Influenza vaccine effectiveness against medically attended influenza illness in Beijing, China, 2014/15 season

ABSTRACT

Influenza vaccination is the most effective way of preventing influenza infections but its coverage is extremely low in China. Poor influenza vaccine effectiveness (VE) was reported in the 2014/15 season in some Northern Hemisphere countries with a predominance of H3N2 viruses belonging to the 3C.2a clade. However, there is limited information regarding the preventive effect of influenza vaccination for the same season in China, in which H3N2 viruses belonging to the 3C.3a clade predominated. Through influenza virological surveillance in Beijing, China during the 2014/15 season, we estimated the influenza VE against medically attended influenza-like illness (ILI) associated with laboratory-confirmed influenza virus infection using a test-negative design, and the effect of prior vaccination on current vaccination was examined. In total, 9297 patients with ILI were enrolled in this study. Among them, 3434 (36.9%) tested positive for influenza viruses: 2167 (63.1%) for A(H3N2), 1261 (36.7%) for influenza B, and 3 (0.1%) for A(H1N1)pdm09. The adjusted VE was estimated as −25% (95% CI: −70%, 8%) against A(H3N2) and −8% (95% CI: −50%, 23%) against B, with an overall VE of −18% (95% CI: −49%, 6%). The overall VE estimate for patients who received 2014/15 vaccination only was −12% (−57%, 20%), while VE for patients who received both 2013/14 and 2014/15 vaccinations was −27% (−72%, 7%). There was no evidence that the influenza vaccine protected vaccinees against medically attended influenza in Beijing, China during the 2014/15 season.

KEYWORDS:

• China
• influenza
• influenza-like illness
• test-negative design
• vaccine effectiveness

Introduction

Influenza vaccination is the most effective way to prevent influenza infection and serious influenza-related complications. ¹ However, the influenza vaccination coverage rates vary greatly by country. ^{2 , 3} For example, in 2013, the coverage rates in the USA and Japan reached >45%, while those in the Republic of Korea, Canada, New Zealand and Australia were around 30%. The coverage rates in several European and South American countries in 2013 ranged from 20–30%, while the rates in some European Central and South American countries ranged from 10–20%. Globally, the coverage rate for most countries in 2013 were under 10%, and no influenza vaccines were distributed in some Africa countries. ³ The coverage rate for influenza vaccination in China in 2013 was under 1%. ³ One potential reason for this low coverage rate in China is that the influenza vaccine is not included in the national immunization program, and citizens in most cities of China, with the exception of several cities such as Beijing, must pay an out-of-pocket fee for their vaccinations. ^{4 , 5} The lack of convenient access to vaccines may also affect the Chinese vaccination uptake rates as only points of vaccination (POV) clinics located in community health service centers are officially authorized to provide immunization services in this country. ^{4 , 5} Other factors potentially contributing to the low rate of uptake of influenza vaccine in China are lack of awareness of the availability of the vaccine, lack of awareness of the benefit of influenza vaccination, and concerns over reported adverse effects of influenza vaccine due to misreporting of perceptions of vaccination.

Influenza vaccine effectiveness (VE) studies play a major role in monitoring how the vaccine works in the target population in specific regions and seasons during which the vaccine strains matched or mismatched the circulating viruses. In recent years, a test-negative design has been extensively used to estimate influenza VE in some countries; ^6-9 however, there is limited information regarding comparable studies in China. ^{10 , 11}

During the 2014/15 Northern Hemisphere influenza season, H3N2 viruses belonging to the 3C.2a clade predominated in many European and North American countries. These viruses mismatched the vaccine strain belonging to the 3C.1 clade that was recommended for use by the World Health Organization (WHO) in that season. Compared to the 2014/15 vaccine strain, the hemagglutinin protein of H3N2 viruses belonging to the 3C.2a clade had multiple amino acid changes, ¹² including the F159Y mutation adjacent to the receptor-binding site which is responsible for major antigenic change, ¹³ and H3N2 viruses belonging to the 3C.2a clade was poorly inhibited by post-infection ferret antisera raised against the 2014/15 vaccine strain. ¹² Thus, low VE estimates were observed in the countries where H3N2 viruses belonging to the 3C.2a clade predominated in the 2014/15 influenza season (the USA: 6%; Canada: −17%; the UK: 29.3%; other European countries: 14.4%). ^14-17 Although H3N2 viruses belonging to the 3C.2a clade predominated in many regions across the world in the 2014/15 influenza season, the viruses belonging to the 3C.3a clade predominated in some countries, including China. ¹⁸ The VE of the H3N2 vaccine strain in countries where H3N2 viruses belonging to the 3C.3a clade predominated in the 2014/15 season remains unclear.

