- Email: [email protected]
Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics
ARTICLE IN PRESS
PCD-533; No. of Pages 9
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
Contents lists available at ScienceDirect
Primary Care Diabetes journal homepage: http://www.elsevier.com/locate/pcd
Review
Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics Ludovic Casanova a,b,c,∗ , Nirvina Gobin c , Patrick Villani a,b,d , Pierre Verger a,b,d a
INSERM, UMR S 912, «Sciences Economiques & Sociales de la Santé et Traitement de l’Information Médicale» (SESSTIM), F-13385 Marseille, France b ORS PACA, Observatoire Régional de la Santé Provence-Alpes-Côte d’Azur, F-13006 Marseille, France c Aix Marseille University, Department of General Practice, Marseille, France d Aix Marseille Université, UMR S 912, IRD, Marseille, F-13385 Marseille, France
a r t i c l e
i n f o
a b s t r a c t
Article history:
Background: The influenza virus is an important cause of morbidity and mortality for diabet-
Received 3 March 2016
ics. The seasonal influenza vaccine’s immunologic effectiveness is proven within the type
Accepted 16 May 2016
1 and type 2 diabetic populations, but the level of evidence is low. This article presents a
Available online xxx
systematic review for the bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics.
Keywords:
Methods: Using systematic review methods, we searched three electronic databases for pub-
Bias
lished literature (MEDLINE, EMBASE and the Cochrane Library) and two grey literature (SIGLE
Diabetes mellitus
and NHS EED) databases, to identify studies published between 1997 and 2013, examining
Influenza, Human
the effect of seasonal influenza vaccination, among diabetics, on any measure for influenza
Influenza vaccines
morbidity or mortality. Results: 725 records were identified from the three databases and screening, short-listing was undertaken independently by two reviewers. After de-duplication, all records were screened by title and then abstract, and 34 short-listed records were reviewed in full, with 7 studies included: 4 cohort studies and 3 case–control studies, conducted in 7 countries. The most common outcome of interest in studies (n = 4) was all-cause mortality among elderly diabetics (>65 years), with individual studies reporting reductions in risk of between 33% [95%CI: 4%–54%] and 68% [95%CI: 58%–75%]. We found only two studies for working-age adult diabetics: one reporting that vaccination prevented hospitalizations due to pneumonia or influenza (vaccine effectiveness [VE] 43%, [95%CI: 28%–54%]) and all-cause hospitalizations (VE: 28% [95%CI: 24%–32%]); and, another reporting no significant decrease in all-cause mortality for working-age adult diabetics. We have identified three major biases: the use of indirect health outcomes, a risk of selection bias (health-seeking bias), and no adjustment for participant pneumococcal vaccination status. The most recent included article finds that morbimortality is still lower during off-season influenza in both vaccinated and non-vaccinated diabetics, indicating important residual confounding.
∗ Corresponding author at: Observatoire régional de la santé Provence-Alpes-Côte d’Azur, 23 rue Stanislas Torrents, 13006 Marseille, France. Tel.: +33 4 91 59 89 14; fax: +33 4 91 59 89 18. E-mail address: [email protected] (L. Casanova). http://dx.doi.org/10.1016/j.pcd.2016.05.005 1751-9918/© 2016 Primary Care Diabetes Europe. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS
2
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
Conclusion: To date, the strength of evidence supporting the routine use of seasonal influenza vaccination is low for diabetics older than 65, and very low for working-age diabetics. © 2016 Primary Care Diabetes Europe. Published by Elsevier Ltd. All rights reserved.
Contents 1. 2. 3.
4.
5.
6.
1.
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .00 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 3.1. Literature research and search strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 3.2. Study eligibility and data synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 3.3. Selection, quality assessment, and data extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 4.1. Inclusion of articles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .00 4.2. Studies’ characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 4.3. Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 4.4. Vaccine effectiveness for diabetics aged 65 and older (Table 4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 4.5. Vaccine effectiveness for diabetics aged 18–65 years old (Table 4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 4.6. Identified biases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 5.1. Strengths and weaknesses of the review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 5.2. Consequences for medical practice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Conflict of interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Appendix A. Supplementary data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
Background
Seasonal influenza is an important cause of morbidity and mortality among diabetics. The World Health Organization recommends that this high risk group properly follow the vaccination schedule. However, the national influenza vaccination campaigns have been criticized [1–3], partly because the strength of evidence is weak regarding the influenza vaccine effectiveness (VE) in certain high risk groups. In fact, with many ethical committees worldwide advising to proceed with caution within diabetic populations, no randomized controlled trial is currently available. A recent systematic review and meta-analysis called for more research to assess effectiveness of seasonal influenza vaccination among diabetics [4].
2.
Introduction
Diabetic patients have an elevated rate of morbidity and mortality linked to infections, [5] as chronic hyperglycaemia can disturb white blood cell functions, such as phagocytosis, chemotaxis and cell adhesion [6]. Diabetic patients have a higher chance of developing pulmonary complications of an influenza infection [7] and their risk of mortality is two to four times higher as compared to non-diabetics [8]. The objective immune response resulting from the
influenza vaccine has been widely studied [9–12] and, more recently in 2011, for the serotype H1N1 [13], because of the immune system’s disorders due to diabetes. The results are consistent whatever the primary outcome of the immune studies (the antibody response among vaccinated, quantitative antibody titres, in vitro humoral and cell-mediated immune responses): the objective immune response by the seasonal influenza vaccine in diabetics patients is similar to non diabetics’ response. The same conclusions were drawn for the pneumococcal vaccine [14]. Although it is plausible that the seasonal influenza vaccine would offer greater protection to diabetics, this has not been directly proven. The public’s demands make it more necessary to prove the VE. The avian influenza epidemic (H5N1) in 2004 and the swine influenza pandemic (H1N1) in 2009–2010 have raised demand for influenza vaccines and many health authorities worldwide have mobilized to provide vaccination for vulnerable populations. Some countries, such as France, have developed prevention measures considered expensive and even unnecessary by the public and some health representatives. Thus ensued a certain level of scepticism from patients, conveyed by insufficient influenza vaccination, even more so by working-age diabetics [15,16]. Doubts on VE that could exist for diabetics patients show that the level of evidence has to improve. In 2015, the ethical committees maintain their refusal to conduct randomized controlled trials. Observational studies are the only solution to improve the level of evidence.
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS 3
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
The aim of this systematic review was to identify studies reporting the VE of seasonal influenza vaccine on morbidity and mortality, and to assess and discuss the quality of included studies.
3.
Records identified through database search Medline = 439 Embase = 286 Cochrane = 0
Methods Titles screened (n = 725)
Our methods are reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement guidelines [17] and checklist is included as an annex (Appendix A).
3.1.
Study eligibility and data synthesis
The population of interest was individuals with Type 1 or Type 2 Diabetes at any age. The intervention was to be the seasonal influenza vaccine. The control group were either those not vaccinated or those receiving a placebo. Any type of study design could have been used to compare the groups, be it case–control studies, cohort studies, quasi-experimental studies or clinical trials. Every criteria of morbimortality was taken into account. No restrictions were applied for the date of publication or language the article was published in.
3.3.
Abstracts screened (n = 120)
Literature research and search strategy
Five databases were searched using MeSH terms and keyword searches with search terms adapted according to the requirements of the databases (Annex 1). Three databases (Medline, Embase and the Cochrane Library) were searched for published literature, and two databases (System for Information of Grey Literature in Europe [SIGLE] and National Health System Economic Evaluation Database [NHS EED]) were searched for grey literature. No particular kind of software was used to do the research. The searches were originally run on 9th September 2014, and then updated on 16th January 2015.
3.2.
Excluded = 605
Selection, quality assessment, and data extraction
Two reviewers (LC & NG) screened titles according to the PICO criteria. Letters to editors, posters and short communications were excluded. The articles were first chosen by their titles by each author separately, then by the abstracts of the previously identified articles. At the completion of each step, reviewers compared and agreed lists of eligible and short-listed records. Any disagreement was resolved by a third reviewer (PV). Then, full articles were reviewed by both reviewers to check that inclusion criteria were met. The selected articles were independently assessed by the two reviewers. The analysis guidelines of literature and the recommended gradation of the National Agency of accreditation and health assessment [18] were used for the articles’ assessments. After pooling the findings, any disagreement between the two authors were solved by consensus. A common synthesis of the extracted data from each article was made. The potential biases of each study were identified, classified and ranked.
Excluded = 85
Full-text articles assessed for eligibility (n = 35)
Articles included in systematic review n=7
Excluded = 28 No data in the subgroup of diabetic patients (n=16) No data about vaccine effectiveness (n=5) Methodology not available: short communication, letter, poster or congress communication (n=4) Study not obtainable (n=3)
Fig. 1 – Flow diagram of study selection procedure for the systematic review.
4.
Results
4.1.
Inclusion of articles
Seven articles were selected and analyzed out of 725 (Fig. 1). Those seven articles were available on the two databases used. No other article was found from manual research and analysis of grey literature. In January 2015, there was no previous systematic review on this topic in the Cochrane Library.
4.2.
