Effect of Influenza Vaccination in Preventing Laboratory-Confirmed Influenza Hospitalization in Patients With Diabetes Mellitus

Abstract

Background

People with diabetes are at high risk of severe influenza complications. The influenza vaccination effect among diabetic patients remains inconclusive. We estimated the average effect of influenza vaccination status in the current and prior seasons in preventing laboratory-confirmed influenza hospitalization in diabetic patients.

Methods

Patients attended in hospitals and primary healthcare centers with influenza-like illness were tested for influenza from the 2013–2014 to 2018–2019 seasons in Navarre, Spain. A test-negative case-control design in diabetic inpatients compared the influenza vaccination status in the current and 5 prior seasons between laboratory-confirmed influenza cases and negative controls. Vaccination status of influenza-confirmed cases was compared between diabetic inpatients and outpatients. Influenza vaccination effect was compared between diabetic patients and older (≥ 60 years) or chronic nondiabetic patients.

Results

Of 1670 diabetic inpatients tested, 569 (34%) were confirmed for influenza and 1101 were test-negative controls. The average effect in preventing influenza hospitalization was 46% (95% confidence interval [CI], 28%–59%) for current-season vaccination and 44% (95% CI, 20%–61%) for vaccination in prior seasons only in comparison to unvaccinated patients in the current and prior seasons. Among diabetic patients with confirmed influenza, current-season vaccination reduced the probability of hospitalization (adjusted odds ratio, 0.35; 95% CI, .15–.79). In diabetic patients, vaccination effect against influenza hospitalizations was not inferior to that in older or chronic nondiabetic patients.

Conclusions

On average, influenza vaccination of diabetic population reduced by around half the risk of influenza hospitalization. Vaccination in prior seasons maintained a notable protective effect. These results reinforce the recommendation of influenza vaccination for diabetic patients.

diabetes mellitus, influenza, vaccine effectiveness, repeated vaccination, test-negative design

Influenza is a serious public health concern that affects a high proportion of the population every year [1]. People with diabetes mellitus or other major chronic conditions are at higher risk of developing severe influenza-related complications [2, 3]. The main preventive measure against influenza and its complications is vaccination [1].

Since patients with diabetes or other major chronic conditions are at increased risk of medical consultations and hospitalizations for influenza infections, annual influenza vaccination in these populations is recommended worldwide [1, 4, 5]. People vaccinated against influenza for one season are more likely to be vaccinated in the next one [6]. Therefore, target subjects for influenza vaccination frequently accumulate several influenza vaccines over successive years.

Antibody response to influenza vaccination in adults with diabetes has been found to be similar to that of healthy adults [7, 8]. Influenza vaccination effect (IVE) in people with diabetes mellitus has been evaluated in several observational studies based on non-laboratory-confirmed clinical endpoints such as hospitalizations or deaths by all causes or specific clinical causes [9–14]. However, the results of these studies were inconclusive due to different sources of bias [15–17]. To the best of our knowledge, one study evaluated IVE in preventing laboratory-confirmed influenza in outpatients with diabetes [18], and another evaluated the effect in preventing hospitalizations in one season [19]; nevertheless, no studies have evaluated the effect in preventing hospital admissions with confirmed influenza over several seasons.

The aim of this study was to estimate the average effect over 6 seasons of the influenza vaccination status in the current and prior seasons in preventing laboratory-confirmed influenza hospitalization in patients with diabetes mellitus. We further aimed to compare these estimates against older or chronic nondiabetic patients.

METHODS

Study Population

This study was performed in the region of Navarre (approximately 650 000 inhabitants), northern Spain, where annual IVE population-based studies have been conducted on a population-based scheme since 2009 [20–26]. The Navarre Health Service provides healthcare to residents, free at the point of use. The trivalent inactivated influenza vaccine was annually recommended and offered free of charge to the target population, which included people aged 60 years or older, and to those with risk factors or major chronic conditions regardless of their age. All patients with diabetes mellitus were considered a target population for vaccination.