Since 2007, a trivalent inactivated influenza vaccine (IIV3) has been provided free of charge to older adults aged ≥ 60 y and elementary and high school students aged 6–17 y in Beijing, the Chinese capital. Infants and children aged 6 months to 5 y and also adults aged <60 y have been encouraged to receive the vaccine at their own expense. ¹⁰ In this study, we used a test-negative design to examine overall or type/subtype-specific VE estimates against medically attended influenza in Beijing, China during the 2014/15 season.

Results

From November 10, 2014 (week 46 in 2014) to April 19, 2015 (week 16 in 2015), the epidemic period of the 2014/15 influenza season, a total of 9905 patients with medically attended influenza-like illness (ILI) were monitored by influenza virological surveillance in Beijing. Of the 9905 patients with ILI, 9297 (94%) had complete surveillance documentation and were enrolled in this study (Fig. 1). Among these 9297 patients, 3434 (36.9%) tested positive for influenza viruses: 2167 (63.1%) for A(H3N2), 1261 (36.7%) for influenza B, 3 (0.1%) for A(H1N1)pdm09, and 3 (0.1%) for influenza A (subtype unknown). The remaining patients (5863, 63.1%) all tested negative for influenza viruses. The proportion of females among the influenza-positive cases was 49.5% (1700/3434), compared with 46.3% (2717/5863) among the influenza-negative controls. The proportions of participants aged 3–17 y among the cases and controls were 27.6% (947/3434) and 30.3% (1778/5863), respectively, and the proportions of those aged ≥ 60 y among the cases and controls were 9.0% (308/3434) and 6.7% (394/5863), respectively. For the cases and controls, 7.3% (251/3434) and 5.4% (318/5863), respectively, had at least one chronic disease. Respiratory specimens were collected within 7 d of ILI onset for all the participants (Table 1).

Figure 1. Flow chart of subject enrollment in the test-negative design study for estimating influenza vaccine effectiveness during the 2014/15 influenza season in Beijing, China.

Table 1. Demographic characteristics of participants by influenza positive/negative status during the 2014/15 season.

Characteristics	Influenza Positive No. (Col.%)	Influenza Negative No. (Col.%)	P value
Overall	3434 (100)	5863 (100)
Sex
Male	1734 (50.5)	3146 (53.7)	0.003
Female	1700 (49.5)	2717 (46.3)
Age groups
6–35 months	167 (4.9)	713 (12.2)	<0.001
3–17 years	947 (27.6)	1778 (30.3)
18–59 years	2012 (58.6)	2978 (50.8)
≥ 60 years	308 (9.0)	394 (6.7)
Presence of at least one chronic disease
Yes	251 (7.3)	318 (5.4)	<0.001
No	3183 (92.7)	5545 (94.6)
Interval between ILI onset and specimen collection
≤ 3 days	3309 (96.4)	5592 (95.4)	0.024
4–7 days	125 (3.6)	271 (4.6)

ILI: influenza-like illness; Col.: column.

^a Pearson's χ² test was used to assess the demographic characteristic differences between influenza-positive cases and influenza-negative controls.

Among the influenza-positive cases, 4.3% (149/3434) received the 2014/15 IIV3, compared with 3.7% (215/5863) among the influenza-negative controls, however, this difference was not statistically significant (p = 0.107). The unadjusted overall VE was estimated as −19% (95% confidence interval [CI]: −47%, 4%) for preventing medically attended influenza (Table 2). After adjusting for sex, age group, chronic disease, and calendar week, the adjusted overall VE was −18% (95% CI: −49%, 6%), with −25% (95% CI: −70%, 8%) against influenza A(H3N2) and −8% (95% CI: −50%, 23%) against influenza B. The adjusted VE estimate was −22% (95% CI: −59%, 6%) for patients aged 3–17 years, −60% (95% CI: −415%, 50%) for those aged 18–59 years, and 16% (95% CI: −54%, 54%) for those aged ≥ 60 years, however, the CIs for all the subgroups overlapped zero.