Studies’ characteristics
Every study included was made in different countries. There were four cohort studies, two of which did matching between the vaccinated and non-vaccinated groups, and three case–control studies, one of which was nestled in a cohort (Table 1). Only three out of seven articles included diabetics aged less than 65 years old, only two of which did a specific analysis on this specific group. All the studies were done from information from medical and administrative databases (Table 2). One study proceeded to collect their data on a single influenza epidemic season (winter 1999–2000) whereas the others studied two to seven influenza outbreaks [20].
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS
4
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
Table 1 – Studies included. First author
Study design
Year
Setting/Country
Age limits
Age mean
Colquhoun
Case control study
1997
≥18 years
Cohort study
2002
≥65 years
52.58 estimated value Missing
1.28
Hak Heymann
Case control study
2004
Leicestershire England 10 statesa United States Israel
≥65 years
Missing
Missing
Looijmans
Case control study in cohort Cohort study with matching Cohort study
2006
Netherlands
≥18 years
Missing
0.603
2012
Taiwan
≥65 years
Missing
1
Cases = 15 556 Controls = 69 097 Cases = 192 Controls = 1561 n = 9035
2012
Tarragona Spain Manitoba Canada
≥65 years
74.4 years (SD: 6.7) 61.83
0.7
n = 2650
1.001
n = 91 605
Wang Rodriguez Lau a
Cohort study with matching
2013
≥18 years
Sex ratio
Missing
Sample Cases = 37 Controls = 77 n = 36 906
Minnesota, Wisconsin, Oregon, Washington area, New-York, New Jersey, Pennsylvania, Connecticut.
4.3.
Outcomes
Table 3 shows the different outcomes; they included: (i) allcause hospitalization; (ii) hospitalization for one reason of admission or a precise reason of admission belonging to a pre-established group; (iii) hospitalization in a specific ward; (iv) death; (v) the union of the previous two outcomes (hospitalization or death). We counted eleven different outcomes, after taking into account all the articles’ primary and secondary outcomes. The outcomes were either classified using groups of codes from the International Classification of Diseases (ICD) referring to the cause of hospitalization, either grouped according to different types of wards. No outcome covering laboratory-confirmed influenza infection during hospitalization has been used in any study. Only one article [20] used hospitalizations for influenza-like illnesses as a primary outcome, relying on results from a previous study: this outcome was developed using data from emergency departments in Edmonton, Canada.
chronic bronchitis, emphysema, asthma, myocardial infarction, congestive heart failure, stroke or diabetes dysregualtion.
4.5. Vaccine effectiveness for diabetics aged 18–65 years old (Table 4) Lau et al. found that hospitalizations for pneumonia/influenza were lowered by 43% [28%–54%] and all-cause hospitalizations were lowered by 28% [24%–32%] for diabetics aged 18–64 years old but did not find a significant decrease in hospitalizations for influenza-like illnesses. Looijmans et al. found a decrease higher than 70% of hospitalizations for a specific cause (pneumonia, acute respiratory disease, prednisolonetreated chronic bronchitis, emphysema, asthma, myocardial infarction, congestive heart failure, stroke or diabetes dysregulation), but did not find a significant decrease in all-cause mortality. We did not find any study including diabetics aged 17 and less.
4.6. 4.4. Vaccine effectiveness for diabetics aged 65 and older (Table 4) All-cause death was an outcome used in four studies [19,21–23], it was thus the outcome that was most analyzed. The risk reduction was highest for all-cause death of vaccinated diabetics, going from 33% CI95 [4%–54%] upto 68% CI95 [58%–75%]. Three studies analyzed the risk reduction for all-cause hospitalization [21,23,20] and there was a significant risk reduction but lower than for all-cause death. Two studies assessed the risk of hospitalization for pneumonia or influenza and found significant but very different risk reductions: respectively 14% [3%–23%] [23] and 45% [34%–53%] [20]. The other eight outcomes were always used in only one study, making comparisons of results impossible. Nonetheless there was a risk reduction for all these outcomes except for hospitalizations for specific causes in Looijmans et al., where there was no significant risk reduction for vaccinated diabetics aged 65 and more. This criterion included hospitalizations for pneumonia, acute respiratory disease, prednisolone-treated
Identified biases
Using indirect outcomes represented a rating bias that was common to all the studies. Unspecific outcomes measured occurrences not linked with influenza. For example, Colquhoun et al. who had measured a significant risk reduction of hospitalization for vaccinated diabetics showed that 86% of the admissions for the hospitalizations were linked to diabetes dysregulation. Likewise, Looijmans et al. who had used hospitalizations for multiple causes as an outcome clarified that most of the observed occurrences were secondary to diabetes dysregulation. The outcome that seemed the most specific was hospitalization for influenza-like illnesses. It was a composite criterion of the ICD-10 codes used by Lau et al. [20]. This criterion was defined depending on the different reasons for admissions for the hospitalizations of patients sent to the emergency rooms (in Edmonton, Canada), with the primary diagnosis of influenza established by the doctors sending these patients to the hospitals. It was for this outcome that the estimates of risk reduction were the lowest: 13% [10%–16%] for the diabetics aged 65 and higher, and non significant decrease for diabetics aged 64 and less.
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
First author
Flu season
Circulating influenza strains predominated by
Vaccine type (strains, year)
Study period
Control period
Adjusted variables
Age, sex, epidemic year, type and duration of diabetes, comorbidity, number of medical visits in the previous year, number of admissions to hospital in the same period and interactions of these variables Age, sex, comorbidity medical conditions, prior health care use, previously hospitalized for pneumonia or influenza. None
Colquhoun
1989–1990 1993–1994
A/Shanghai/11/87 (H3N2) A/Beijing/32/92 (H3N2)
Not available
Influenza season according to the Royal college of general practitioners
None
Multivariate logistic regression model
Hak
1996–1997 (62.9%) 1997–1998 (63.5%)
H3N2 H3N2
Not available
Influenza season according to Centers for Disease Control and Prevention
None
Multivariate logistic regr + G3:K3ession model
Heymann
2000–2001 (48.8%)
Missing
Not available
1999–2000
H3N1 + H3N2 (Sydney Type)
Trivalent 1999–2000 season ref = 16
Yes Summer period None
Univariate logistic
Looijmans
Influenza season non justified Influenza season non justified
Wang
2001–2009
Multiple
Missing
All year-round non justified
None
Rodriguez
2002–2003 (59.80%) 2003–2004 (65.80%) 2004–2005 (70.80%) 2005–2006 (71.30%)
Multiple
Missing
Influenza season not justified
Yes Summer period
Cox proportional hazards regression model Logistic discrete time-hazard regression modelling
Lau
2000–2008
Multiple
Missing
Influenza season according to provincial surveillance data
Yes Pre and post epidemic period
Mulitvariate logistic regression model
Poisson regression models
Age, sex, health care insurance, presence of heart or lung disease, high-risk disease, number of medications, number of medical visits in the previous year Age, sex, comorbidity
To adjust for potential selection bias, they calculated a propensity score by using a logit model with the dependent variable influenza vaccine status Sex, age, income, pneumococcal vaccination receipt, number of medical visits in the previous year, comorbid health status
ARTICLE IN PRESS
Statistic model
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
5
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
Table 2 – Characteristics of included studies.
PCD-533; No. of Pages 9
ARTICLE IN PRESS
6
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
Table 3 – Outcomes used in the included studies. Author, year Colquhoun, 1997
Hak, 2002 Heymann, 2004
Looijmans, 2006
Wang, 2012
Rodriguez, 2012 Lau, 2013
a b
Outcomes 1. Hospitalization for primary viral and bacterial and acute and chronic bronchitis, ketocidosis or diabetic coma 1. Hospitalization for pneumonia/influenza or all cause death 1. Hospitalization in internal medicine and geriatric wards 2. All-cause death 1. Hospitalization for pneumonia, acute respiratory disease, prednisolone-treated chronic bronchitis, emphysema, asthma, myocardial infarction, congestive heart failure, stroke or diabetes dysregulation 2. All-cause death 3. Both
Classification
Code
ICD-9-CMa available
466, 480.8–483, 484.8–487, 490, 491, 250–250.2, 250.7, 250.9
ICD-9-CM available
480–487 or death
Useless
Useless
Useless ICPC available
Useless R80, R81, R78, R91, R95, R96, K75, K77, K90, T90
Useless ICPC available
Useless R80, R81, R78, R91, R95, R96, K75, K77, K90, T90 or death Useless 480–487
1. Total hospitalizations 2. Hospitalization for pneumonia/influenza 3. Hospitalization for respiratory failure 4. Hospitalization in an intensive care unit 5. All-cause death 1. All-cause death 1. Hospitalization for influenza-like illnesses
Useless ICD-9-CM available
2. Hospitalization for pneumonia/influenza 3. Total hospitalizations
ICD-10-CA available
518.81–518.84, 799.1 Useless Useless Useless J00, J01, J32, J02, J04, J06.8, J06.9, J09, J10, J11, J12, J20, J40, J44.8, J21, J22, J44.0, J441, R05, R09.1, J13, J14, J15, J16, J18 J10, J11, J12, J13, J14, J15, J16, J18
Useless
Useless
ICD-9-CM available Useless Useless Useless ICD-10-CAb available
9th revision of the International Statistical Classification of Diseases and Related Health Problems. 10th revision of the International Statistical Classification of Diseases and Related Health Problems.
