.5)Genotype IV CYD14866830.1 (0.1; 0.2)1933060.6 (0.3; 0.9)79.2(54.1; 91.4)64.0(39.7; 78.5)Genotype V CYD1553270160.2 (0.1; 0.3)76134340.6 (0.4; 0.7)65.3(50.9; 75.7)54.9(40.7; 65.6)p-value*0.52130.5400Serotype 248.6(27.4; 63.7)47.1(31.3; 59.2)American/Asian CYD1548270350.2 (0.1; 0.2)50134610.4 (0.3; 0.5)52.2(28.9; 67.9)50.2(32.6; 63.2)Asian I CYD14CYD1266870.2 (0.1; 0.3)933300.3 (0.1; 0.5)33.6(−62.7; 71.9)34.6(−27.4; 65.7)Cosmopolitan CYD1456687<0.1 (0.0; 0.2)433300.1 (0.0; 0.3)37.8(−151; 83.5)40.3(−41.4; 74.3)p-value*0.77360.7253Serotype 376.0(62.3; 84.7)73.6(64.4; 80.4)Genotype I CYD1446715<0.1 (0.0; 0.2)633470.2 (0.1; 0.4)66.8(−16.3; 91.5)61.2(−4.1; 86.1)Genotype II CYD14CYD067150.0 (0.0; 0.1)33347<0.1 (0.0; 0.3)100.0(55.4; 100.0)80.1(7.6; 97.1)Genotype III CYD1416715<0.1 (0.0; 0.1)23347<0.1 (0.0; 0.2)75.1(−160; 98.8)75.1(−27.4; 96.6)Genotype III CYD152327060<0.1 (0.1; 0.1)47134590.3 (0.3; 0.5)75.7(60.5; 85.5)74.2(64.3; 81.4)Genotype III CYD14 + CYD152433775<0.1 (0.0; 0.1)49168060.3 (0.2; 0.4)75.7(60.8; 85.3)74.3(64.7; 81.4)p-value*0.59280.6985Serotype 485.2(74.6; 91.4)83.2(76.2; 88.2)Genotype I CYD1436716<0.1 (0.0; 0.1)1233270.4 (0.2; 0.6)87.6(60.9; 97.2)86.2(63.6; 94.8)Genotype II CYD14CYD36716<0.1 (0.0; 0.1)1433270.4 (0.2; 0.7)89.4(67.7; 97.6)89.6(70.5; 96.3)Genotype II CYD15CYD1127063<0.1 (0.0; 0.1)31134420.2 (0.2; 0.3)82.4(66.0; 91.5)80.8(71.2; 87.3)Genotype II CYD14 + CYD15CYD1433779<0.1 (0.0; 0.1)45167690.3 (0.2; 0.4)84.6(72.6; 91.8)82.6(74.7; 88.1)p-value*1.00000.6678 Cases: number of subjects with at least one sequenced symptomatic virologically-confirmed dengue episode during the active phase of follow-up. Density incidence: data indicate cases per 100 person-years at risk. *The p-value was obtained by testing the heterogeneity of genotype distribution between groups (within each serotype) using a Chi2 (or Fisher’s exact test). CYD Genotype of the serotype-specific CYD-TDV vaccine component. Table 3 Estimation of the interaction between genotype and vaccine group for symptomatic VCD detected during the active phase of follow-up by serotype in all participants who received >= 1 injection (intention to treat) (CYD14/CYD15). The estimate of the interaction term between genotype and vaccine group is derived from Cox proportional hazards regression models including the vaccine group, the genotype and the interaction. https://doi.org/10.7554/eLife.24196.027 Estimated interaction with observed vaccine efficacyEstimated interaction with vaccine efficacy with imputationSerotypeParameterParameter estimate95%Parameter estimate95% Serotype 1Genotype IV vs Genotype I−0.058[−0.858; 0.743]0.095[−0.475; 0.665]Genotype V vs Genotype I−0.167[−0.936; 0.603]0.067[−0.492; 0.625] Serotype 2American/Asian vs Asian I−0.732[−1.486; 0.022]−0.471[−1.032; 0.089]Cosmopolitan vs Asian I−0.404[−1.259; 0.451]−0.344[−0.966; 0.267] Serotype 3Genotype II vs Genotype I−12.748[−729.203; 703.707]−1.079[−2.754; 0.596]Genotype III vs Genotype I−0.251[−1.208; 0.705]−0.459[−1.116; 0.198] Serotype 4Genotype II vs Genotype I−1.114[−1.943; −0.285]−0.8184[−1.434; −0.203] Table 4 Estimation of the interaction between genotype and vaccine group for symptomatic VCD detected during the active phase of follow-up by serotype in subjects older than 9 years of age who received >= 1 injection (intention to treat) (CYD14/CYD15). The estimate of the interaction term between genotype and