Influenza surveillance was based on automatic reporting of cases of medically attended influenza-like illness (ILI) from all primary healthcare centers and hospitals in Navarre [23]. ILI was defined as the sudden onset of any general symptom (fever or feverishness, malaise, headache, or myalgia) in addition to any respiratory symptom (cough, sore throat, or shortness of breath). The protocol for influenza cases in hospitals established early detection and double swabbing, nasopharyngeal and pharyngeal, at admission of all hospitalized patients with ILI. A sentinel network composed of a representative sample of 16% of primary healthcare physicians was trained to perform a double swab, after verbal informed consent, from all patients diagnosed with ILI and whose symptoms had begun within the previous 5 days. Swabs were tested for influenza virus by reverse-transcription polymerase chain reaction (RT-PCR).

In every influenza vaccination campaign, diagnoses of diabetes mellitus and other major chronic conditions (cardiovascular disease, respiratory disease, renal disease, cancer, liver cirrhosis, dementia, stroke, immunodeficiency, rheumatic disease, and body mass index ≥ 40 kg/m²) were obtained from the electronic medical records of primary healthcare, where these conditions are codified according to the International Classification of Primary Care, second edition [27]. This source of information has demonstrated 98% sensitivity and 99% specificity to detect diagnosed cases of diabetes [28].

Influenza vaccination status in the current and 5 previous seasons and the pneumococcal vaccination status were obtained from the regional vaccination register [6]. Only vaccines received ≥ 14 days prior to illness onset were considered.

Study Design

This study evaluated the average effect over 6 seasons of the influenza vaccination status in the current and prior seasons, regardless of variations in the effectiveness from season to season. In a test-negative case-control study [29] nested in the cohort of population covered by the Navarre Health Service, we pooled 6 influenza seasons, from 2013–2014 to 2018–2019. The study included only patients with continuous residence in the region during the 5 years before the analyzed influenza season. Children younger than 9 years, healthcare workers, and nursing homes residents were excluded. All patient information was linked using an individual identification number. The Navarre Ethical Committee for Medical Research approved the study protocol.

The primary analysis evaluated IVE in preventing hospitalizations with laboratory-confirmed influenza in diabetic patients. Cases were diabetic patients who were hospitalized for ILI and confirmed for influenza virus by RT-PCR, and controls were patients with diabetes mellitus admitted for ILI who tested negative for any influenza virus.

The effect of influenza vaccination on the probability of being hospitalized in patients who presented an influenza infection was evaluated in a case-to-case comparison [30]. We compared the hospitalized diabetic patients with confirmed influenza against diabetic patients who received assistance in primary healthcare centers for ILI and were confirmed for influenza virus by RT-PCR.

Furthermore, a test-negative case-control design was applied to the whole influenza vaccination target population to assess IVE in preventing laboratory-confirmed influenza hospitalization. IVE was compared between diabetic patients and the rest of the target population for vaccination.

Statistical Analysis

Frequencies of characteristics of cases and controls were compared by χ ² test. Logistic regression was used to calculate the adjusted odds ratios (aORs) for age groups (9–64, 65–84, and ≥ 85 years), other major chronic conditions, and month–season of sample collection. The 95% confidence intervals (CIs) of the aOR were obtained. Sensitivity analyses were performed for additional adjustment for sex, pneumococcal vaccination status, and hospitalization in the prior 12 months. The IVE was estimated as (1 – aOR) × 100%.

We evaluated the protective effect of influenza vaccination status in the current and prior seasons from the patient’s point of view instead of evaluating the effect of a single vaccine dose. IVE estimates were obtained for combinations of the influenza vaccination status in the current and prior seasons. When differences between estimates in the categories were not statistically significant, these categories were aggregated to simplify the presentations of results as follow: current-season vaccination regardless of prior doses, vaccination in prior seasons only, and neither current-season vaccination nor prior doses as the reference category.

The interaction terms between the vaccination status and sex or age groups were tested. Stratified analyses were carried out by sex, age group (9–64 and ≥ 65 years), presence of other major chronic conditions, influenza season, and virus (sub)type (A/H1N1, A/H3N2, and B). Only seasons with a minimum number of cases of a given (sub)type were included from the pooled analysis for that outcome.

The IVE in diabetic and nondiabetic populations was estimated and compared in a logistic regression model considering the different categories that combined vaccination status and the diagnosis of diabetes mellitus.