The healthy vaccinee bias also represented a selection bias that was common to all the studies. It stems from the fact that vaccinated people are more likely than others to adopt behaviours in order to be health-conscious and thus be healthier. The vaccination bias leads to an overestimation of VE. This difference in behaviours can be explained through unobserved factors such as the clinical and follow-ups, compliance to treatment recommendations, drug compliance, and is linked to the social and financial statuses [24–26]. None of the seven studies could neutralize nor lessen this healthy vaccinee bias [18]. This healthy vaccinee bias was illustrated by Rodriguez et al. and Lau et al., who showed outcomes differences between groups of vaccinated and non-vaccinated diabetics during influenza off-season. Thus, the VE estimates of the influenza vaccine were partly linked to nonobserved confounding factors. A confounding bias linked to previous pneumoccal vaccinations is also possible. Six out of seven studies made no adjustment on the pneumoccal vaccine. The compliances to the influenza vaccine and the pneumoccal vaccine share common factors [29]. Hence there is a higher pneumoccal vaccination rate in people vaccinated against influenza. Rodriguez et al. found that 70% of diabetics vaccinated against influenza and 22% of non-vaccinated diabetics also had had the pneumococcal vaccine (while looking back on 5 years before their analysis). One of the most frequently described forms of a pneumococcal infection is a respiratory tract
infection: pneumococcal and influenza infections share common symptoms. Furthermore, one of the most frequent complications of a influenza infection is a pulmonary pneumococcal infection. Most of the outcomes listed above included pneumonias, this major confounding bias was hence not taken into account. The indication bias was a minor bias. This bias can be explained by the fact that patients having more severe symptoms are more likely to be vaccinated. Only two out seven studies have done an adjustment on the severity of diabetes: Colquhoun et al. adjusted their analyses on the duration of diabetes in three categories and Akker et al. adjusted theirs on the number of medications, considered as a proxy variable of the diabetes’ severity. All the articles, except for Heymann et al., took into account cardiovascular or renal comorbidities. For all the studies that were identified, information from medical and administrative databases were used for the samples of diabetic patients, identification of the vaccination status and for defining outcomes. These databases allowed the correct designation of diabetic patients and vaccinated people or not. However, defining outcomes solely on diagnosis codes worsened their already low specificity. Classification errors are possible due to selective tests or if the diagnosis was only made from clinical observations. Laboratory confirmation has been advised as a basic requirement for studies assessing the efficiency or effectiveness of a vaccine against influenza.
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS 7
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
Table 4 – Influenza vaccine’s effectiveness measured in terms of risk reduction (%) for each studied outcome. Outcomes
All-cause mortality
Total hospitalizations
Hospitalization for pneumonia/influenza Hospitalization for pneumonia, acute respiratory disease, prednisolone-treated chronic bronchitis, emphysema, asthma, myocardial infarction, congestive heart failure, stroke or diabetes dysregulation or death Hospitalization for pneumonia/influenza or all-cause death in 96–97 Hospitalization for pneumonia/influenza or all-cause death in 97–98 Hospitalization for pneumonia, acute respiratory disease, prednisolone-treated chronic bronchitis, emphysema, asthma, myocardial infarction, congestive heart failure, stroke or diabetes dysregulation Hospitalization for primary viral and bacterial and acute and chronic bronchitis, ketocidosis or diabetic coma Hospitalization for respiratory failure Hospitalization in an intensive care unit Hospitalization for influenza-like illnesses
5.
Authors
18–65 years
>65 years
Adjusted risk reduction |IC95%]
p
26% [(−760)–93]
0.819
Adjusted risk reduction [IC95%]
p
67.9% [58.3–75.3] 44% [20–96] 33% [4–53] 56% [46–64] 33% [30–36] 12% [4–19] 13% [3–23] 14% [3–23] 45% [34–53] 39% [(−5)–65]
Missing 0.039 0.031 <0.01 <0.001 <0.01 Missing <0.05 <0.001 0.076
Hak
50% [37–60]
<0.001
Hak
21% [6–34]
0.009
14% [(−88)–60]
0.706
Colquhoun
79% [19–95]
Missing
Wang Wang Lau
33% [14–48] 70% [53–81] 13% [10–6]
<0.01 <0.01 <0.001
Heymann Looijmans Rodriguez Wang Lau Wang Heymann Wang Lau Looijmans
Looijmans
Discussion
Many outcomes based on morbimortality have measured the seasonal influenza vaccine’s effectiveness on patients with diabetes aged 65 and higher. These outcomes have found a decrease in morbimortality for vaccinated diabetics aged 65 and higher. For working-age patients with diabetes, studies are rare and the effectiveness measured sometimes non significant for some outcomes. Several biases restricted the demonstration of the seasonal influenza vaccine’s effectiveness for patients with diabetes. The influenza vaccination guidelines and recommendations having been published before any clinical trials could have proven the vaccine’s efffectiveness for diabetic patients. The recommendations were retrospectively justified from observational studies’ results. There was the healthy vaccinee bias in these studies, which led to an overestimation of the VE. Heymann and Wang described that mortality was lowered by 68% and 56% respectively. Logically the influenza vaccine should only lower all-cause mortality only by impacting the mortality excess caused by the influenza during influenza season. However, studies have shown that excess mortality linked to the influenza is around 5–10% [1,27] during winters. It is hence not plausible that seasonal influenza vaccination of diabetic patients lowers all-cause mortality by 50% during winter season. This statement can probably be explained by non-controlled confounding factors in those studies.
28% [24–32]
<0.001
43% [28–54] 72% [46–85]
<0.001 <0.001
70% [39–85]
1% [(−1)–3]
0.001
0.402
Particularly, mastering the healthy vaccinee bias is a key methodological aspect to work on in future articles. Studies have tried to identify differences between groups vaccinated and not vaccinated against influenza [26,28]. Amongst potential confounding factors, male gender, age, to have comorbidities and the number of consultations with a general practitioner have been taken into account in most of the studies included in this review (Table 2). Other variables linked to the quality of the follow-up of diabetes and to patients’ financial and social statuses have not been factored in. Low incomes [29], the profession [30], a low level of education [31] are connected to poor compliance with the influenza vaccine schedule. A Canadian study shows that vaccinated diabetics have a better follow-up and a better treatment compliance [24]. A French study shows that a low income is linked to a lower chance of being vaccinated [29]. Finally, diabetics who have a high risk of infections (a combination of comorbidities and high levels of glycosylated haemoglobin) do not properly follow the vaccine schedules for vaccinations against influenza, hepatitis B and pneumococcal vaccines [25]. Thus, the healthy vaccine bias could be reduced by an adjustment of the variables linked to the quality of the follow-up of diabetes and the financial and social statuses of diabetic patients. The second important common limit in all the studies was the use of numerous, indirect and unspecific outcomes. The clinical diagnosis of a influenza infection is imprecise, having low clinical specificity and sensibility. Lab confirmation by PCR is the gold standard but using this test is costly and difficult
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS
8
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
due to the high number of subjects needed in studies for vaccine effectiveness. However, the specificity varies depending on the outcomes used in the studies. The use of composite criteria may be a useful alternative. This was done by Lau and colleagues albeit with a non-validated set of composite criteria. New diagnostic tests should measure this composite criteria’s sensibility and specificity in order to be considered as indicators of the incidence of influenza cases. The residual confounding between vaccinated and nonvaccinated diabetics has been measured by analysing the persistence of a decrease of morbimortality outside of the influenza season. Out of the three studies that have done this, only one observed the persistence of a decrease of morbimortality for vaccinated diabetics outside of the influenza season. It would be interesting in future works to use this approach in order to test the effect of adjustment of the healthy vaccine bias. The savage influenza strains and the strains used in influenza vaccines change yearly [32], thus explaining the changing degrees of severity of the influenza epidemics and the different population’s sub-groups most concerned. Furthermore, the concordance between savage strains and vaccine strains fluctuate (Rodriguez et al.). It would be advised to study the seasonal influenza vaccine effectiveness during many influenza seasons, this being the case for five out of seven included studies.
5.1.
Strengths and weaknesses of the review
This review is the second systematic review that has analyzed the seasonal influenza vaccine’s effectiveness for diabetic patients. Compared to the first review, we have chosen to include only articles published as full articles in order to analyze the chapter Methods. We have limited the search equation to studies specifically focusing on diabetics, compared to the first systematic review. The choice was made to not include studies in the general population that did not have a specific analysis of the diabetic population. Given the high prevalence of diabetes mellitus and the diabetic population’s specificities, a study protocol should be specifically made for this population. The previous review had identified eleven studies. A complete study was excluded because there was only one analysis in the diabetic sub-group of the general population of the study, and three short publications were excluded because the protocol and methods were not available (letters to editors or short communication). This study’s strength is to focus on the methodological limits and biases while showcasing in detail the outcomes, the adjustment variables and the statistical models that were used. Our detailed analysis of the methodological limits of previous studies of the seasonal influenza vaccine’s effectiveness allows us to offer precise suggestions of protocol improvements of future observational studies on this topic, knowing that clinical trials will not be possible. Using medical and administrative databases presents numerous advantages.
5.2.
Consequences for medical practice
Doctors, either working in hospitals or independently offer the seasonal influenza vaccination on an annual basis. Despite
this, the level of proof of its effectiveness is low. Improving the awareness about the influenza VE is a prerequisite for the vaccine to be accepted by the general population.
