Influenza vaccine effectiveness – Patients with diabetes
Colquhoun, 1997 (ALL AGES)
- Estimate vaccine effectiveness in reducing hospital admissions for diabetic patients during periods when vaccine and wild strains were antigenically close.
- Case-control study.
- Data source
o Leicestershire Diabetes Register, which was specific for IDDM for the first influenza season studied (this bias should be similar in both cases and controls).
o Influenza epidemics = 1989-90, and 1993.
o Cases = admitted to hospital with ICD-9 codes for respiratory conditions and acute sequalae of diabetes. Exclusions of non-influenza etiologies (e.g.: adenovirus) where determined.
o Two controls per case were selected randomly from the Diabetes Register.
- Data collection
o GP notes of cases and controls
o Age, sex, duration of diabetes, other chronic medical disorders, number of previous GP consultations, influenza vaccination during the 3 years before the epidemic.
- 80 cases identified, but only 37 available for study.
- 166 controls selected, but only 77 available for study.
- Non-participation due mainly to lack of GP consent.
- Cases and controls similar WRT demographics co-morbidities, and HC utilization.
- Modal age band was 45-64, modal duration of diabetes = <9 years.
- More cases than controls had IDDM (49% vs 27%), likely due to sampling for acute diabetic complications as part of the outcome definition.
- Results
o Crude OR for hospital admission = 0.19 [0.05, 0.70].
o Hospital admission due to previous vaccination – OR = OR = 0.62 [0.15, 2.65] NS
o Authors suggest that this means no evidence of cross-reactivity. However, I believe this may indicate bias.
o Interactions NS during multiple logistic regression modeling – low numbers.
o Adjusted OR – hospital admission – age, sex, IDDM vs NIDDM, year of epidemic, and 12 month number of GP visits-adjusted model – OR = 0.21 [0.05, 0.81] – VE = 79%.
- No non-epidemic control period. Small sample size. Diabetic admissions may not have been influenza-attributable – outcome mis-classification?
- Authors: Results support the current UK guidelines for annual vaccination, including people with diabetes.
Hak E, Nordin J, Wei F, Mullooly J, Poblete S, Strikas R, Nichol KL. Influence of high-risk medical conditions on the effectiveness of influenza vaccination among elderly members of 3 large managed-care organizations. Clin Infect Dis, 35(4): 2002
- ELDERLY ONLY
- See Nichol et al. for methods (Nichol KL, Nordin J, Mullooly J, Lask R, Fillbrandt K, Iwane M. Influenza vaccination and reduction in hospitalizations for cardiac disease and stroke among the elderly. NEJM, 348(14): 2003).
- Classification of subject into those with combined pulmonary and cardiac disease, pulmonary disease, cardiac disease, diabetes and other endocrine disorders, immunosuppression, and being a healthy elderly individual.
- Outcome = composite of hospitalization for P&I or death from any cause.
- Data analysis – adjustment for age, sex, comorbidity, previous health care utilization, and previous hospitalization for P&I.
- 1997-98 – Poor match.
- Results
o Overall combined outcome – VE = 48% [42, 52] and 31% [26, 37].
o Effective in all sub-groups. Absolute benefits varied by subgroup.
§ VE
· All – 48% [42, 52], 31% [26, 37]
· Healthy – 46% [34, 56], 42% [28, 52]
· Diabetes – 50% [37, 60], 21% [6, 34]
· Lung – 48%, 27%
· Heart – 49%, 30%
· Heart and lung disease – 47%, 28%.
· VE roughly similar in 1996-97, but more lower and variable in 1997-98, a year of poor vaccine match.
o Vaccine prevented AR – healthy elderly – 3.8/1000 and 3.5/1000.
o Vaccine prevented AR – high-risk elderly – 18.0/1000 and 8.5/1000.
o NNT – 26-56 high risk, 264 healthy; 50-150 high risk, 290 healthy.
o Diabetes – outcome rates – vaccinated vs unvaccinated
§ 1996-97 – 185/9390 (2.0%) vs 548/12596 (4.4%).
§ 1997-98 –323/13966 (2.3%) vs 255/8025 (3.2%).
- Both healthy and high-risk elderly people derive substantial benefits from vaccination, and age-based strategies have been more effective than risk condition-based vaccination strategies.
- However, elderly people with underlying medical conditions do have significantly higher rates of hospitalization and death. Though all persons aged >= 65 years benefit and should be targeted for vaccination, efforts should be renewed to ensure vaccination among those with high-risk conditions.
Heymann, 2004 (ELDERLY ONLY)
- Estimate the effectiveness of influenza vaccination of community-dwelling, diabetic, elderly individuals.
- Retrospective cohort study.