Table 2. Percentage of participants receiving the seasonal 2014/15 influenza vaccine among influenza-positive cases and influenza-negative controls, and the unadjusted and adjusted estimates of influenza VE in the 2014/15 season in Beijing, China.

	Influenza positive	Influenza negative	Unadjusted VE	Adjusted VE
Groups	Vaccinated No./total (Row %)	Vaccinated No./total (Row %)	VE% (95% CI)	VE% (95% CI)
Overall	149/3434 (4.3)	215/5863 (3.7)	−19 (−47 to 4)	−18 (−49 to 6)
Influenza type/subtype
Influenza A (H3N2)	95/2167 (4.4)	215/5863 (3.7)	−20 (−54 to 6)	−25 (−70 to 8)
Influenza B	54/1261 (4.3)	215/5863 (3.7)	−18 (−59 to 13)	−8 (−50 to 23)
Age groups (yrs)
6–35 months	0/167 (0.0)	4/713 (0.6)	NA	NA
3–17 years	120/947 (12.7)	174/1778 (9.8)	−34 (−71 to −4)	−22 (−59 to 6)
18–59 years	7/2012 (0.3)	6/2978 (0.2)	−73 (−415 to 42)	−60 (−415 to 50)
≥ 60 years	22/308 (7.1)	31/394 (7.9)	10 (−59 to 49)	16 (−54 to 54)

ILI: influenza-like illness; VE: vaccine effectiveness; CI: confidence interval.

* VE was calculated as 100% × (1-odds ratio [the odds of vaccination among influenza-positive cases vs. the odds of vaccination among influenza-negative controls]) using unconditional logistic regression models.

^a Participants were considered to be vaccinated if the last dose of the vaccine was given ≥ 14 d before ILI onset.

^b Unconditional logistic regression models were adjusted for sex, age group, chronic diseases, and calendar week.

^c Not available because the models did not converge due to small sample sizes.

We also estimated the overall effectiveness of exposure to each vaccine combination in the subjects (i.e., 2014/15 vaccination only, 2013/14 vaccination only, or both 2014/15 and 2013/14 vaccinations) compared with the unvaccinated subjects in both seasons (the reference group). In the adjusted model, the VE after receipt of the 2014/15 vaccination only, the 2013/14 vaccination only, and vaccinations in both seasons was −12% (95% CI: −57%, 20%), −31% (95% CI: −84%, 7%), and −27% (95% CI: −72%, 7%), respectively (Table 3).

Table 3. Percentage of participants vaccinated with the 2014/15 influenza vaccine and the 2013/14 influenza vaccine among influenza-positive cases and influenza-negative controls, with estimates of influenza VE in the 2014/15 season in Beijing, China.

	Influenza Positive	Influenza Negative	Unadjusted VE	Adjusted VE
Vaccination status	Vaccinated No. (Col.%)	Vaccinated No. (Col.%)	VE% (95% CI)	VE% (95% CI)
Overall	3434 (100)	5863 (100)
Unvaccinated	3218 (93.7)	5555 (94.7)	Reference	Reference
2014/15 vaccination only	63 (1.8)	105 (1.8)	−4 (−42 to 24)	−12 (−57 to 20)
2013/14 vaccination only	67 (2.0)	93 (1.6)	−24 (−71 to 9)	−31 (−84 to 7)
Both 2014/15 and 2013/14 vaccinations	86 (2.5)	110 (1.9)	−35 (−80 to −1)	−27 (−72 to 7)

ILI: influenza-like illness; VE: vaccine effectiveness; CI: confidence interval; Col.: column.

* VE was calculated as 100% × (1-odds ratio [the odds of vaccination among influenza-positives cases vs. the odds of vaccination among influenza-negative controls]) using unconditional logistic regression models.

^a Participants were considered to be vaccinated if the last dose of the vaccine was given ≥ 14 d before ILI onset.

^b Unconditional logistic regression models were adjusted for sex, age group, chronic diseases, and calendar week.

Discussion

We used a test-negative design to estimate the effectiveness of the influenza vaccine in Beijing, China in the 2014/15 influenza season. Our results suggest that the effect of vaccination in the 2014/15 influenza season in Beijing was negligible if any in terms of protection against influenza A and influenza B viruses.