6.
Conclusion
WHO and numerous countries’ health authorities recommend a seasonal influenza vaccination for diabetic patients and people aged 65 and higher. There is currently no other primary preventive measure and this vaccine has a good safety record [32]. Thus, in order to improve the global health care provided for patients, studies with higher levels of evidence, especially randomized clinical trials, are more and more necessary.
Conflict of interest We confirm that each of the listed authors meets the authorship requirements as stated in the Uniform Requirements for Manuscripts Submitted to Biomedical Journals, and that no potential conflicts of interest relevant to this article were reported.
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.pcd.2016.05.005.
references
[1] L. Simonsen, R.J. Taylor, C. Viboud, M.A. Miller, L.A. Jackson, Mortality benefits of influenza vaccination in elderly people: an ongoing controversy, Lancet Infect. Dis. 7 (10) (2007) 658–666. [2] M.T. Osterholm, N.S. Kelley, A. Sommer, E.A. Belongia, Efficacy and effectiveness of influenza vaccines: a systematic review and meta-analysis, Lancet Infect. Dis. 12 (1) (2012) 36–44. [3] P. Doshi, Influenza: marketing vaccine by marketing disease, BMJ 346 (May (1)) (2013), f3037–f3037. [4] C. Remschmidt, O. Wichmann, T. Harder, Vaccines for the prevention of seasonal influenza in patients with diabetes: systematic review and meta-analysis, BMC Med. 13 (2015) 53. [5] A.G. Bertoni, S. Saydah, F.L. Brancati, Diabetes and the risk of infection-related mortality in the US, Diab. Care 24 (6) (2001) 1044–1049. [6] M. Delamaire, D. Maugendre, M. Moreno, M.-C. Le Goff, H. Allannic, B. Genetet, Impaired leucocyte functions in diabetic patients, Diab. Med. 14 (1) (1997) 29–34. [7] K.P. Bouter, R.J.A. Diepersloot, L.K.J. van Romunde, R. Uitslager, N. Masurel, J.B.L. Hoekstra, et al., Effect of epidemic influenza on ketoacidosis, pneumonia and death in diabetes mellitus: a hospital register survey of 1976-1979 in The Netherlands, Diab. Res. Clin. Pract. 12 (1) (1991) 61–68. [8] R. Valdez, K.M. Narayan, L.S. Geiss, M.M. Engelgau, Impact of diabetes mellitus on mortality associated with pneumonia and influenza among non-Hispanic black and white US adults, Am. J. Publ. Health 89 (11) (1999) 1715–1721. [9] Q. Zhu, H. Chang, Y. Chen, F. Fang, C. Xue, F. Zhang, et al., Protection of inactivated influenza virus vaccine against
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
lethal influenza virus infection in diabetic mice, Biochem. Biophys. Res. Commun. 329 (1) (2005) 87–94. B.J. Feery, L.J. Hartman, A.W. Hampson, J. Proietto, Influenza immunization in adults with diabetes mellitus, Diab. Care 6 (5) (1983) 475–478. P. Pozzilli, E.A. Gale, N. Visalli, M. Baroni, P. Crovari, V. Frighi, et al., The immune response to influenza vaccination in diabetic patients, Diabetologia 29 (12) (1986) 850–854. J.E. McElhaney, M.J. Pinkoski, D. Au, K.E. Lechelt, R.C. Bleackley, G.S. Meneilly, Helper and cytotoxic T lymphocyte responses to influenza vaccination in healthy compared to diabetic elderly, Vaccine 14 (6) (1996) 539–544. J.S. Nam, A.R. Kim, J.C. Yoon, Y. Byun, S.A. Kim, K.R. Kim, et al., The humoral immune response to the inactivated influenza A (H1N1) 2009 monovalent vaccine in patients with Type 2 diabetes mellitus in Korea, Diab. Med. 28 (7) (2011) 815–817. M.M. Lederman, H.M. Rodman, B.Z. Schacter, P.K. Jones, G. Schiffman, Antibody response to pneumococcal polysaccharides in insulin-dependent diabetes mellitus, Diab. Care 5 (1) (1982) 36–39. OMS|Grippe (saisonnière) [Internet], 2015 [cité 15 janv 2015]. Disponible sur: http://www.who.int/mediacentre/ factsheets/fs211/fr/. P. Tuppin, S. Choukroun, S. Samson, A. Weill, P. Ricordeau, H. Allemand, Vaccination contre la grippe saisonnière en France en 2010 et 2011: diminution des taux de couverture et facteurs associés, La Presse Médicale 41 (11) (2012) e568–e576. D. Moher, A. Liberati, J. Tetzlaff, D.G. Altman, PRISMA. Group, Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement, BMJ 339 (2009) b2535. Guide d’analyse de la littérature et gradations des recommandations [Internet]. Agence nationale d’accréditation et d’évaluation de la santé, 2000 [cité 15 janv 2015]. Disponible sur: http://www.has-sante.fr/portail/ upload/docs/application/pdf/analiterat.pdf. I.L.-V. Akker, T.J.M. den Verheij, E. Buskens, K.L. Nichol, G.E.H.M. Rutten, E. Hak, Clinical effectiveness of first and repeat influenza vaccination in adult and elderly diabetic patients, Diab. Care 29 (8) (2006) 1771–1776. D. Lau, D.T. Eurich, S.R. Majumdar, A. Katz, J.A. Johnson, Effectiveness of influenza vaccination in working-age adults with diabetes: a population-based cohort study, Thorax 68 (7) (2013) 658–663. A.D. Heymann, Y. Shapiro, G. Chodick, V. Shalev, E. Kokia, E. Kramer, et al., Reduced hospitalizations and death associated with influenza vaccination among patients with and without diabetes, Diab. Care 27 (11) (2004) 2581–2584. T. Rodriguez-Blanco, A. Vila-Corcoles, C. de Diego, O. Ochoa-Gondar, E. Valdivieso, F. Bobe, et al., Relationship
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
9
between annual influenza vaccination and winter mortality in diabetic people over 65 years, Hum. Vac. Immunother. 8 (3) (2012) 363–370. I.-K. Wang, C.-L. Lin, Y.-C. Chang, P.-C. Lin, C.-C. Liang, Y.-L. Liu, et al., Effectiveness of influenza vaccination in elderly diabetic patients: a retrospective cohort study, Vaccine 31 (4) (2013) 718–724. K.A. Achtymichuk, J.A. Johnson, F. Al Sayah, D.T. Eurich, Characteristics and health behaviors of diabetic patients receiving influenza vaccination, Vaccine 33 (30) (2015) 3549–3555. M.J. Alcusky, J. Pawasauskas, Adherence to guidelines for hepatitis B, pneumococcal, and influenza vaccination in patients with diabetes, Clin. Diab. 33 (3) (2015) 116–122. I. Jimenez-Trujillo, A. López-de Andrés, V. Hernández-Barrera, P. Carrasco-Garrido, J.M. Santos-Sancho, R. Jiménez-García, Influenza vaccination coverage rates among diabetes sufferers, predictors of adherence and time trends from 2003 to 2010 in Spain, Hum. Vac. Immunother. 9 (6) (2013) 1326–1332. L. Simonsen, T.A. Reichert, C. Viboud, W.C. Blackwelder, R.J. Taylor, M.A. Miller, Impact of influenza vaccination on seasonal mortality in the US elderly population, Arch. Intern. Med. 165 (3) (2005) 265–272. H. Lai, W.S. Aronow, A.H. Gutwein, Prevalence of influenza vaccination and pneumococcal vaccination in elderly and high-risk patients seen in a university general medicine clinic, Am. J. Ther. 15 (6) (2008) 528–530. P. Verger, S. Cortaredona, C. Pulcini, L. Casanova, P. Peretti-Watel, O. Launay. Characteristics of patients and physicians correlated with regular influenza vaccination in patients treated for type 2 diabetes: a follow-up study from 2008 to 2011 in southeastern France. Clin. Microbiol. Infect. [Internet] [cité 2 juill 2015];0(0). Disponible sur: http://www.clinicalmicrobiologyandinfection.com/article/ S1198743X1500662X/abstract. P.M.S.B. Francisco, M.B. Barros, A. de, M.R.D. Cordeiro, Influenza vaccination among elders: prevalence, associated factors, and reasons for noncompliance in Campinas, São Paulo State, Brazil, Cad Saude Publ. 27 (3) (2011) 417–426. I. Jimenez-Trujillo, R. Jiménez-García, J. Esteban-Hernández, V. Hernández-Barrera, P. Carrasco Garrido, M.A. Salinero-Fort, et al., Predictors of adherence to multiple clinical preventive recommendations among adults with diabetes in Spain, PLoS ONE 10 (6) (2015) e0131844. HCSP, Vaccination contre la grippe saisonnière chez les personnes âgées et les professionnels de santé [Internet], Haut Conseil de la Santé Publique, Paris, 2014 mars [cité 17 nov 2015]. Disponible sur: http://www.hcsp.fr/explore. cgi/avisrapportsdomaine?clefr=424.