RESULTS

Characteristics of Cases and Controls

A total of 1670 diabetic patients hospitalized with ILI were tested during the 6 influenza seasons. Influenza was confirmed in 569 (34%) cases: 296 A(H3N2), 142 A(H1N1), 128 influenza B, and 3 influenza A not subtyped. These laboratory-confirmed cases were compared against 1101 influenza-negative hospitalized patients.

Renal disease was less frequent among influenza cases than in test-negative controls (22% vs 30%; P = .001). No other statistically significant differences by sex, age, or presence of other comorbidities were observed. Among the case patients, 71% had been vaccinated against influenza in any of the 5 prior seasons and 58% had received the current-season vaccine in comparison to 82% and 68% of the control patients, respectively (P < .001 for both) (Table 1).

Table 1.

Characteristics of Diabetic Patients Hospitalized for Influenza-like Illness According to Influenza Infection Status, 2013–2014 to 2018–2019 Influenza Seasons

Characteristic	Laboratory-Confirmed Influenza Cases, No. (%)	Influenza-Negative Controls, No. (%)	P Value
Total	569 (100)	1101 (100)
Age group, y			.424
9–64	97 (17)	161 (15)
65–84	348 (61)	689 (62)
≥ 85	124 (22)	251 (23)
Sex			.310
Male	328 (58)	663 (60)
Female	241 (42)	438 (40)
Other major chronic conditions
Cancer	130 (23)	286 (26)	.161
Liver cirrhosis	37 (7)	89 (8)	.246
Renal disease	127 (22)	328 (30)	.001
Immunodeficiency	15 (3)	37 (3)	.419
Respiratory disease	198 (35)	419 (38)	.191
Cardiovascular disease	267 (47)	562 (51)	.110
Dementia	27 (5)	56 (5)	.761
Stroke	69 (12)	136 (12)	.894
Rheumatic disease	10 (2)	24 (2)	.562
Body mass index ≥ 40 kg/m2	53 (9)	103 (9)	.978
Vaccination in the current season			< .001
No	237 (42)	351 (32)
Yes	332 (58)	750 (68)
Influenza vaccines in the 5 prior seasons			< .001
0	163 (29)	203 (18)
1–2	64 (11)	138 (13)
3–5	342 (60)	760 (69)
Influenza season			.001
2013–2014	63 (11)	93 (8)
2014–2015	47 (8)	66 (6)
2015–2016	75 (13)	132 (12)
2016–2017	118 (21)	203 (18)
2017–2018	164 (29)	294 (27)
2018–2019	102 (18)	313 (28)

Vaccination Effect Among Patients With Diabetes Mellitus

The pooled analysis of the 6 influenza seasons considered the combinations of the influenza vaccination in the current and prior seasons. Categories of influenza vaccination in the current season with and without vaccination in prior seasons were aggregated since their estimates did not present statistically significant differences (Supplementary Tables 1 and 2). On average, the protective effect against hospitalization for influenza was 46% (95% CI, 28%–59%) in persons vaccinated in the current season regardless of prior doses, and 44% (95% CI, 20%–61%) in persons vaccinated in prior seasons only in comparison to unvaccinated individuals in the current and 5 prior seasons (Figure 1).

Figure 1.

Effect of influenza vaccination in the current and 5 prior seasons in preventing laboratory-confirmed influenza hospitalization among diabetic patients. Combined analysis of the 2013–2014 to 2018–2019 influenza seasons. ^aVaccination effect adjusted by age group (9–64, 65–84, and ≥ 85 years), other major chronic conditions, and month–season of sample collection. Abbreviation: CI, confidence interval.

The IVE estimate for current-season vaccination was higher in females than in males (58% vs 39%; P for interaction = .271), as well as in subjects aged 9–64 years in comparison to those ≥ 65 years (58% vs 40%; P for interaction = .340), with no statistical significance in either case. In diabetic patients with other comorbidities, the protective effect of current-season vaccination was 50% (95% CI, 32%–63%), similar to the overall analysis results.

In the sensitivity analyses adjusted for sex, pneumococcal vaccination status, and hospitalization within the prior 12 months, the estimated IVE remained almost unchanged (Supplementary Table 3). IVE estimates by influenza season showed high variability among seasons, ranging between 21% and 84% (Supplementary Table 4).