During the 2014/15 season, H3N2 viruses belonging to the 3C.3a clade predominated in China, ¹⁸ while H3N2 viruses belonging to the 3C.2a clade predominated in Europe and North America. ^{15-17 , 19} Our adjusted VE estimate against H3N2 (−25%; 95% CI: −70%, 8%) in ambulatory care settings in Beijing during the 2014/15 season was close to that of the same season in Canada (−17%; 95% CI: −50%, 9%). ¹⁵ However, it was slightly lower than the VE estimate in the USA (6%; 95% CI: −5%, 17%), ¹⁹ and much lower than that in the UK (29.3%; 95% CI: 8.6%, 45.3%) ¹⁶ and lower also than the pooled estimate for other European countries (14.4%; 95% CI: −6.3%, 31.0%). ¹⁷ In addition, the specific VE against the antigenically drifted H3N2 viruses belonging to the 3C.2a clade in Canada was −13% (95% CI: −51%, 15%) ¹⁵ and only 1% (95% CI: −14%, 14%) in the USA. ¹⁹ The absent or low influenza VE against H3N2 viruses during the 2014/15 season is likely attributed to the mismatch between the H3N2 vaccine strain belonging to the 3C.1 clade and the predominantly circulating H3N2 viruses belonging to the 3C.2a or 3C.3a clades. ^{12 , 18} If there is sufficient mismatch so as to lead to very low VE, due to the long manufacturing cycle for the vaccine, there is nothing that can be done to change the vaccine during the season, such that one needs a group-common influenza vaccine when / if it becomes available in the distant future. Furthermore, the variation in overall VE estimates against H3N2 viruses between countries in the 2014/15 season may result from variations in the proportions of circulating viruses belonging to 3C.2a, 3C.3a, 3C.3 and 3C.b clades in these countries and the VE against each respective clade. ^{15-17 , 19}

The VE estimate against influenza B viruses in the 2014/15 influenza season in Beijing was −8% (95% CI: −50%, 23%), an estimate much lower than the end-of-season VE estimates in the USA (55%; 95% CI: 43%, 65%), ¹⁴ Canada (42%; 95% CI: 10%, 62%), ¹⁵ and the UK (46.3%; 95% CI: 13.9%, 66.5%). ¹⁶ However, a study conducted in Beijing showed that the effectiveness of the influenza vaccine for preventing influenza B hospitalizations was also very low (−31.5%, 95% CI: −153.9%, 31.9%). ¹¹ There are several potential reasons for the low VE against influenza B viruses in Beijing during the 2014/15 season. First, the predominant influenza B viruses in northern China during this season belonged to the Yamagata lineage clade 3, ²⁰ which differs from the B/Massachusetts/2/2012 vaccine strain (Yamagata lineage clade 2) recommended for inclusion in the vaccine for the 2014/15 influenza season. As reported by the WHO, vaccines containing B/Massachusetts/2/2012 antigens induced anti-HA antibodies that reacted well against the Yamagata lineage clade 2 viruses, while significant reductions in the geometric mean titers were observed more frequently when testing clade 3 viruses. ^{12 , 18} Second, influenza B viruses were circulating in late in the 2014/15 influenza season in Beijing, China (data not shown), thus any protective effects elicited by the pre-season vaccinations may have waned over time. A study using the combined data from Europe from the 2010/11 season to the 2014/15 season that investigated the time since vaccination effects on influenza VE showed that the VE against influenza B virus declined from 70.7% 44 d after vaccination to 21.4% at season end. ²¹

VE studies conducted in Europe and Canada have indicated that negative interference from prior vaccination on current vaccination appeared in these regions during the 2014/15 season, ^{15 , 17} although this particular effect was not observed in our study or in the USA VE study for the same season. ¹⁴ Further studies with stricter designs, such as randomized controlled trials, are warranted to shed light on the precise effect of prior vaccination on current vaccination.

This study has provided new knowledge on VE in the influenza field. First, previous VE studies on the 2014/15 influenza season focused mainly on countries where H3N2 viruses belonging to the 3C.2a clade predominated, while this study provided VE estimates from a country where H3N2 viruses belonging to the 3C.3a clade predominated. Our findings augment current information on the VE of the 2014/15 H3N2 vaccine strain against the mismatched H3N2 viruses circulating at that time. Second, the data published on the effect of prior vaccination on current vaccination is inconsistent in different VE studies for the 2014/15 season, ^{14 , 15 , 17} and the findings of our study expanded current knowledge in this respect.