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
Contents lists available at ScienceDirect
Primary Care Diabetes journal homepage: http://www.elsevier.com/locate/pcd
Review
Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics Ludovic Casanova a,b,c,∗ , Nirvina Gobin c , Patrick Villani a,b,d , Pierre Verger a,b,d a
INSERM, UMR S 912, «Sciences Economiques & Sociales de la Santé et Traitement de l’Information Médicale» (SESSTIM), F-13385 Marseille, France b ORS PACA, Observatoire Régional de la Santé Provence-Alpes-Côte d’Azur, F-13006 Marseille, France c Aix Marseille University, Department of General Practice, Marseille, France d Aix Marseille Université, UMR S 912, IRD, Marseille, F-13385 Marseille, France
a r t i c l e
i n f o
a b s t r a c t
Article history:
Background: The influenza virus is an important cause of morbidity and mortality for diabet-
Received 3 March 2016
ics. The seasonal influenza vaccine’s immunologic effectiveness is proven within the type
Accepted 16 May 2016
1 and type 2 diabetic populations, but the level of evidence is low. This article presents a
Available online xxx
systematic review for the bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics.
Keywords:
Methods: Using systematic review methods, we searched three electronic databases for pub-
Bias
lished literature (MEDLINE, EMBASE and the Cochrane Library) and two grey literature (SIGLE
Diabetes mellitus
and NHS EED) databases, to identify studies published between 1997 and 2013, examining
Influenza, Human
the effect of seasonal influenza vaccination, among diabetics, on any measure for influenza
Influenza vaccines
morbidity or mortality. Results: 725 records were identified from the three databases and screening, short-listing was undertaken independently by two reviewers. After de-duplication, all records were screened by title and then abstract, and 34 short-listed records were reviewed in full, with 7 studies included: 4 cohort studies and 3 case–control studies, conducted in 7 countries. The most common outcome of interest in studies (n = 4) was all-cause mortality among elderly diabetics (>65 years), with individual studies reporting reductions in risk of between 33% [95%CI: 4%–54%] and 68% [95%CI: 58%–75%]. We found only two studies for working-age adult diabetics: one reporting that vaccination prevented hospitalizations due to pneumonia or influenza (vaccine effectiveness [VE] 43%, [95%CI: 28%–54%]) and all-cause hospitalizations (VE: 28% [95%CI: 24%–32%]); and, another reporting no significant decrease in all-cause mortality for working-age adult diabetics. We have identified three major biases: the use of indirect health outcomes, a risk of selection bias (health-seeking bias), and no adjustment for participant pneumococcal vaccination status. The most recent included article finds that morbimortality is still lower during off-season influenza in both vaccinated and non-vaccinated diabetics, indicating important residual confounding.
∗ Corresponding author at: Observatoire régional de la santé Provence-Alpes-Côte d’Azur, 23 rue Stanislas Torrents, 13006 Marseille, France. Tel.: +33 4 91 59 89 14; fax: +33 4 91 59 89 18. E-mail address: [email protected] (L. Casanova). http://dx.doi.org/10.1016/j.pcd.2016.05.005 1751-9918/© 2016 Primary Care Diabetes Europe. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS
2
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
Conclusion: To date, the strength of evidence supporting the routine use of seasonal influenza vaccination is low for diabetics older than 65, and very low for working-age diabetics. © 2016 Primary Care Diabetes Europe. Published by Elsevier Ltd. All rights reserved.
Contents 1. 2. 3.
4.
5.
6.
1.
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .00 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 3.1. Literature research and search strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 3.2. Study eligibility and data synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 3.3. Selection, quality assessment, and data extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 4.1. Inclusion of articles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .00 4.2. Studies’ characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 4.3. Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 4.4. Vaccine effectiveness for diabetics aged 65 and older (Table 4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 4.5. Vaccine effectiveness for diabetics aged 18–65 years old (Table 4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 4.6. Identified biases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 5.1. Strengths and weaknesses of the review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 5.2. Consequences for medical practice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Conflict of interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 Appendix A. Supplementary data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
Background
Seasonal influenza is an important cause of morbidity and mortality among diabetics. The World Health Organization recommends that this high risk group properly follow the vaccination schedule. However, the national influenza vaccination campaigns have been criticized [1–3], partly because the strength of evidence is weak regarding the influenza vaccine effectiveness (VE) in certain high risk groups. In fact, with many ethical committees worldwide advising to proceed with caution within diabetic populations, no randomized controlled trial is currently available. A recent systematic review and meta-analysis called for more research to assess effectiveness of seasonal influenza vaccination among diabetics [4].
2.
Introduction
Diabetic patients have an elevated rate of morbidity and mortality linked to infections, [5] as chronic hyperglycaemia can disturb white blood cell functions, such as phagocytosis, chemotaxis and cell adhesion [6]. Diabetic patients have a higher chance of developing pulmonary complications of an influenza infection [7] and their risk of mortality is two to four times higher as compared to non-diabetics [8]. The objective immune response resulting from the
influenza vaccine has been widely studied [9–12] and, more recently in 2011, for the serotype H1N1 [13], because of the immune system’s disorders due to diabetes. The results are consistent whatever the primary outcome of the immune studies (the antibody response among vaccinated, quantitative antibody titres, in vitro humoral and cell-mediated immune responses): the objective immune response by the seasonal influenza vaccine in diabetics patients is similar to non diabetics’ response. The same conclusions were drawn for the pneumococcal vaccine [14]. Although it is plausible that the seasonal influenza vaccine would offer greater protection to diabetics, this has not been directly proven. The public’s demands make it more necessary to prove the VE. The avian influenza epidemic (H5N1) in 2004 and the swine influenza pandemic (H1N1) in 2009–2010 have raised demand for influenza vaccines and many health authorities worldwide have mobilized to provide vaccination for vulnerable populations. Some countries, such as France, have developed prevention measures considered expensive and even unnecessary by the public and some health representatives. Thus ensued a certain level of scepticism from patients, conveyed by insufficient influenza vaccination, even more so by working-age diabetics [15,16]. Doubts on VE that could exist for diabetics patients show that the level of evidence has to improve. In 2015, the ethical committees maintain their refusal to conduct randomized controlled trials. Observational studies are the only solution to improve the level of evidence.
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS 3
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
The aim of this systematic review was to identify studies reporting the VE of seasonal influenza vaccine on morbidity and mortality, and to assess and discuss the quality of included studies.
3.
Records identified through database search Medline = 439 Embase = 286 Cochrane = 0
Methods Titles screened (n = 725)
Our methods are reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement guidelines [17] and checklist is included as an annex (Appendix A).
3.1.
Study eligibility and data synthesis
The population of interest was individuals with Type 1 or Type 2 Diabetes at any age. The intervention was to be the seasonal influenza vaccine. The control group were either those not vaccinated or those receiving a placebo. Any type of study design could have been used to compare the groups, be it case–control studies, cohort studies, quasi-experimental studies or clinical trials. Every criteria of morbimortality was taken into account. No restrictions were applied for the date of publication or language the article was published in.
3.3.
Abstracts screened (n = 120)
Literature research and search strategy
Five databases were searched using MeSH terms and keyword searches with search terms adapted according to the requirements of the databases (Annex 1). Three databases (Medline, Embase and the Cochrane Library) were searched for published literature, and two databases (System for Information of Grey Literature in Europe [SIGLE] and National Health System Economic Evaluation Database [NHS EED]) were searched for grey literature. No particular kind of software was used to do the research. The searches were originally run on 9th September 2014, and then updated on 16th January 2015.
3.2.
Excluded = 605
Selection, quality assessment, and data extraction
Two reviewers (LC & NG) screened titles according to the PICO criteria. Letters to editors, posters and short communications were excluded. The articles were first chosen by their titles by each author separately, then by the abstracts of the previously identified articles. At the completion of each step, reviewers compared and agreed lists of eligible and short-listed records. Any disagreement was resolved by a third reviewer (PV). Then, full articles were reviewed by both reviewers to check that inclusion criteria were met. The selected articles were independently assessed by the two reviewers. The analysis guidelines of literature and the recommended gradation of the National Agency of accreditation and health assessment [18] were used for the articles’ assessments. After pooling the findings, any disagreement between the two authors were solved by consensus. A common synthesis of the extracted data from each article was made. The potential biases of each study were identified, classified and ranked.
Excluded = 85
Full-text articles assessed for eligibility (n = 35)
Articles included in systematic review n=7
Excluded = 28 No data in the subgroup of diabetic patients (n=16) No data about vaccine effectiveness (n=5) Methodology not available: short communication, letter, poster or congress communication (n=4) Study not obtainable (n=3)
Fig. 1 – Flow diagram of study selection procedure for the systematic review.
4.
Results
4.1.
Inclusion of articles
Seven articles were selected and analyzed out of 725 (Fig. 1). Those seven articles were available on the two databases used. No other article was found from manual research and analysis of grey literature. In January 2015, there was no previous systematic review on this topic in the Cochrane Library.
4.2.
Studies’ characteristics
Every study included was made in different countries. There were four cohort studies, two of which did matching between the vaccinated and non-vaccinated groups, and three case–control studies, one of which was nestled in a cohort (Table 1). Only three out of seven articles included diabetics aged less than 65 years old, only two of which did a specific analysis on this specific group. All the studies were done from information from medical and administrative databases (Table 2). One study proceeded to collect their data on a single influenza epidemic season (winter 1999–2000) whereas the others studied two to seven influenza outbreaks [20].
