Thirty-one abstracts were reviewed and critiqued by each author independently. Twenty-four abstracts were excluded because the inclusion criteria were not met (Fig. 1), leaving seven studies for inclusion (Table 2).16–22
Jackson et al.19 conducted a large, retrospective cohort study in which patients were included if they were stable on warfarin therapy as defined by at least three INR levels within therapeutic range (≥2 to ≤4) for at least three consecutive months and received any one of four vaccinations between the years 1992 and 2003. Of the 4923 patients who received a trivalent, inactivated influenza vaccination, mean INR values 28 days following influenza vaccination did not differ significantly from mean values outside of the 28-day post-vaccination period even after adjusting for potential confounders (2·53 vs. 2·54 respectively; mean INR difference 0·01; 95% CI −0·01 to 0·03). No safety parameters were evaluated as part of this study.
Prospective studies – uncontrolled
MacCallum et al.16 conducted an audit of 78 patients receiving long-term warfarin therapy who reported receiving an influenza vaccine within 10 days prior to a clinic visit. To reduce the influence of intrapatient variability on INR levels, post-vaccination INR levels were standardized against patient-specific INR data within 1 year of the study vaccination date. Comparable patterns of variability were noticed between outlying INR values post-vaccination and similar outliers independent of influenza vaccination. The study also plotted standardized INR values for each patient between days 1 and 10 post-vaccination. The influenza vaccine had no effect on INR because any variation was not dependent on time elapsed from vaccination. None of the patients reported experiencing any adverse bleeding or thrombotic events.
In a study by Arnold et al.,22 changes in INR following influenza vaccination were assessed in nine patients who were on a stable dose of warfarin for at least 2 months prior to study enrollment. Following influenza vaccination, patient INRs were assessed at seven different time points over the course of 30 days. Median INRs following vaccination were not significantly different from median baseline INR values. One patient experienced epistaxis during the final week of the study, and it was noted that the patient’s INR during this time was lower than at baseline.
Prospective studies – case controlled
Ninety patients considered stable on anticoagulant therapy for 3 months prior to study inclusion were assessed by Paliani et al.17 Ninety-eight percent of these patients received warfarin, while 2% received acenocourmarol. Patient INR values were assessed three times prior to influenza vaccination against a matched comparator group, with the values taken 5–7 days before immunization and 7–10 days post-vaccination. At study completion, the study group had an increase in mean INR of 0·56 from baseline. The difference between mean INR in the two groups was considered significant (3·35 ± 1·04 study vs. 2·59 ± 0·90 comparator, P < 0·00005). In a subgroup analysis, 49 of 90 patients experienced a ≥0·5 change in mean INR (2·64 ± 0·95 before vs. 3·85 ± 0·98 after, P < 0·00001), and two reports of epistaxis and muscular hematoma were reported within this subgroup. The remaining patients did not experience any INR changes, and no other adverse bleeding events were reported.
Poli et al.18 evaluated patients considered stable on warfarin therapy for at least 6 months prior to study enrollment. All 73 patients in the study group who completed the trial received a single SQ influenza vaccination. Patient INRs in both the study and comparator group were evaluated during the 3 months before and after vaccination. Mean INRs were not reported, but no differences were found prior to or following vaccination within either group. In a subgroup analysis of patients older than 70 years of age, time spent below the therapeutic INR range appeared to be significantly longer in the study group following immunization (10% before immunization vs. 27% after, P = 0·001). A similar observation was not noted in the comparator group during the same time period. None of the patients experienced an adverse bleeding event during the course of the study.
Prospective studies – placebo controlled
Farrow et al.21 completed a single-blind study evaluating the effects of influenza and pneumococcal vaccines in 69 patients with a stable INR for at least 3 months prior to study entry while on warfarin. Twenty-five patients each were randomized to receive either an influenza vaccine or a saline control injection; the remaining patients received a pneumococcal vaccine. INR levels were taken immediately prior to vaccination and 2, 7 and 21 days post-vaccination. There were no statistically significant differences between groups in mean INR values at any time point post-vaccination. Two patients, one in the influenza vaccination group and one in the control group, required a dose reduction after INR levels rose to 4·5 on day 7 and 5·0 on day 21, respectively. Four patients in the influenza vaccine group vs. one patient in the control group required small increases in warfarin dosage to maintain an INR >2·0. No adverse events were reported in the patients receiving influenza vaccination.