There are several limitations associated with this study. First, in light of its observational study design, the possibility of unidentified confounders exists. Second, the convenience sampling of eligible patients may lead to selection biases on VE estimates. However, the exact effects resulting from these potential selection biases are unknown and require further investigations. Third, because of the small sample sizes for the specific age groups, we could not estimate VE stratified by age group. Fourth, in this study, the vaccination coverage rate was very low, and this limited the study's statistical power and resulted in the absence of statistical significance, especially for subanalyses on, for example, the effect of repeat vaccinations. Fifth, as very few institutions in Beijing, such as hospitals, provide influenza vaccines to their employees free of charge, these vaccinees would not be registered by the staff at the POV clinics, which are the places officially authorized to provide immunization services in Beijing, China, and these could be missed through the Beijing Management System of Information for the Immunization Program. However, given the extremely small number of these missed vaccinees, the impact on VE estimation is likely to be minimal.

In conclusion, our results have indicated that there was no evidence of the influenza vaccine protecting vaccine recipients against medically attended influenza in Beijing, China during the 2014/15 influenza season. An effect of prior vaccination on current vaccination was not observed in this test-negative design study. Dynamic monitoring of the genetic and antigenic features of circulating influenza viruses and optimization of influenza vaccine recommendations and manufacturing processes are warranted to assist the development of an effective influenza vaccine.

Materials and methods

Influenza virological surveillance

During the 2014/15 influenza season, 23 sentinel hospitals (11 national and 12 municipal surveillance sites) and 17 collaborating laboratories across 16 districts of Beijing participated in influenza virological surveillance in Beijing, China. According to the procedures as described previously, ¹⁰ doctors at the ambulatory care clinics of sentinel hospitals screened and enrolled ILI (defined as temperature ≥ 38°C with either cough or sore throat) patients within 3 d of symptom onset after obtaining their informed consent. Convenience sampling was performed with the aim of enrolling a weekly sample size of 20 patients/national sentinel hospital and 15/municipal sentinel hospital according to the sample size requirements stated in the documents issued by the national health administrative authority of China and the local health administrative authority of Beijing, China. In Beijing, ILI patients traditionally seek medical attention at hospitals rather than at private clinics, ²² thus these patients with ILI include both mild and severe cases. Nurses collected epidemiologic information and oral pharyngeal (OP) swabs from the enrolled patients with ILI. OP swabs were temporarily stored in viral transport medium at 4°C, and then transported to the corresponding collaborating laboratory for typing and subtyping of influenza viruses using real-time reverse-transcription polymerase chain reaction following the procedures recommended by the WHO Collaborating Center for Reference and Research on Influenza at the Chinese National Influenza Center. Influenza-positive OP specimens were subject to virus isolation using Madin–Darby canine kidney (MDCK) cells. Information on the ILI patients enrolled from the national sentinel hospitals and the influenza virus strains isolated from these patients had to be submitted to the Chinese National Influenza Center.

Study design

Influenza-positive patients were continuously observed in the influenza virological surveillance each week from 2014 to 2015 in Beijing. Consequently, we were unable to adopt a study period covering the time between the week of onset for the first influenza case and the week of onset for the last influenza case during the influenza season, a method used in other studies. ^{11 , 14 , 17 , 23} We defined the start of the study period as at least 14 d after commencement of the influenza vaccination campaign in Beijing (usually October 15 during each season) because vaccinees are typically considered to be vaccinated 14 d after receiving their influenza vaccinations. ^{14 , 16 , 17} In view of the above-mentioned points, we used the influenza epidemic period as the study period. The influenza epidemic period was defined as the weeks in which the respective weekly proportion of influenza-positive patients during virological surveillance was >20% of the proportion in the peak week during the influenza season (i.e., the proportion of influenza-positive patients every week in the epidemic period >20% * the highest weekly proportion). ²⁴ For example, if the proportion of influenza-positive patients in the peak week is 70% (i.e., the highest weekly proportion in the influenza season), the weeks with individual proportions >14% (i.e., 70% * 20%) in this season would be included in the study period.

