Impact of two-dose varicella vaccination in Guangzhou: An interrupted time-series study

ABSTRACT

In November 2017, Guangzhou implemented a self-paid two-dose varicella vaccination program. This study evaluated the program’s impact on varicella incidence trends across the general population and specific age groups. An interrupted time-series analysis was conducted using surveillance data from January 2011 to December 2024. The pre-intervention period spanned 2011–2017, and the post-intervention period 2018–2024. Seasonal-trend decomposition and segmented regression were employed to analyze changes in level and trend of varicella incidence. A total of 225,497 varicella cases were reported between 2011 and 2024. The annual incidence rate decreased by 28.1% from the pre-intervention to post-intervention period, from 126.80 to 91.14 per 100,000 population. Prior to intervention, there was no evidence of a monthly change in varicella incidence (0.3%; t = 2.27, P = .294). Significant upward trends were observed among 10–14-y-olds and ≥20-y-olds (0.6% monthly increase, both P < .01). Post-intervention, a significant monthly decrease of 1.8% was observed overall (t = -9.956, P < .001). The most pronounced reduction occurred in 4–6-y-olds (3.8% monthly decrease), followed by 7–9-y-olds (3.0%), 0–3-y-olds (2.3%), 10–14-y-olds (1.9%), ≥20-y-olds (1.1%), and 15–19-y-olds (0.5%). Varicella incidence exhibited consistent seasonal variation, with a primary peak occurring from October to January. The two-dose varicella vaccination program was associated with a significant reduction in varicella incidence across all age groups in Guangzhou, demonstrating both direct protection among vaccinated children and indirect herd effects in older populations. These findings support incorporating the two-dose varicella vaccine into routine national immunization programs.

Introduction

Varicella, or chickenpox, is a highly contagious infectious disease primarily affecting children.¹ The disease can occur in all seasons of the year, but is more prevalent in winter and spring.² Varicella can spread rapidly in collective gathering places like childcare institutions and schools, which can result in widespread transmission that proves challenging to control.³ Varicella outbreaks in schools can lead to significant days of schooling lost for students and substantially impact students’ physical well-being. Although most primary varicella cases present as self-limiting skin lesions (papules, vesicles, crusts), severe complications such as encephalitis or pneumonia can occasionally occur in high-risk populations, such as infants, adolescents, pregnant women, the immunocompromised, and unvaccinated individuals. The optimal strategy for safeguarding children at risk against varicella is immunization.

The varicella live attenuated vaccine (Oka strain) was developed in 1974 by Japanese scientist Dr Takahashi using the virus isolated from the vesicular fluid of a child with typical varicella⁴ and then used commercially in 1987.⁵ Since then, the Oka strain vaccine has been widely used worldwide, and China also introduced the varicella vaccine in 1997.⁶ China has not yet incorporated the varicella vaccine into its National Immunization Program (NIP). In most provinces, the varicella vaccine remains classified as a Category 2 vaccine, requiring voluntary self-paid vaccination. In recent years, however, several provinces and municipalities, including Qingdao, Tianjin, Shanghai, Nanjing, Suzhou, Shenzhen, and others, have integrated the varicella vaccine into their local immunization programs.⁷

Since 1998, Guangzhou has implemented a voluntary single-dose vaccination program using the live attenuated varicella vaccine (Oka strain), targeting children aged ≥1 y. Studies have demonstrated that single-dose administration of the varicella vaccine exhibits a favorable safety profile and robust immunogenicity.^8,9 However, long-term epidemiological surveillance has revealed a substantial incidence of breakthrough varicella (BV) cases (0.8% to 2.0%) following single-dose vaccination.¹⁰ A study showed that during the 10-y observation period, the risk of developing varicella >42 d postvaccination was 3.3-fold lower (P < .001) in children who received two injections than in those who received one injection (2.2% vs. 7.3%, respectively).¹¹ Considering the increasing trend in BV incidence and relatively high number of BV cases, the Advisory Committee on Immunization Practices in the United States recommended two routine doses of varicella vaccine in 2006.^5,12 Guangzhou adjusted the self-paid strategy of varicella vaccination from one dose to two doses in 2017. Children aged 12–24 months are recommended to receive the first dose of the varicella vaccine, and children aged 4–6 y are recommended to receive the second dose.¹³ However, some studies found no difference in clinical presentations between one-dose and two-dose recipients with breakthrough varicella.^14,15 Some studies have also pointed out that the research on the protective effect of the second dose varicella vaccine through methods such as retrospective cohort studies and analysis of serum antibody levels in the population is based on short-term observations, and the long-term trend of vaccine effectiveness is uncertain.¹⁶

Interrupted time series (ITS) is the strongest quasi-experimental design for analyzing the impact of intervention measures on time series, and it has been extensively applied to evaluate the causal impacts of social, public health, and environmental policies. In Guangzhou, there has been little empirical evidence about the impact of a two-dose varicella vaccination strategy at the population level. Therefore, our study aimed to investigate the effectiveness of a two-dose varicella vaccination strategy in controlling varicella, using the ITS, providing a reference for optimizing the strategy of varicella vaccine immunization.