A double-blind crossover study by Iorio et al.20 assessed 100 patients receiving warfarin for >6 months and had a minimum of three consecutive therapeutic INRs that were documented at least 3 weeks apart. The study spanned 70 days and consisted of two 28-day study periods separated by a 14-day washout period. Fifty study subjects received an influenza vaccine during the first study period and a placebo injection during the second period, while the remaining study subjects received the injections in the reverse sequence. INR levels were assessed weekly during each study period. Following study completion, differences between mean INR, mean weekly doses of warfarin, and percentage of time spent outside the therapeutic range after treatment or placebo were not statistically significant. Likewise, data analysis using a linear mixed-effects model confirmed that vaccination did not significantly affect INR (regression coefficient −0·095, 95% CI −0·253 to 0·064; P = 0·24) or weekly warfarin dose (regression coefficient 0·228, 95% CI −0·902 to 1·357; P = 0·69). For adverse bleeding events, six events were reported during vaccination periods as compared to five events during placebo periods. More events were reported overall in the group that received the vaccine during the first study period (nine events in the vaccine to placebo group vs. two events in the placebo to vaccine group). The most common event reported was epistaxis; all events noted were considered minor and occurred within subtherapeutic to therapeutic INR ranges (1·5–3·3) with the exception of one bruising event that occurred at an INR of 6·9.
With the exception of the results of subgroup analyses and the study by Paliani et al.,17 significant changes in INR were not observed post-vaccination.16–22 In addition to INR, adverse bleeding events were also assessed in this review. No adverse bleeding episodes occurred in three of the seven studies.16,18,21 The remaining reported events were considered minor in nature and not always associated with increases in INR.17,22 Interestingly, in the study conducted by Iorio et al.,20 most of the bleeding events occurred in patients who received the vaccine prior to placebo. In evaluating the nature of the bleeding events reported overall, this trend may be attributed to variability in anticoagulation response between and within patients rather than to the vaccine itself.
The studies reviewed included only patients stable on their current warfarin therapy, and this may not reflect clinical practice. Both interpatient and intrapatient variability in warfarin response should be considered when interpreting these studies. Three of the studies reviewed attempted to control for this potential confounder,16,19,20 which may explain Kramer’s index case and the inconsistent results seen in the studies reviewed. It remains questionable as to whether the reported studies had adequate statistical power. Iorio et al.20 noted that a post hoc analysis demonstrated that the study was sufficiently powered. Such post hoc analyses may lead to flawed conclusions.
Notably, two of the studies reviewed documented a significant difference in INR changes post-vaccination despite relatively small sample sizes.17,18 Paliani et al.17 reported that mean INR values were significantly potentiated post-vaccination in the study group. However, this comparison was made between the study group and the comparator group. It is not known whether the increase in INR observed in the study group was statistically significant when compared with the mean baseline INR of the same set of patients. Making this comparison would have been beneficial to account for potential intrapatient variability in anticoagulation response over time as addressed by the studies conducted by MacCallum et al.16 and Jackson et al.19
Through post hoc analysis, Poli et al.18 detected a significant difference in time spent below the therapeutic range in a subgroup of elderly patients >70 years of age. As mean INR values were not reported, it may be possible that these patients had baseline INR values that were already on the threshold between therapeutic and subtherapeutic classifications. On another note, while post hoc subgroup analyses may provide further insight into the study data, such analyses are not considered to be ideal because there is an increased possibility of misinterpreting the data and detecting a significant difference when one may not exist.23 Thus, conclusions made from observations noted in the subgroups mentioned by both Paliani et al.17 and Poli et al.18 should be made with caution.
Lastly, heterogeneity in methodology existed between the different studies in regard to the route of influenza vaccine administration and time of follow-up. Both of these factors could be considered as limitations in assessing the effect of influenza vaccine on INR levels. However, after further evaluation, neither of these factors are expected to have significantly impacted the results reported. For route of administration, subcutaneous injections were utilized in two studies in which significant differences were documented either overall or within a subgroup.17,18 While IM injection is the recommended route for the influenza vaccine, there is some belief that patients on chronic anticoagulation should receive the immunization subcutaneously to decrease the risk of muscular hematoma.24 However, studies have shown that patients on anticoagulation therapy may receive the vaccine via the IM route without an increased risk of hematoma because the immunological response following either an IM or SQ injection is similar.11,25 Therefore, the route of administration is not expected to be a confounder in assessing the effect of influenza vaccine on INR. Likewise, time after follow-up did not appear to affect whether a difference was observed in INR levels following immunization. In the index case study published by Kramer et al.,5 an adverse bleeding effect was seen within 10 days of influenza vaccination. As described previously, Paliani et al.17 noted a significant difference in INRs post-vaccination compared with the control group within the same time period. However, in the remaining six studies, INR levels were not significantly different up to 3 months after immunization.16,18–22 This time frame is beyond that in which INR changes are expected.