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS
4
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
Table 1 – Studies included. First author
Study design
Year
Setting/Country
Age limits
Age mean
Colquhoun
Case control study
1997
≥18 years
Cohort study
2002
≥65 years
52.58 estimated value Missing
1.28
Hak Heymann
Case control study
2004
Leicestershire England 10 statesa United States Israel
≥65 years
Missing
Missing
Looijmans
Case control study in cohort Cohort study with matching Cohort study
2006
Netherlands
≥18 years
Missing
0.603
2012
Taiwan
≥65 years
Missing
1
Cases = 15 556 Controls = 69 097 Cases = 192 Controls = 1561 n = 9035
2012
Tarragona Spain Manitoba Canada
≥65 years
74.4 years (SD: 6.7) 61.83
0.7
n = 2650
1.001
n = 91 605
Wang Rodriguez Lau a
Cohort study with matching
2013
≥18 years
Sex ratio
Missing
Sample Cases = 37 Controls = 77 n = 36 906
Minnesota, Wisconsin, Oregon, Washington area, New-York, New Jersey, Pennsylvania, Connecticut.
4.3.
Outcomes
Table 3 shows the different outcomes; they included: (i) allcause hospitalization; (ii) hospitalization for one reason of admission or a precise reason of admission belonging to a pre-established group; (iii) hospitalization in a specific ward; (iv) death; (v) the union of the previous two outcomes (hospitalization or death). We counted eleven different outcomes, after taking into account all the articles’ primary and secondary outcomes. The outcomes were either classified using groups of codes from the International Classification of Diseases (ICD) referring to the cause of hospitalization, either grouped according to different types of wards. No outcome covering laboratory-confirmed influenza infection during hospitalization has been used in any study. Only one article [20] used hospitalizations for influenza-like illnesses as a primary outcome, relying on results from a previous study: this outcome was developed using data from emergency departments in Edmonton, Canada.
chronic bronchitis, emphysema, asthma, myocardial infarction, congestive heart failure, stroke or diabetes dysregualtion.
4.5. Vaccine effectiveness for diabetics aged 18–65 years old (Table 4) Lau et al. found that hospitalizations for pneumonia/influenza were lowered by 43% [28%–54%] and all-cause hospitalizations were lowered by 28% [24%–32%] for diabetics aged 18–64 years old but did not find a significant decrease in hospitalizations for influenza-like illnesses. Looijmans et al. found a decrease higher than 70% of hospitalizations for a specific cause (pneumonia, acute respiratory disease, prednisolonetreated chronic bronchitis, emphysema, asthma, myocardial infarction, congestive heart failure, stroke or diabetes dysregulation), but did not find a significant decrease in all-cause mortality. We did not find any study including diabetics aged 17 and less.
4.6. 4.4. Vaccine effectiveness for diabetics aged 65 and older (Table 4) All-cause death was an outcome used in four studies [19,21–23], it was thus the outcome that was most analyzed. The risk reduction was highest for all-cause death of vaccinated diabetics, going from 33% CI95 [4%–54%] upto 68% CI95 [58%–75%]. Three studies analyzed the risk reduction for all-cause hospitalization [21,23,20] and there was a significant risk reduction but lower than for all-cause death. Two studies assessed the risk of hospitalization for pneumonia or influenza and found significant but very different risk reductions: respectively 14% [3%–23%] [23] and 45% [34%–53%] [20]. The other eight outcomes were always used in only one study, making comparisons of results impossible. Nonetheless there was a risk reduction for all these outcomes except for hospitalizations for specific causes in Looijmans et al., where there was no significant risk reduction for vaccinated diabetics aged 65 and more. This criterion included hospitalizations for pneumonia, acute respiratory disease, prednisolone-treated
Identified biases
Using indirect outcomes represented a rating bias that was common to all the studies. Unspecific outcomes measured occurrences not linked with influenza. For example, Colquhoun et al. who had measured a significant risk reduction of hospitalization for vaccinated diabetics showed that 86% of the admissions for the hospitalizations were linked to diabetes dysregulation. Likewise, Looijmans et al. who had used hospitalizations for multiple causes as an outcome clarified that most of the observed occurrences were secondary to diabetes dysregulation. The outcome that seemed the most specific was hospitalization for influenza-like illnesses. It was a composite criterion of the ICD-10 codes used by Lau et al. [20]. This criterion was defined depending on the different reasons for admissions for the hospitalizations of patients sent to the emergency rooms (in Edmonton, Canada), with the primary diagnosis of influenza established by the doctors sending these patients to the hospitals. It was for this outcome that the estimates of risk reduction were the lowest: 13% [10%–16%] for the diabetics aged 65 and higher, and non significant decrease for diabetics aged 64 and less.
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
First author
Flu season
Circulating influenza strains predominated by
Vaccine type (strains, year)
Study period
Control period
Adjusted variables
Age, sex, epidemic year, type and duration of diabetes, comorbidity, number of medical visits in the previous year, number of admissions to hospital in the same period and interactions of these variables Age, sex, comorbidity medical conditions, prior health care use, previously hospitalized for pneumonia or influenza. None
Colquhoun
1989–1990 1993–1994
A/Shanghai/11/87 (H3N2) A/Beijing/32/92 (H3N2)
Not available
Influenza season according to the Royal college of general practitioners
None
Multivariate logistic regression model
Hak
1996–1997 (62.9%) 1997–1998 (63.5%)
H3N2 H3N2
Not available
Influenza season according to Centers for Disease Control and Prevention
None
Multivariate logistic regr + G3:K3ession model
Heymann
2000–2001 (48.8%)
Missing
Not available
1999–2000
H3N1 + H3N2 (Sydney Type)
Trivalent 1999–2000 season ref = 16
Yes Summer period None
Univariate logistic
Looijmans
Influenza season non justified Influenza season non justified
Wang
2001–2009
Multiple
Missing
All year-round non justified
None
Rodriguez
2002–2003 (59.80%) 2003–2004 (65.80%) 2004–2005 (70.80%) 2005–2006 (71.30%)
Multiple
Missing
Influenza season not justified
Yes Summer period
Cox proportional hazards regression model Logistic discrete time-hazard regression modelling
Lau
2000–2008
Multiple
Missing
Influenza season according to provincial surveillance data
Yes Pre and post epidemic period
Mulitvariate logistic regression model
Poisson regression models
Age, sex, health care insurance, presence of heart or lung disease, high-risk disease, number of medications, number of medical visits in the previous year Age, sex, comorbidity
To adjust for potential selection bias, they calculated a propensity score by using a logit model with the dependent variable influenza vaccine status Sex, age, income, pneumococcal vaccination receipt, number of medical visits in the previous year, comorbid health status
ARTICLE IN PRESS
Statistic model
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
5
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
Table 2 – Characteristics of included studies.
PCD-533; No. of Pages 9
ARTICLE IN PRESS
6
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
Table 3 – Outcomes used in the included studies. Author, year Colquhoun, 1997
Hak, 2002 Heymann, 2004
Looijmans, 2006
Wang, 2012
Rodriguez, 2012 Lau, 2013
a b
Outcomes 1. Hospitalization for primary viral and bacterial and acute and chronic bronchitis, ketocidosis or diabetic coma 1. Hospitalization for pneumonia/influenza or all cause death 1. Hospitalization in internal medicine and geriatric wards 2. All-cause death 1. Hospitalization for pneumonia, acute respiratory disease, prednisolone-treated chronic bronchitis, emphysema, asthma, myocardial infarction, congestive heart failure, stroke or diabetes dysregulation 2. All-cause death 3. Both
Classification
Code
ICD-9-CMa available
466, 480.8–483, 484.8–487, 490, 491, 250–250.2, 250.7, 250.9
ICD-9-CM available
480–487 or death
Useless
Useless
Useless ICPC available
Useless R80, R81, R78, R91, R95, R96, K75, K77, K90, T90
Useless ICPC available
Useless R80, R81, R78, R91, R95, R96, K75, K77, K90, T90 or death Useless 480–487
1. Total hospitalizations 2. Hospitalization for pneumonia/influenza 3. Hospitalization for respiratory failure 4. Hospitalization in an intensive care unit 5. All-cause death 1. All-cause death 1. Hospitalization for influenza-like illnesses
Useless ICD-9-CM available
2. Hospitalization for pneumonia/influenza 3. Total hospitalizations
ICD-10-CA available
518.81–518.84, 799.1 Useless Useless Useless J00, J01, J32, J02, J04, J06.8, J06.9, J09, J10, J11, J12, J20, J40, J44.8, J21, J22, J44.0, J441, R05, R09.1, J13, J14, J15, J16, J18 J10, J11, J12, J13, J14, J15, J16, J18
Useless
Useless
ICD-9-CM available Useless Useless Useless ICD-10-CAb available
9th revision of the International Statistical Classification of Diseases and Related Health Problems. 10th revision of the International Statistical Classification of Diseases and Related Health Problems.