Materials and methods

Data source

Varicella case data from 1 January 2011 to 31 December 2024 and corresponding population data were retrieved from the China Information System for Disease Control and Prevention (CISDCP). The extraction criteria were based on clinical onset dates and patients’ residential addresses.

This study was reviewed and approved by the Ethics Committee of the Guangzhou Center for Disease Control and Prevention (GZCDC-ECHR-2023P003) as the data were fully anonymized before being accessed for analysis and posed no more than minimal risk to the subjects.

Varicella vaccination strategy and surveillance in Guangzhou

Following the introduction of a self-paid one-dose varicella vaccination program for children aged 1–2 y in Guangzhou in 1998,¹⁷ a self-paid two-dose varicella vaccination program was implemented in November 2017. Varicella is not yet a notifiable infectious disease in China. Since 1995, varicella has been designated a Category C notifiable infectious disease in Guangzhou.^18,19 This classification mandates standardized case reporting and public health monitoring.

Statistical analysis

Descriptive study

The monthly varicella incidence data were organized, and a descriptive epidemiological study was conducted to characterize the varicella incidence in the general population of Guangzhou from 2011 to 2024. The analysis indicators included the seasonal index and the incidence rate. The seasonal index, defined as the average cases in a given month divided by the overall monthly mean cases from 2011 to 2024, was used to identify seasonal patterns. A value exceeding 1 indicated a high-incidence month, whereas values near or below 1 suggested absent or minimal seasonality.

The incidence rate was used as an index to evaluate the effect of intervention; the calculation method can be expressed as:

Incidence rate = (number of patients in the year/average population in the year) × 100,000.

Since Guangzhou changed its varicella vaccination program strategy in late 2017, we designate 2011–2017 as the pre-intervention period and 2018–2024 as the post-intervention period. The annual average incidence rate during the pre-intervention period (2011–2017) and the post-intervention incidence rate (2018–2024) were calculated using the formula, respectively:

Incidence = (Mean annual case count/Mean annual population) × 100,000.

ITS model building

ITS is an increasingly popular time-series analysis method that Box and Tiao proposed to assess the impact of intervention when randomization is impractical.²⁰ ITS can control for preexisting regression trends in outcomes before and after the intervention, while evaluating immediate changes in post-intervention levels. Additionally, ITS enables multidimensional temporal analysis by differentiating the intervention’s short-term (acute phase) and long-term (sustained phase) effects on outcome trajectories.

We employed an ITS analysis based on Quasi-Poisson Regression models to evaluate the impact of the two-dose varicella vaccination strategy implemented at the end of 2017. The analysis compared outcome measures between pre-intervention (2011–2017) and post-intervention (2018–2024) periods, specifically assessing:

1. Immediate-level change: Abrupt shift in outcome indicators at policy implementation.

2. Long-term slope change: Altered trend trajectories during the post-policy phase.

Model building

Guangzhou adjusted the varicella vaccination program to two doses in November 2017, so this study used 2018 as the time point for adjusting the vaccine immunization strategy. Eighty-four months from January 2011 to December 2017 was the pre-intervention period for the varicella strategy, and 84 months from January 2018 to December 2024 was the post-intervention period. The monthly reported incidence rate of varicella was used as the outcome variable to construct the interrupted time series segmented regression model. The model is shown below:

$$log{Y}_t={\beta }_0+{\beta }_1\mathit{T}+{\beta }_2{X}_t+{\beta }_3\mathit{T}{X}_t+{\beta }_{4}sin\frac{2\mathit{\pi t}}{12}+{\beta }_{5}cos\frac{2\mathit{\pi t}}{12}+{\beta }_{6}sin\frac{4\mathit{\pi t}}{12}+{\beta }_{7}cos\frac{4\mathit{\pi t}}{12}+log\mathit{N}$$