Influenza-positive and influenza-negative patients aged ≥ 6 months that were identified through influenza virological surveillance during the study period of the 2014/15 influenza season were included in this study. Influenza-positive cases were ILI patients who were swabbed and tested positive for influenza, while influenza-negative controls were ILI patients who were swabbed and tested negative for influenza. In this study, a test-negative design was used to estimate VE against medically attended influenza in this study, ^{25 , 26} focusing on mild illness.

Data collection

Epidemiological information for each patient enrolled during influenza virological surveillance was collected by nurses using a standardized questionnaire. Variables included name, sex, birth date, home address, illness onset date, specimen collection date, and presence of chronic diseases. Presence of chronic diseases was defined as having any one of the following diseases: asthma, tuberculosis, pulmonary fibrosis, chronic tracheitis or bronchitis, emphysema, chronic obstructive pulmonary disease, diabetes, anemia, oncological diseases, diseases of the immune system, cardiovascular and cerebrovascular diseases, renal diseases, hepatopathy and neurological diseases. The questionnaire was sent along with respiratory specimens to the corresponding collaborating laboratory. Influenza vaccination records were documented from the information obtained from the Beijing Management System of Information for the Immunization Program.

A trivalent inactivated influenza vaccine is used in Beijing, China. The composition of this vaccine for the 2014/15 influenza season recommended by WHO included A/California/7/2009 (H1N1)pdm09, A/Texas/50/2012 (H3N2), and B/Massachusetts/2/2012-like viruses. Demographic information and influenza vaccination history on the vaccinees were deposited in the Beijing Management System of Information for the Immunization Program by staff at the POV clinics located in community health service centers, which are the places officially authorized to provide immunization services in Beijing, China. This system includes adults aged ≥ 60 y and elementary and high school students who received influenza vaccinations free of charge, and vaccinees that paid for their own influenza vaccines. According to the individual names, genders, birth dates, and home addresses of the ILI patients enrolled in this study, trained staff from our research team identified the influenza vaccination status of each participant by querying the Beijing Management System of Information for the Immunization Program. According to the package information supplied with the influenza vaccines produced by manufacturers licensed by the China Food and Drug Administration, people aged ≥ 3 y need one adult dose of influenza vaccine, while children aged 6–35 months require 2 doses of the influenza vaccine for children. In this study, participants aged ≥ 3 y were considered to be vaccinated if they received the 2014/15 influenza vaccine at least 14 d before the illness onset, and those who did not receive an influenza vaccine in the current season or received it <14 d before the illness onset were considered to be unvaccinated. Participants aged 6–35 months who received 2 doses of the influenza vaccine at least 14 d before illness onset were regarded as vaccinated. Participants aged 6–35 months of age with other vaccination statuses, such as unvaccinated, single-dose vaccination only, one dose of vaccination at least 14 d before illness onset but another dose <14 d later, or who received both vaccination doses <14 d before illness onset, were considered to be unvaccinated.

Data analyses

Statistical analyses were performed with SPSS 20.0 software (SPSS Inc., Chicago, IL, USA). Percentages were calculated for categorical variables. Unconditional univariate and multivariate logistic regression models, without and with adjusting for variables, were used to estimate odds ratios (OR; the odds of influenza vaccination among influenza-positive cases divided by the odds of influenza vaccination among influenza-negative controls). VE was estimated as 100% × (1-OR). ²⁵ The adjustment variables included gender, age group (6–35 months, 3–17 years, 18–59 years, and ≥ 60 years), presence of at least one chronic disease, and calendar week. Stratified VE estimates were also calculated by age group and by influenza virus type. Additionally, as recent studies have indicated that the effectiveness of the influenza vaccine received in the current season is influenced by previous seasons vaccinations, ^{27 , 28} the modification effect of vaccination in the 2013/14 season on vaccination in the 2014/15 season was also assessed. All statistical tests were 2-sided, and statistical significance was defined as p < 0.05 or the lower bound of the 95% CI for VE >0.

The Institutional Review Board and Human Research Ethics Committee of the Beijing Center for Disease Prevention and Control approved this study.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding

This work was supported by grants from the Beijing Science and Technology Planning Project of Beijing Science and Technology Commission (D141100003114002), the Capital Health Research and Development of Special (2014–1–1011), Beijing Health System High Level Health Technology Talent Cultivation Plan (2013–3–098), and Beijing Talents Fund (2014000021223ZK36).