The healthy vaccinee bias also represented a selection bias that was common to all the studies. It stems from the fact that vaccinated people are more likely than others to adopt behaviours in order to be health-conscious and thus be healthier. The vaccination bias leads to an overestimation of VE. This difference in behaviours can be explained through unobserved factors such as the clinical and follow-ups, compliance to treatment recommendations, drug compliance, and is linked to the social and financial statuses [24–26]. None of the seven studies could neutralize nor lessen this healthy vaccinee bias [18]. This healthy vaccinee bias was illustrated by Rodriguez et al. and Lau et al., who showed outcomes differences between groups of vaccinated and non-vaccinated diabetics during influenza off-season. Thus, the VE estimates of the influenza vaccine were partly linked to nonobserved confounding factors. A confounding bias linked to previous pneumoccal vaccinations is also possible. Six out of seven studies made no adjustment on the pneumoccal vaccine. The compliances to the influenza vaccine and the pneumoccal vaccine share common factors [29]. Hence there is a higher pneumoccal vaccination rate in people vaccinated against influenza. Rodriguez et al. found that 70% of diabetics vaccinated against influenza and 22% of non-vaccinated diabetics also had had the pneumococcal vaccine (while looking back on 5 years before their analysis). One of the most frequently described forms of a pneumococcal infection is a respiratory tract
infection: pneumococcal and influenza infections share common symptoms. Furthermore, one of the most frequent complications of a influenza infection is a pulmonary pneumococcal infection. Most of the outcomes listed above included pneumonias, this major confounding bias was hence not taken into account. The indication bias was a minor bias. This bias can be explained by the fact that patients having more severe symptoms are more likely to be vaccinated. Only two out seven studies have done an adjustment on the severity of diabetes: Colquhoun et al. adjusted their analyses on the duration of diabetes in three categories and Akker et al. adjusted theirs on the number of medications, considered as a proxy variable of the diabetes’ severity. All the articles, except for Heymann et al., took into account cardiovascular or renal comorbidities. For all the studies that were identified, information from medical and administrative databases were used for the samples of diabetic patients, identification of the vaccination status and for defining outcomes. These databases allowed the correct designation of diabetic patients and vaccinated people or not. However, defining outcomes solely on diagnosis codes worsened their already low specificity. Classification errors are possible due to selective tests or if the diagnosis was only made from clinical observations. Laboratory confirmation has been advised as a basic requirement for studies assessing the efficiency or effectiveness of a vaccine against influenza.
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS 7
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
Table 4 – Influenza vaccine’s effectiveness measured in terms of risk reduction (%) for each studied outcome. Outcomes
All-cause mortality
Total hospitalizations
Hospitalization for pneumonia/influenza Hospitalization for pneumonia, acute respiratory disease, prednisolone-treated chronic bronchitis, emphysema, asthma, myocardial infarction, congestive heart failure, stroke or diabetes dysregulation or death Hospitalization for pneumonia/influenza or all-cause death in 96–97 Hospitalization for pneumonia/influenza or all-cause death in 97–98 Hospitalization for pneumonia, acute respiratory disease, prednisolone-treated chronic bronchitis, emphysema, asthma, myocardial infarction, congestive heart failure, stroke or diabetes dysregulation Hospitalization for primary viral and bacterial and acute and chronic bronchitis, ketocidosis or diabetic coma Hospitalization for respiratory failure Hospitalization in an intensive care unit Hospitalization for influenza-like illnesses
5.
Authors
18–65 years
>65 years
Adjusted risk reduction |IC95%]
p
26% [(−760)–93]
0.819
Adjusted risk reduction [IC95%]
p
67.9% [58.3–75.3] 44% [20–96] 33% [4–53] 56% [46–64] 33% [30–36] 12% [4–19] 13% [3–23] 14% [3–23] 45% [34–53] 39% [(−5)–65]
Missing 0.039 0.031 <0.01 <0.001 <0.01 Missing <0.05 <0.001 0.076
Hak
50% [37–60]
<0.001
Hak
21% [6–34]
0.009
14% [(−88)–60]
0.706
Colquhoun
79% [19–95]
Missing
Wang Wang Lau
33% [14–48] 70% [53–81] 13% [10–6]
<0.01 <0.01 <0.001
Heymann Looijmans Rodriguez Wang Lau Wang Heymann Wang Lau Looijmans
Looijmans
Discussion
Many outcomes based on morbimortality have measured the seasonal influenza vaccine’s effectiveness on patients with diabetes aged 65 and higher. These outcomes have found a decrease in morbimortality for vaccinated diabetics aged 65 and higher. For working-age patients with diabetes, studies are rare and the effectiveness measured sometimes non significant for some outcomes. Several biases restricted the demonstration of the seasonal influenza vaccine’s effectiveness for patients with diabetes. The influenza vaccination guidelines and recommendations having been published before any clinical trials could have proven the vaccine’s efffectiveness for diabetic patients. The recommendations were retrospectively justified from observational studies’ results. There was the healthy vaccinee bias in these studies, which led to an overestimation of the VE. Heymann and Wang described that mortality was lowered by 68% and 56% respectively. Logically the influenza vaccine should only lower all-cause mortality only by impacting the mortality excess caused by the influenza during influenza season. However, studies have shown that excess mortality linked to the influenza is around 5–10% [1,27] during winters. It is hence not plausible that seasonal influenza vaccination of diabetic patients lowers all-cause mortality by 50% during winter season. This statement can probably be explained by non-controlled confounding factors in those studies.
28% [24–32]
<0.001
43% [28–54] 72% [46–85]
<0.001 <0.001
70% [39–85]
1% [(−1)–3]
0.001
0.402
Particularly, mastering the healthy vaccinee bias is a key methodological aspect to work on in future articles. Studies have tried to identify differences between groups vaccinated and not vaccinated against influenza [26,28]. Amongst potential confounding factors, male gender, age, to have comorbidities and the number of consultations with a general practitioner have been taken into account in most of the studies included in this review (Table 2). Other variables linked to the quality of the follow-up of diabetes and to patients’ financial and social statuses have not been factored in. Low incomes [29], the profession [30], a low level of education [31] are connected to poor compliance with the influenza vaccine schedule. A Canadian study shows that vaccinated diabetics have a better follow-up and a better treatment compliance [24]. A French study shows that a low income is linked to a lower chance of being vaccinated [29]. Finally, diabetics who have a high risk of infections (a combination of comorbidities and high levels of glycosylated haemoglobin) do not properly follow the vaccine schedules for vaccinations against influenza, hepatitis B and pneumococcal vaccines [25]. Thus, the healthy vaccine bias could be reduced by an adjustment of the variables linked to the quality of the follow-up of diabetes and the financial and social statuses of diabetic patients. The second important common limit in all the studies was the use of numerous, indirect and unspecific outcomes. The clinical diagnosis of a influenza infection is imprecise, having low clinical specificity and sensibility. Lab confirmation by PCR is the gold standard but using this test is costly and difficult
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS
8
p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
due to the high number of subjects needed in studies for vaccine effectiveness. However, the specificity varies depending on the outcomes used in the studies. The use of composite criteria may be a useful alternative. This was done by Lau and colleagues albeit with a non-validated set of composite criteria. New diagnostic tests should measure this composite criteria’s sensibility and specificity in order to be considered as indicators of the incidence of influenza cases. The residual confounding between vaccinated and nonvaccinated diabetics has been measured by analysing the persistence of a decrease of morbimortality outside of the influenza season. Out of the three studies that have done this, only one observed the persistence of a decrease of morbimortality for vaccinated diabetics outside of the influenza season. It would be interesting in future works to use this approach in order to test the effect of adjustment of the healthy vaccine bias. The savage influenza strains and the strains used in influenza vaccines change yearly [32], thus explaining the changing degrees of severity of the influenza epidemics and the different population’s sub-groups most concerned. Furthermore, the concordance between savage strains and vaccine strains fluctuate (Rodriguez et al.). It would be advised to study the seasonal influenza vaccine effectiveness during many influenza seasons, this being the case for five out of seven included studies.
5.1.
Strengths and weaknesses of the review
This review is the second systematic review that has analyzed the seasonal influenza vaccine’s effectiveness for diabetic patients. Compared to the first review, we have chosen to include only articles published as full articles in order to analyze the chapter Methods. We have limited the search equation to studies specifically focusing on diabetics, compared to the first systematic review. The choice was made to not include studies in the general population that did not have a specific analysis of the diabetic population. Given the high prevalence of diabetes mellitus and the diabetic population’s specificities, a study protocol should be specifically made for this population. The previous review had identified eleven studies. A complete study was excluded because there was only one analysis in the diabetic sub-group of the general population of the study, and three short publications were excluded because the protocol and methods were not available (letters to editors or short communication). This study’s strength is to focus on the methodological limits and biases while showcasing in detail the outcomes, the adjustment variables and the statistical models that were used. Our detailed analysis of the methodological limits of previous studies of the seasonal influenza vaccine’s effectiveness allows us to offer precise suggestions of protocol improvements of future observational studies on this topic, knowing that clinical trials will not be possible. Using medical and administrative databases presents numerous advantages.
5.2.
Consequences for medical practice
Doctors, either working in hospitals or independently offer the seasonal influenza vaccination on an annual basis. Despite
this, the level of proof of its effectiveness is low. Improving the awareness about the influenza VE is a prerequisite for the vaccine to be accepted by the general population.
6.