Where Y_t is the monthly incidence of varicella and T is the time count variable since the start of the study, T = 1,2, … … n, in this study T takes the value of 1–168; X_t is the intervention variable that takes the value of 0 before the intervention and 1 after the intervention; TX_t is the interaction term between time and intervention; N denotes either the overall study population or age-stratified subpopulations; β₀ is the level of the varicella incidence at the beginning; β₁ is the pre-intervention slope, indicating the pre-intervention trend of change; β₂ is the amount of change in level at the time of intervention, reflecting the short-term effect of the intervention; and β₃ is the difference between the pre- and post-intervention slopes, indicating the amount of change in slope or change in trend before and after the intervention, reflecting the long-term effect of the intervention²¹; β₁₊β₃ is the slope of the reported incidence of varicella after the intervention, means the monthly change after intervention. Due to the pronounced seasonality of infectious diseases such as varicella, two pairs of sine and cosine functions were incorporated into the model to improve fitting accuracy and control for seasonal effects. To ensure the validity of parameter estimates and the robustness of significance tests, the Newey–West method was applied to adjust for potential autocorrelation or heteroscedasticity, following the assessment of serial autocorrelation using the Durbin–Watson test.

In addition to the primary analysis encompassing all ages, we performed subgroup analyses by categorizing the population into six age groups (0–3, 4–6, 7–9, 10–14, 15–19, and 20 y and older). This division is based on the hypothesis that the intervention’s effectiveness would vary according to school enrollment status, influencing contact patterns, and age-specific vaccination coverage rates before the policy change.

Data analysis

WPS Office software (version 12.1.0.21171) was used for data collation and statistical descriptions. We performed all analyses in RStudio (version 2024.12.1). We compared annual incidence rates using Pearson’s chi-square test and defined a two-tailed p-value of <.05 as statistically significant.

Results

Basic information

There were 225,497 varicella cases reported between January 2011 and December 2024. The annual average reported incidence rate from 2011 to 2024 was 106.59 per 100,000. The pre-intervention (2011–2017) annual average reported incidence rate was 126.80 per 100,000, while the post-intervention (2018–2024) annual average reported incidence rate was 91.14 per 100,000. The annual incidence rate decreased by 28.1% from the pre-intervention to post-intervention period (from 126.80 to 91.14; (126.80 − 91.14)/126.80 × 100%). The reported incidence rates ranged from 53.75 per 10,000 (2022) to 190.29 per 10,000 (2017), peaking in 2017. The distribution of annual incidence rates differed significantly by year over the period 2011–2024 (χ² = 38662.74, P < .001) (Figure 1).

Figure 1.

Annual reported varicella incidence rate in Guangzhou, 2011–2024.

ITS analysis result

Interrupted time series regression model for the monthly incidence rate of varicella in the total population of Guangzhou

The ITS results indicated that the annual reported incidence rate of varicella in Guangzhou was 8.97 × 10⁻⁵ per 100,000 at the baseline of the time series. During the pre-intervention period, varicella incidence demonstrated relative stability, with a statistically nonsignificant increase of 0.3% per month (t = 2.217, P = .294). The post-intervention level change in varicella incidence in Guangzhou was 7.719 per 100,000 (t = 10.024, P < .001). The post-intervention slope change in varicella incidence rates in Guangzhou was −2.1% (t = −9.956, P < .001), demonstrating a statistically significant decrease in varicella incidence after intervention. The post-intervention incidence trend slope was calculated as −0.018 (β₁₊β₃ = −0.018), indicating the varicella incidence rate declined at a monthly rate of 1.8% in the post-intervention period. All seasonal coefficients (β₄ to β₇) demonstrated statistical significance (P < .05), indicating seasonal variations in varicella incidence (Table 1, Figure 2).

Table 1.

Regression model coefficients for Guangzhou’s total population.

Variable	Estimated value	t-Value	P-Value	exp(Est.)
intercept	−9.330	−124.272	<.001	8.97 × 10−5
β1	0.003	2.217	.294	1.003
β2	2.044	10.024	<.001	7.719
β3	−0.021	−9.956	<.001	0.979
β4	0.083	2.117	.086	1.086
β5	−0.109	−2.998	.001	0.897
β6	0.176	5.157	.001	1.193
β7	0.269	7.349	<.001	1.309

Figure 2.

Incidence and the trend of varicella in Guangzhou, 2011–2024.