The average effect observed in people with current-season vaccination in preventing hospitalizations was high against influenza B (71% [95% CI, 50%–84%]), moderate against influenza A(H1N1) cases (57% [95% CI, 32%–73%]), and low against influenza A(H3N2) (21% [95% CI, −16% to 46%]). Vaccination in previous seasons only had a considerable protective effect against influenza A(H3N2) and B cases, and a possible effect against influenza A(H1N1) (Table 2).

Table 2.

Effect of Influenza Vaccination in the Current and 5 Prior Seasons in Preventing Laboratory-Confirmed Influenza Hospitalization by (Sub)type Among Diabetic Patients—Combined Analysis of the 2013–2014 to 2018–2019 Influenza Seasons

Influenza (Sub)type	Cases/Controls	Vaccination Effect, % (95% CI)a	P Value
A(H1N1) subtypeb
Never vaccinated	52/136	Reference
Vaccination in prior seasons only	26/124	32 (−24 to 62)	.211
Current season vaccination	64/572	57 (32–73)	< .001
A(H3N2) subtypec
Never vaccinated	57/152	Reference
Vaccination in prior seasons only	37/156	41 (2–64)	.039
Current season vaccination	193/661	21 (−16 to 46)	.227
B typed
Never vaccinated	39/69	Reference
Vaccination in prior seasons only	22/87	61 (23–81)	.007
Current season vaccination	64/336	71 (50–84)	< .001

Abbreviation: CI, confidence interval.

^aVaccination effect adjusted by age group (9–64, 65–84, and ≥ 85 years), other major chronic conditions, and month–season of sample collection.

^bPooled A(H1N1) analysis includes 2013–2014, 2015–2016, 2017–2018, and 2018–2019 seasons.

^cPooled A(H3N2) analysis includes 2013–2014, 2014–2015, 2016–2017, 2017–2018, and 2018–2019 seasons.

^dPooled B analysis includes 2014–2015, 2015–2016, and 2017–2018 seasons.

Among diabetic patients with laboratory-confirmed influenza, the comparison of hospitalized patients and those attended in primary healthcare showed that current-season vaccination was associated to a 65% (aOR, 0.35 [95% CI, .15–.79]) reduction in the probability of hospitalization, whereas no decrease of this probability was detected in subjects who received vaccination in prior seasons only (aOR, 1.00 [95% CI, .34–2.95]) (Supplementary Table 5).

Comparison of Diabetic and Nondiabetic Patients

A total of 5171 patients with major chronic conditions or aged 60 years or older were hospitalized with ILI and tested for influenza during the study period. In a joint multivariate analysis, the IVE was compared between 1670 (32%) diabetic and 3501 (68%) nondiabetic patients. The risk of hospitalization for confirmed influenza in unvaccinated persons did not differ significantly between diabetic and nondiabetic individuals (aOR, 1.20 [95% CI, .92–1.55]). The risk of hospitalization for confirmed influenza was slightly lower, with no statistically significant difference, in diabetic than in nondiabetic patients for vaccination in the current season (aOR, 0.89 [95% CI, .76–1.05]), and vaccination in prior seasons only (aOR, 0.86 [95% CI, .62–1.19]). The estimates derived from subanalyses in people younger and older than 65 years were consistent (Table 3).

Table 3.

Comparison of the Risk of Laboratory-Confirmed Influenza Hospitalization Between Diabetic Patients and Other Target Populations for Influenza Vaccination According to Influenza Vaccination Status in the Current and 5 Prior Seasons—Combined Analysis of the 2013–2014 to 2018–2019 Influenza Seasons^a