- Data sources
o Databases of Maccabi Healthcare Services.
o Patients at least 65 years of age on October 1, 2000.
o Reference group was low-risk = absence of heart disease, lung disease, diabetes or endocrine disorders, renal disease, stroke or dementia, vasculitis, rheumatologic disease, and cancer.
- Vaccination status available from the database.
- Outcome = any hospitalization or death.
- Influenza season: October through February of 2000-2001.
- Negative control: June through September of 2000.
- 15556 (48.8% vaccinated) patients with diabetes, 69097 (42.0% vaccinated) patients in the reference group.
- Results
o Vaccine protective, NS for differences between diabetic and reference group
o Diabetic patients: 8.29% vs 9.44%, OR = 0.87 [0.77, 0.97], VE = 13%.
o Reference patients: 6.99% vs 9.08%, OR = 0.77 [0.75, 0.80], VE = 23%.
o No effect in either group during the summer control period. Diabetic patients had 10% more events during this period than reference patients.
o Vaccination significantly associated with reduction in hospitalization rates in all age and sex categories examined, except a few diabetes substrata. Likewise mortality rate of older women with diabetes not significantly improved.
- Weaknesses: PPV may have been more common in reference patients. Diabetic influenza signal may have been cancelled out by generally higher event rates in diabetic patients. Cause-of-hospitalization or death not available. Types of diabetes mixed. No adjustment for comorbidity.
- This study supports the use of annual influenza vaccination in elderly patients, including those with diabetes.
Looijmans-Van Den Akker, 2006 (RESULTS PROVIDED FOR ELDERLY AND NON-ELDERLY)
- Determine the effectiveness of influenza vaccination in reducing occurrence of hospitalization and death from any cause in adults with diabetes during an influenza epidemic.
- Nested case control study – prospective data collection.
o Cases were hospitalizations identified from a variety of ICD-codes, including MI, CHF, stroke, and diabetes dysregulation, all-cause deaths.
o Controls were randomly selected in a roughly 1:4 ratio.
- Data source
o Primary care-based Prevention of Influenza, Surveillance and Management (PRISMA) study cohort, including 90 general practices from 1999 to 2002.
o Prospective recording of data in routine medical practice.
o 1999-2000 influenza season included.
o 75235 patients under observation during the 1999-2000 A(H3N1) epidemic were considered. 9238 patients with diabetes were eligible.
- 192 cases and 1561 controls were included. Vaccinated control subjects were older, more likely to have chronic heart or lung disease, took more medication in the previous year.
- Results
o VE adjusted for age, sex, health care insurance, presence of heart or lung disease, or other high-risk disease, and number of meds and GP visits in the previous year.
§ Hospitalization VE = 54% [26, 71]
§ Death from any case VE = 58% [13. 80]
§ Hospitalization or death VE = 56% [36, 70]
o Adjusted VEs were higher for patients aged 18-64, which may explain much lower findings in Heymann compared with Colquhoun and the present study.
§ Hospitalization VE = 70% [39, 85]
§ Death VE = 24% [-706, 93]
§ Hospitalization or death VE = 72% [46, 85]
o Adjusted VEs, patients aged >= 65 – non-significant prevention of hospitalizations.
§ Hospitalization VE = 14% [-88, 60]
§ Death VE = 56% [4, 80]
§ Hospitalization or death VE = 39% [-5, 65]
- NNTs – Calculated by Darren – using Table 3 – crude incidence rates of total complications in the source population of 9238 patients with diabetes.
o Aged >= 65 years – 22.3 / 1000 – 29.9 / 1000 = AR = -7.6 / 1000. NNT = 132.
o Aged 18-64 – 14.0/1000 – 28.3/1000 = AR = -14.3/1000. NNT = 70.
- Adjustment in adults aged 18-64 appeared to increase VE estimates.
- IDDM mixed with NIDDM. Specificity of outcomes for influenza unclear. No non-epidemic control period.
- Most hospitalizations due to diabetes dysregulation.
- Patients with type 2 diabetes, like other high-risk individuals, benefit from annual influenza vaccination regardless of age.
Influenza vaccine effectiveness – Primary observational studies showing healthy user bias
Jackson LA, Nelson JC, Benson P, Neuzil KM, Reid RJ, Psaty BM, Heckbert SR, Larson EB, Weiss NS. Functional status is a confounder of the association of influenza vaccine and risk of all cause mortality in seniors. International Journal of Epidemiology, 35(2): 2006
- Objective: Explore the hypothesis that functional status and disease severity are confounders of the association of influenza vaccination and risk of death.
- Nested case-control study.