Interrupted time series regression model for the monthly incidence rate of varicella in different age groups in Guangzhou

Analysis of varicella incidence trends among different age groups in Guangzhou before and after implementing the two-dose immunization strategy revealed distinct patterns. Prior to the intervention, the 10–14-y-olds (t = 4.073, P < .01) and ≥20-y-olds (t = 4.701, P < .01) exhibited significant upward trends, with both showing a monthly incidence increase of 0.6%. The remaining four age groups manifested no significant upward or downward trends. Post-intervention showed that all six age groups experienced an immediate increase in incidence levels post-intervention. Following the intervention, a significant downward trend in varicella incidence was seen in all six age groups, with the 4–6-y-olds exhibiting the fastest reduction in incidence rate, at 3.8% per month. Followed by the 7–9-y-olds with a monthly decrease of 3.0%, the 0–3-y-olds with 2.3%, the 10–14-y-olds with 1.9%, the ≥20-y-olds with 1.1%, and the 15–19-y-olds with 0.5% (Supplementary Table S1, Figure 3).

Figure 3.

The trends in the varicella incidence in different age groups from 2011 to 2024.

Seasonal index

From 2011 to 2024, varicella cases in Guangzhou exhibited a bimodal distribution, peaking primarily between October and January (accounting for 42.1% of all cases, n = 95,023) with a secondary peak in May. (Figure 4).

Figure 4.

Seasonal index of varicella in Guangzhou, 2011–2024.

Discussion

This study analyzed the changing trends in varicella incidence across all age groups and within specific age groups in Guangzhou by the two-dose varicella vaccination strategy. A significant decline in monthly varicella incidence was observed in five of the six age groups studied, with the most modest change seen in the 15–19 y cohort. This demonstrates that the intervention was effective at initiating a decrease even in groups where rates had previously stabilized, and at reversing the upward trend in groups with previously increasing rates. The two-dose vaccination schedule demonstrated significant long-term effectiveness across all age groups. Due to the vaccination age and antibody persistence duration, there are certain differences in the changes in incidence rates among different age groups after the implementation of the intervention.

In this study, a statistically significant change in slope (β₃) of varicella incidence was observed in every age group following the intervention, collectively indicating a positive long-term reduction trend. While statistically significant across the board (all P < .01), the magnitude of the reduction varied considerably, with β₃ values ranging from −0.036 to −0.009. The general population showed a highly significant estimate of −0.021 (P < .001). This finding aligns with the investigation reported by Lin et al.², Duan et al.²² Furthermore, Pan et al.²³ conducted a retrospective cohort study that included 107,324 local children, confirming the protective efficacy of receiving two doses of the varicella vaccine was significantly higher than that of a single dose. A real-world effectiveness study of varicella vaccination demonstrated that a single dose of varicella vaccine provides limited protection (46%), while two doses offer strong and lasting protection – up to 95% effectiveness – even when the doses are given years apart.²⁴ A varicella seropositivity study revealed that subjects who received two doses of the vaccine had a higher antibody level and positive rate than those who received one dose at any time since vaccination, and at any age.¹³ These findings confirm that implementing the two-dose varicella vaccination policy can effectively prevent and control varicella infection.

The findings from our interrupted time-series analysis provide compelling evidence for direct and indirect (herd) protection effects of the two-dose varicella vaccination program. The most pronounced direct effect was observed in the 4–6-y-old group, which exhibited the most rapid monthly decline in incidence (3.8%). This group is the target for the second vaccine dose, and this sharp reduction may be a direct consequence of increased immunization coverage and the resultant high level of individual protection. The 0–3-y-old group also exhibited a moderate decline in incidence. This age group is also age-eligible for the one-dose varicella vaccine before intervention. This likely explains why the reduction was less pronounced than in the 4–6-y-old group. The significant downward trends in age groups outside the second-dose recommendation strongly suggest the operation of indirect protection. The intervention indirectly lowered the risk of exposure and infection for adult, older, partially vaccinated, or completely unvaccinated individuals. Furthermore, the significant downward trends in the 7–9 and 10–14-y-old age groups, while strongly suggestive of indirect protection, should be interpreted with consideration of the aging of vaccinated cohorts. Children who received the two-dose vaccine in the 4–6-y-old group at the start of the intervention would have aged into these older groups during the post-intervention period. This demographic movement likely contributed, at least in part, to the observed declines in these age categories, acting alongside the population-wide herd effects resulting from reduced viral circulation. Nevertheless, the concurrent and significant decline in incidence among adults (≥20 y old), a group unaffected by such cohort aging effects, provides compelling and independent evidence for a substantial indirect protection effect of the vaccination program.

Guangzhou has implemented a two-dose varicella vaccine immunization strategy, aiming to ultimately reduce the incidence of varicella by enhancing the herd immunity level. The short-term effects in the initial policy implementation stage have shown a certain degree of complexity. This study observed an increase in varicella incidence across all age groups during the initial policy implementation phase. Theoretically, as it takes some time for the varicella vaccine to provide effective protection and the coverage rate of the second dose has not yet reached the level required to establish a stable herd immunity barrier (WHO-recommended target ≥80%) in the short term,¹⁶ there is a lag in the expected manifestation of policy effects. The possibility of a significant reduction in the incidence rate in the short term is relatively low. Overall, the short-term increase in incidence initially after the policy is likely due to a combination of heightened surveillance sensitivity, overlap with the seasonal peak, and underlying long-term trends.