Population	Cases/Controls	Adjusted OR (95% CI)b	P Value
All patients
Nondiabetic target population unvaccinated	393/578	Reference
Diabetic patients unvaccinated	149/175	1.20 (.92–1.55)	.184
Nondiabetic target population vaccinated in prior seasons only	178/333	Reference
Diabetic patients vaccinated in prior seasons only	88/176	0.86 (.62–1.19)	.355
Nondiabetic target population vaccinated in the current season	674/1345	Reference
Diabetic patients vaccinated in the current season	332/750	0.89 (.76–1.05)	.163
Aged 9–64 y
Nondiabetic target population unvaccinated	168/268	Reference
Diabetic patients unvaccinated	56/54	1.55 (.98–2.44)	.062
Nondiabetic target population vaccinated in prior seasons only	45/89	Reference
Diabetic patients vaccinated in prior seasons only	15/42	0.78 (.37–1.64)	.510
Nondiabetic target population vaccinated in the current season	63/176	Reference
Diabetic patients vaccinated in the current season	26/65	1.12 (.63–1.99)	.694
Aged ≥ 65 y
Nondiabetic target population unvaccinated	225/310	Reference
Diabetic patients unvaccinated	93/121	1.07 (.77–1.49)	.693
Nondiabetic target population vaccinated in prior seasons only	133/244	Reference
Diabetic patients vaccinated in prior seasons only	73/134	0.91 (.63–1.30)	.592
Nondiabetic target population vaccinated in the current season	611/1169	Reference
Diabetic patients vaccinated in the current season	306/685	0.86 (.72–1.02)	.084

Abbreviations: CI, confidence interval; OR, odds ratio.

^aTarget population for influenza vaccination included people ≥ 60 years of age or with major chronic conditions.

^bOdds ratio adjusted by age group (9–64, 65–84, and ≥ 85 years), other major chronic conditions, and month–season of sample collection.

DISCUSSION

To the best of our knowledge, this is the first study that evaluates the average effect over several seasons of influenza vaccination in the current and previous seasons in preventing laboratory-confirmed influenza hospitalization in individuals with diabetes mellitus. In the pooled analysis of 6 influenza seasons, current-season vaccination had a moderate effect in preventing influenza hospitalization (46%), and vaccination in prior seasons only retained a substantial protective effect (44%). Furthermore, we found that current-season influenza vaccination decreased the probability of hospitalization among patients with diabetes and laboratory-confirmed influenza, possibly explained as a vaccine-related reduction in the severity in patients to whom the influenza infection was not prevented [30]. This was not observed for vaccination in prior seasons only. All of these results support the benefit of influenza vaccination among patients with diabetes mellitus, which is of special interest as diabetes has been associated with an increased risk of influenza-related complications and deaths [2, 3].

Influenza vaccination is recommended to all patients with high-risk conditions, including patients with diabetes. Thus, it is not ethically accepted to conduct placebo-controlled trials; and assessment of IVE should be based on observational studies [29]. Several studies have evaluated IVE against nonspecific outcomes such as hospitalizations with clinical diagnosis of pneumonia or influenza and all-cause deaths [15, 16]; however, they are prone to relevant misclassification because clinical ILI due to influenza virus is indistinguishable from other respiratory infections not preventable by influenza vaccination [9–14].

The test-negative case-control design has demonstrated many advantages in estimating IVE. Only laboratory-confirmed cases were included and compared against test-negative controls with similar symptoms, and all were recruited in the same healthcare settings before either the patients and the physicians knew the laboratory results, which ensures comparability between cases and controls and reduces selection bias [29, 31, 32]. This case-control study was nested in a population-based cohort for which extensive and reliable databases were available. Study subjects were treated in hospitals with protocols for swabbing all ILI patients. All analyses were adjusted for the most commonly recognized confounding variables. All of this helps to prevent unmeasured confounding [33].

IVE estimates were slightly higher in individuals aged 9–64 years in comparison to older patients, with no statistically significant difference, which may be due to immunosenescence. We found that IVE over 6 seasons was on average high in preventing hospitalizations for influenza B cases, moderate for influenza A(H1N1), and low for influenza A(H3N2). Shang et al reported a similar IVE pattern by virus sub(type) in diabetic outpatients [18]. Other studies have communicated consistent results within the general population [31, 34].

Most studies evaluating the IVE did not consider the history of previous vaccinations, although they may preserve a significant protective effect in subsequent years [34–37]. Since annual administration of the influenza vaccine is recommended in patients with diabetes mellitus, they may accumulate several doses of vaccines over the years. So far no study had evaluated IVE considering the history of previous vaccinations in this type of patient. Here, we show a considerable protective effect of the vaccines received in prior seasons among diabetic people unvaccinated in the current season. Immunogenicity studies have shown that the level of antibodies peaks a few weeks after vaccination and then progressively declines, although it may persist for years [38]. Additional doses of vaccine may boost the response, particularly when vaccine components remain unchanged or the antigenic distance is small, which is frequently the case. On average, the annual drift of influenza viruses was not sufficient to completely evade the protective effect of the influenza vaccines from prior seasons [36].