Conclusion
WHO and numerous countries’ health authorities recommend a seasonal influenza vaccination for diabetic patients and people aged 65 and higher. There is currently no other primary preventive measure and this vaccine has a good safety record [32]. Thus, in order to improve the global health care provided for patients, studies with higher levels of evidence, especially randomized clinical trials, are more and more necessary.
Conflict of interest We confirm that each of the listed authors meets the authorship requirements as stated in the Uniform Requirements for Manuscripts Submitted to Biomedical Journals, and that no potential conflicts of interest relevant to this article were reported.
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.pcd.2016.05.005.
references
[1] L. Simonsen, R.J. Taylor, C. Viboud, M.A. Miller, L.A. Jackson, Mortality benefits of influenza vaccination in elderly people: an ongoing controversy, Lancet Infect. Dis. 7 (10) (2007) 658–666. [2] M.T. Osterholm, N.S. Kelley, A. Sommer, E.A. Belongia, Efficacy and effectiveness of influenza vaccines: a systematic review and meta-analysis, Lancet Infect. Dis. 12 (1) (2012) 36–44. [3] P. Doshi, Influenza: marketing vaccine by marketing disease, BMJ 346 (May (1)) (2013), f3037–f3037. [4] C. Remschmidt, O. Wichmann, T. Harder, Vaccines for the prevention of seasonal influenza in patients with diabetes: systematic review and meta-analysis, BMC Med. 13 (2015) 53. [5] A.G. Bertoni, S. Saydah, F.L. Brancati, Diabetes and the risk of infection-related mortality in the US, Diab. Care 24 (6) (2001) 1044–1049. [6] M. Delamaire, D. Maugendre, M. Moreno, M.-C. Le Goff, H. Allannic, B. Genetet, Impaired leucocyte functions in diabetic patients, Diab. Med. 14 (1) (1997) 29–34. [7] K.P. Bouter, R.J.A. Diepersloot, L.K.J. van Romunde, R. Uitslager, N. Masurel, J.B.L. Hoekstra, et al., Effect of epidemic influenza on ketoacidosis, pneumonia and death in diabetes mellitus: a hospital register survey of 1976-1979 in The Netherlands, Diab. Res. Clin. Pract. 12 (1) (1991) 61–68. [8] R. Valdez, K.M. Narayan, L.S. Geiss, M.M. Engelgau, Impact of diabetes mellitus on mortality associated with pneumonia and influenza among non-Hispanic black and white US adults, Am. J. Publ. Health 89 (11) (1999) 1715–1721. [9] Q. Zhu, H. Chang, Y. Chen, F. Fang, C. Xue, F. Zhang, et al., Protection of inactivated influenza virus vaccine against
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005
PCD-533; No. of Pages 9
ARTICLE IN PRESS p r i m a r y c a r e d i a b e t e s x x x ( 2 0 1 6 ) xxx–xxx
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
lethal influenza virus infection in diabetic mice, Biochem. Biophys. Res. Commun. 329 (1) (2005) 87–94. B.J. Feery, L.J. Hartman, A.W. Hampson, J. Proietto, Influenza immunization in adults with diabetes mellitus, Diab. Care 6 (5) (1983) 475–478. P. Pozzilli, E.A. Gale, N. Visalli, M. Baroni, P. Crovari, V. Frighi, et al., The immune response to influenza vaccination in diabetic patients, Diabetologia 29 (12) (1986) 850–854. J.E. McElhaney, M.J. Pinkoski, D. Au, K.E. Lechelt, R.C. Bleackley, G.S. Meneilly, Helper and cytotoxic T lymphocyte responses to influenza vaccination in healthy compared to diabetic elderly, Vaccine 14 (6) (1996) 539–544. J.S. Nam, A.R. Kim, J.C. Yoon, Y. Byun, S.A. Kim, K.R. Kim, et al., The humoral immune response to the inactivated influenza A (H1N1) 2009 monovalent vaccine in patients with Type 2 diabetes mellitus in Korea, Diab. Med. 28 (7) (2011) 815–817. M.M. Lederman, H.M. Rodman, B.Z. Schacter, P.K. Jones, G. Schiffman, Antibody response to pneumococcal polysaccharides in insulin-dependent diabetes mellitus, Diab. Care 5 (1) (1982) 36–39. OMS|Grippe (saisonnière) [Internet], 2015 [cité 15 janv 2015]. Disponible sur: http://www.who.int/mediacentre/ factsheets/fs211/fr/. P. Tuppin, S. Choukroun, S. Samson, A. Weill, P. Ricordeau, H. Allemand, Vaccination contre la grippe saisonnière en France en 2010 et 2011: diminution des taux de couverture et facteurs associés, La Presse Médicale 41 (11) (2012) e568–e576. D. Moher, A. Liberati, J. Tetzlaff, D.G. Altman, PRISMA. Group, Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement, BMJ 339 (2009) b2535. Guide d’analyse de la littérature et gradations des recommandations [Internet]. Agence nationale d’accréditation et d’évaluation de la santé, 2000 [cité 15 janv 2015]. Disponible sur: http://www.has-sante.fr/portail/ upload/docs/application/pdf/analiterat.pdf. I.L.-V. Akker, T.J.M. den Verheij, E. Buskens, K.L. Nichol, G.E.H.M. Rutten, E. Hak, Clinical effectiveness of first and repeat influenza vaccination in adult and elderly diabetic patients, Diab. Care 29 (8) (2006) 1771–1776. D. Lau, D.T. Eurich, S.R. Majumdar, A. Katz, J.A. Johnson, Effectiveness of influenza vaccination in working-age adults with diabetes: a population-based cohort study, Thorax 68 (7) (2013) 658–663. A.D. Heymann, Y. Shapiro, G. Chodick, V. Shalev, E. Kokia, E. Kramer, et al., Reduced hospitalizations and death associated with influenza vaccination among patients with and without diabetes, Diab. Care 27 (11) (2004) 2581–2584. T. Rodriguez-Blanco, A. Vila-Corcoles, C. de Diego, O. Ochoa-Gondar, E. Valdivieso, F. Bobe, et al., Relationship
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
9
between annual influenza vaccination and winter mortality in diabetic people over 65 years, Hum. Vac. Immunother. 8 (3) (2012) 363–370. I.-K. Wang, C.-L. Lin, Y.-C. Chang, P.-C. Lin, C.-C. Liang, Y.-L. Liu, et al., Effectiveness of influenza vaccination in elderly diabetic patients: a retrospective cohort study, Vaccine 31 (4) (2013) 718–724. K.A. Achtymichuk, J.A. Johnson, F. Al Sayah, D.T. Eurich, Characteristics and health behaviors of diabetic patients receiving influenza vaccination, Vaccine 33 (30) (2015) 3549–3555. M.J. Alcusky, J. Pawasauskas, Adherence to guidelines for hepatitis B, pneumococcal, and influenza vaccination in patients with diabetes, Clin. Diab. 33 (3) (2015) 116–122. I. Jimenez-Trujillo, A. López-de Andrés, V. Hernández-Barrera, P. Carrasco-Garrido, J.M. Santos-Sancho, R. Jiménez-García, Influenza vaccination coverage rates among diabetes sufferers, predictors of adherence and time trends from 2003 to 2010 in Spain, Hum. Vac. Immunother. 9 (6) (2013) 1326–1332. L. Simonsen, T.A. Reichert, C. Viboud, W.C. Blackwelder, R.J. Taylor, M.A. Miller, Impact of influenza vaccination on seasonal mortality in the US elderly population, Arch. Intern. Med. 165 (3) (2005) 265–272. H. Lai, W.S. Aronow, A.H. Gutwein, Prevalence of influenza vaccination and pneumococcal vaccination in elderly and high-risk patients seen in a university general medicine clinic, Am. J. Ther. 15 (6) (2008) 528–530. P. Verger, S. Cortaredona, C. Pulcini, L. Casanova, P. Peretti-Watel, O. Launay. Characteristics of patients and physicians correlated with regular influenza vaccination in patients treated for type 2 diabetes: a follow-up study from 2008 to 2011 in southeastern France. Clin. Microbiol. Infect. [Internet] [cité 2 juill 2015];0(0). Disponible sur: http://www.clinicalmicrobiologyandinfection.com/article/ S1198743X1500662X/abstract. P.M.S.B. Francisco, M.B. Barros, A. de, M.R.D. Cordeiro, Influenza vaccination among elders: prevalence, associated factors, and reasons for noncompliance in Campinas, São Paulo State, Brazil, Cad Saude Publ. 27 (3) (2011) 417–426. I. Jimenez-Trujillo, R. Jiménez-García, J. Esteban-Hernández, V. Hernández-Barrera, P. Carrasco Garrido, M.A. Salinero-Fort, et al., Predictors of adherence to multiple clinical preventive recommendations among adults with diabetes in Spain, PLoS ONE 10 (6) (2015) e0131844. HCSP, Vaccination contre la grippe saisonnière chez les personnes âgées et les professionnels de santé [Internet], Haut Conseil de la Santé Publique, Paris, 2014 mars [cité 17 nov 2015]. Disponible sur: http://www.hcsp.fr/explore. cgi/avisrapportsdomaine?clefr=424.
Please cite this article in press as: L. Casanova, et al., Bias in the measure of the effectiveness of seasonal influenza vaccination among diabetics, Prim. Care Diab. (2016), http://dx.doi.org/10.1016/j.pcd.2016.05.005