It is well recognized that varicella displays distinct seasonality. Our seasonal index analysis revealed a bimodal distribution; the primary peak occurs from October to January, with a secondary peak in May. This pattern is consistent with findings from studies in Shaanxi,²⁵ Chongqing,²⁵ and Kaifeng.²⁵

Although varicella is not a legally notifiable infectious disease in China, Guangzhou has proactively included it in the local infectious disease management system since 1995. This nearly three-decade-long standardized management mechanism has provided a crucial guarantee for the systematicity, completeness, and reliability of varicella epidemic data, and has also laid a solid data foundation for the analysis of the effectiveness of vaccination policies in this study.

However, this study still has several limitations. Although the ITS method can effectively analyze the varicella policy’s short-term and long-term effects, it is difficult to fully control all potential confounding factors (such as changes in other public health intervention measures during the same period).²⁶ Furthermore, data on first- and second-dose varicella vaccination coverage rates were not obtained, nor were climate or other potentially relevant factors incorporated into the model. Additionally, this study did not assess the impact of vaccination on herpes zoster (HZ) incidence. Given that childhood HZ is relatively uncommon in China and systematic baseline surveillance data are lacking, coupled with the short duration since program initiation and the fact that a routine second dose has not been implemented, evaluating its long-term effect on HZ is currently not feasible. Moreover, the observed decline in varicella incidence after 2018 overlapped with the COVID-19 pandemic beginning in 2020 and its associated non-pharmaceutical interventions (such as mask-wearing and social distancing), which may have contributed additionally to the reduction. However, we consider the two-dose vaccination program to be the predominant driver, based on the following: a decline in varicella incidence, particularly observable in the 4–6-y-old age group, was already evident during the initial phase of the two-dose policy implementation (2018–2019). This trend emerged 1–2 y prior to the onset of the COVID-19 pandemic and the introduction of strict NPIs in Guangzhou (which began in early 2020); The declining trend in varicella incidence established post-2018 was sustained after 2020. Therefore, the observed effect is likely the result of combined vaccination and non-pharmaceutical interventions, with vaccination serving as the fundamental driver. In future evaluations, acquiring vaccination data could enable further analysis to assess the effects of administering the second vaccine dose to student populations at different coverage levels. This would provide direct evidence for optimizing vaccination strategies (such as target populations and timing of vaccination) and potentially allow for conducting cost-effectiveness analyses to support resource allocation decisions. Furthermore, establishing a long-term HZ surveillance system to assess the enduring benefits of immunization programs represents an important direction for future research.

Conclusions

Through a comprehensive analysis of the epidemic data before and after implementing the two-dose varicella vaccination policy in Guangzhou, this study confirms that the strategy has played a significant and positive role in controlling the spread of varicella. The results indicate that the current immunization schedule (“first dose at 1 y of age, booster dose at 4–6 y of age”) not only effectively reduces the incidence rate and outbreak risk among the target pediatric age group, but also indirectly protects susceptible populations who are unable to receive the vaccine (such as infants and adults) by establishing a robust herd immunity barrier. Therefore, it contributes to effectively controlling the varicella epidemic across the entire population. Therefore, we strongly recommend maintaining and promoting the two-dose vaccination program in Guangzhou’s NIP. In the future, conducting assessments of the preventive effects and cost-effectiveness of different varicella vaccination strategies based on Guangzhou’s specific circumstances is essential to identify the optimal strategy for the city.

Biography

Qing He, Master of Science in Epidemiology, serves as a Director Physician. With extensive experience in public health, she has remained dedicated to the prevention and control of vaccine-preventable infectious diseases. Her work encompasses epidemiological surveillance, outbreak investigation and vaccine effectiveness studies.

Funding Statement

Guangzhou Municipal Science and Technology Program [2024A03J0419; 2025B04J003] and The Key Project of Medicine Discipline of Guangzhou [No. 2025–2027-11] funded this research and the APC.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data presented in this study are available at the request of the corresponding author for legal reasons.

Generative artificial intelligence

During the preparation of this work, the author used Deepseek to improve language fluency, correct grammatical errors, and enhance overall readability.

Supplementary material legends

Supplemental data for this article can be accessed online at https://doi.org/10.1080/21645515.2025.2610903.