vaccine group is derived from Cox proportional hazards regression models including the vaccine group, the genotype and the interaction. https://doi.org/10.7554/eLife.24196.028 Estimated interaction with observed vaccine efficacyEstimated interaction with vaccine efficacy with imputationSerotypeParameterParameter estimate95%Parameter estimate95% Serotype 1Genotype IV vs Genotype I−0.574[−1.917; 0.768]0.153[−0.760; 1.066]Genotype V vs Genotype I−0.061[−1.177; 1.054]0.385[−0.416; 1.186] Serotype 2American/Asian vs Asian I−0.327[−1.277; 0.624]−0.270[−0.987; 0.448]Cosmopolitan vs Asian I−0.064[−1.637; 1.510]−0.089[−1.151; 0.972] Serotype 3Genotype II vs Genotype I−13.019[−943.634; 917.597]−0.664[−2.578; 1.250]Genotype III vs Genotype I−0.309[−1.665; 1.047]−0.405[−1.443; 0.633] Serotype 4Genotype II vs Genotype I0.226[−1.174; 1.626]0.244[−0.789; 1.277] For DENV-1, vaccine efficacy estimates against the three different genotypes were highly similar in the all ages group and in participants 9–16 years of age (Tables 1 and 2). Additionally, the genotype interaction parameter estimates in the all ages group (Table 3) were close to zero and had reasonably tight 95% confidence bounds. This suggests it is unlikely that an interaction exists between genotype and vaccine efficacy, but if such an interaction does exist, it is small. Amongst participants 9–16 years of age, the interaction parameter estimates had 95% confidence intervals that bounded zero and were wider than the all ages group, making conclusions relatively difficult to draw. For DENV-2, vaccine efficacy estimates against the American-Asian genotype (50.2%; 95% CI: 32.6–63.2%) and the Cosmopolitan genotype (43.8%; 95% CI: 16.1–62.2%) were similar, and both were higher than against the Asian I genotype (19.8%; 95% CI: −30.0–49.6%) in the all ages group (Table 1). However, the genotype interaction estimates had 95% confidence intervals that, although reasonably tight, included zero in the all ages group (Table 3) and in participants ≥ 9 years (Table 4). We note however that the upper bound of the confidence interval was very close to zero for the Asian/American versus Asian I genotype interaction (all ages, Table 3), leaving open the possibility that true heterogeneity may exist. Within DENV-3, the confidence intervals for the interaction estimates were very wide when comparing genotype II versus genotype I in the all ages population (Table 3) and in participants ≥ 9 years (Table 4), and thus no conclusions could be drawn from these data. For DENV-3 genotype III versus genotype I, the 95% confidence intervals around the interaction estimates (Tables 3 and 4) were much tighter but nonetheless passed through zero. This suggested an interaction between genotype and vaccine efficacy remained possible but unlikely, and that more data would be needed to address this question. Against DENV-4, vaccine efficacy was significantly lower against genotype I (58.3%, 95% CI: 29.9–75.2%), which circulates endemically only in Asia, compared to the globally distributed genotype II (81.8%, 95% CI: 74.3–87.1%, across CYD14/CYD15) in the all ages population (p=0.009) (Table 1). Confidence intervals around estimates of the interaction between genotype I and genotype II and vaccine group exclude zero, consistent with a lower efficacy against genotype I relative to genotype II (Table 3). However, when efficacy against