- Data source
o Group Health Cooperative databases. Cohort = subjects aged >= 65 on January 1, 1998.
o Community-dwelling.
o Cases = all subjects who died in Jan-Mar 1998.
o Controls – 2 or 3 controls per case, matched by age and sex, randomly selected – risk set sampling.
- The time period chosen was an influenza season of low vaccine matching – so the correct VE should be quite low.
- Comorbidity classification
o Group 1 – ICD-9-CM groupings
o Group 2 – Sub-groupings of group 1 variables for severity. E.g.: No cancer, non-serious cancer, serious or metastatic cancer; no diabetes, diabetes without complications, diabetes with complications.
o Group 3 – Functional status indicators – MRR – dementia, non-home residence, independent ambulation, assistance needed for bathing.
- 252 cases, 576 controls.
- Results
o Unadjusted OR = 0.59 [0.41, 0.83]
o Adjustment for Group 1 – OR = 0.45 [0.30, 0.68]
§ Comorbid disease associated with both mortality and vaccination.
o However, within comorbidity severity sub-strata - Group 2 variables
§ More severe disease associated with higher risk of death
§ E.g.: severe cancer – 11-fold increased risk of death – no difference in vaccination vs no cancer
§ Less severe strata were associated with increased vaccination, but only small or no differences in the risk of death.
§ E.g.: Non-serious cancer – 3 times higher risk of vaccination – no difference in risk of death.
o Adjustment for Group 2 – OR = 0.51 [0.33, 0.78]
1. Residual confounding among those with illness – those assigned a chronic renal disease code – proportion with creatinine value >= 2.4 mg/dl – 85% vs 37%, cases vs controls.
2. Residual confounding among those without illness – presence of functional impairments – 73% vs 21%.
§ Cases had more severe disease, less functional capacity, not captured by diagnostic codes.
o Adjustment for Group 3 – functional status – OR = 0.71 [0.47, 1.06]
o Restriction to create more homogeneous subgroups.
§ ORs much closer to the null
§ E.g.: Subjects who met the criteria for at least one of the group 1 variables – adjusted for functional status – OR = 0.83 [0.48, 1.41].
§ Subjects with a serum creatinine value recorded – OR = 0.93 [0.53, 1.64].
§ Subjects with an optometry department visit – OR = 0.86 [0.35, 2.12].
- Comorbidity classifications that do not distinguish severity jointly classify as diseased both persons with codes associated with an increased likelihood of vaccination and persons with codes associated with an increased risk of death.
- The combined group that appears to be at increased risk of death also appears more likely to be vaccinated – individually, though the opposite may be true, that those at most increased risk of death do not receive vaccine.
- An ecologic fallacy is committed.
- Residual confounding is also likely to be present in illness severity and functional status.
- Results suggest that functional status limitations identified by chart review are important confounders of the association of influenza vaccination and risk of death in seniors.
- Diagnostic codes incompletely address, if not worsen, confounding.
o Absence of diagnostic code not sensitive for absence of illness. Misclassification is differential.
o Heterogeneity in disease status among persons classified by broad groupings of diagnosis codes.
- Restriction produced estimates much closer to the null.
- Results suggest that further research is needed on methods to reduce selection bias in observational studies of influenza vaccine effectiveness.
Jackson LA, Jackson ML, Nelson JC, Neuzil KM, Weiss NS. Evidence of bias in estimates of influenza vaccine effectiveness in seniors. International Journal of Epidemiology, 35(2): 2006.
- Hypothesis – the magnitude of the underlying differences that predispose to death and hospitalization may diminish over time – depletion of the susceptible – groups become more similar.
- Objective – Evaluate the possible influence of bias. Evaluate ability of covariates defined by diagnosis codes and indicators of medical utilization to remove selection bias.
- Retrospective cohort study
- Data source
o Group Health Cooperative
o Subjects aged >= 65 years each September 1, entered every September 1.
o Cohort entry September 1995 to September 2002.
o Follow-up to August 31, 2003.
- Outcomes
o All cause mortality, P&I hospitalization
o Secondary – hospitalization – stroke, ischemic heart disease, CHF, and injury or trauma.
- Covariates – TVC for vaccination status, ICD diagnoses as secondary analyses.
- Time periods
o Influenza season – first and last weeks with >= 50 influenza positive isolates.
o Pre-season, post-influenza – to May 31, summer – to August 31.
- 75527 subjects – 338264 person-years of observation – 8 year study period.
- Vaccination coverage – 68 to 74%.
- Results
o RR – increased progressively in pre-influenza, influenza, and post-influenza time periods.
o Mortality, age-sex adjusted.
§ Before influenza –RR = 0.39 [0.33, 0.47]
§ Influenza – RR = 0.56 [0.52, 0.61]
§ Post influenza – RR = 0.74 [0.67, 0.80]