Failure in annual vaccination has been related to neglect, a recent diagnosis with a new major condition, or a recent hospitalization [39]. In these cases, the residual effect of prior vaccines may be essential to prevent influenza.

In our study, a similar IVE in preventing hospitalization was observed in diabetic patients and in older or chronic nondiabetic population. Similarly, a recent study reported no differences in IVE in preventing outpatient consultations among individuals with different high-risk medical conditions, including diabetes [18]. Thus, annual IVE-obtained estimates for overall risk populations may be valid for diabetic patients. Furthermore, some studies found no differences in antibody response acquired by influenza vaccination in adults with diabetes compared to a group of healthy adults [7, 8].

This study has several strengths. The test-negative design is currently accepted as the reference design for IVE studies [29, 32]. All cases were laboratory-confirmed for influenza, and controls were negative. Study subjects were recruited within the same population and during the same influenza seasons, received the same vaccine, and were exposed to the same circulating viruses. The pooled analyses of 6 influenza seasons increased the power of the study and achieved sufficient representation of different virus (sub)types and patient characteristics. The vaccination history was obtained from the regional vaccination registry [6], and the study was limited to the population with stable residence in the region to avoid biases due to vaccination information.

This study may also be subject to some limitations. We were not able to distinguish between diabetes types 1 and 2 and assumed that most cases were type 2 diabetes since the study was restricted to people aged 9 years or older and most of participants were older than 65. Diabetes diagnosis was obtained from primary healthcare electronic records and, although some misclassification is possible, the validity of this diagnosis has been previously reported [28]. Undiagnosed diabetes may be frequent and was not included in this study. This study was carried out in a single place where vaccination was recommended in persons over 60 years and in those with risk factors, and only inactivated trivalent vaccine was administered. Consequently, care must be taken when generalizing our results to other places with different indications for vaccination, different vaccination coverage, or where other types of vaccines are used. As the statistical power in the analysis of a single season was reduced, the results should be interpreted cautiously. The results of the pooled analysis of 6 influenza seasons should be understood as an average. Since observational studies may be affected by confounding bias, the analyses were adjusted for age, presence of other comorbidities, and season of sample collection. Sex, pneumococcal vaccination, and hospitalizations in the preceding year have been described as potential confounding factors [13, 40], but sensitivity analyses ruled out their effect on the IVE estimates. Although we had no information about the severity of the diabetes, its effect on the estimates may be controlled by the test-negative design and variables we considered in the analysis such as other chronic comorbidities and hospitalization in the prior 12 months.

In conclusion, influenza vaccination reduces the risk of laboratory-confirmed influenza hospitalizations among patients with diabetes mellitus by about half on average of several influenza seasons. As IVE in people with diabetes was similar to that in people with chronic conditions other than diabetes, the estimates obtained for all chronic patients may be valid for the diabetic population. Influenza vaccines received in recent seasons may retain some protective effect in people who were not vaccinated in the current season. When influenza vaccination fails to prevent influenza disease in diabetic patients, it can still reduce the probability of hospitalization. Our results reinforce the recommendations of annual vaccination for influenza in patients with diabetes.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. All authors participated in the design, implementation, analysis, and/or interpretation of the study and the development of this manuscript. All authors had full access to the data and gave final approval before submission.

Acknowledgments. The authors thank the Primary Health Care Sentinel Network and the Network for Influenza Surveillance in Hospitals of Navarre for the recruitment of patients to the study.

Financial support. This work was supported by the Influenza-Monitoring of Vaccine Effectiveness (I-MOVE) Network supported by the European Centre for Disease Prevention and Control; by the Horizon 2020 program of the European Commission (I-MOVE-plus, agreement number 634446); and by the Instituto de Salud Carlos III with the European Regional Development Fund (grant numbers PI17/00868, PI12/00087, and INT19/00028).

Potential conflicts of interest. The authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.