Reduced risk of poststroke pneumonia in thrombolyzed stroke patients with continued statin treatment
- Conflicts of interest: ME has received a research grant from AstraZeneca for studies with rosuvastatin.
- Funding: ME receives grant support from the Volkswagen Stiftung (Lichtenberg program to ME), Deutsche Forschungsgemeinschaft and German Ministry of Education and Research (Center for Stroke Research Berlin, CSB).
Pneumonia is a frequent complication after stroke with strong impact on clinical outcome. Statins have pleiotropic immunmodulatory properties and were recently shown to exert beneficial effects on the development and clinical course of pneumonia.
We aimed to investigate whether statin use is associated with a reduced risk of poststroke pneumonia in acute ischemic stroke patients treated with tissue plasminogen activator within 4·5hours.
Data was extracted from a local register including all consecutive stroke patients who received thrombolysis at our institution. Prior statin use was identified retrospectively from clinical records and had to be continued after hospital admission. Poststroke pneumonia was diagnosed according to standardized criteria of US Centers for Disease Control and Prevention. Mortality and functional outcome at three-months were further assessed.
Overall, 481 ischemic stroke patients were analyzed. Continued statin use was documented in 17% of the patients. Frequency of pneumonia was 11%. Patients with statin use were less likely to develop poststroke pneumonia (5% vs. 13%, P = 0·04). After multivariable adjustment for known risk factors for poststroke pneumonia (age, stroke severity, dysphagia, male sex and diabetes), statin treatment was negatively associated with pneumonia (OR 0·31; 95% CI 0·10–0·94). Occurrence of pneumonia independently predicted three-month mortality and functional outcome.
Use of statins in acute ischemic stroke patients who receive thrombolysis might reduce the risk of poststroke pneumonia. Further studies are warranted to validate this finding.
Pneumonia is a serious complication after stroke and can be observed in 7% to 22% of the patients treated on a stroke unit [1, 2]. About 10% of all deaths during the stay in an acute care hospital after stroke can be attributed to this type of infection . Stroke patients who develop pneumonia during the hospital stay are more likely to remain functionally dependent [2, 4, 5].
Pneumonia after stroke is generally caused by aspiration due to stroke-related deficits such as dysphagia and immobility. Of note, stroke patients are particularly susceptible to pneumonia caused by stroke-induced impairment of the immune system [6, 7].
Statins significantly lower serum cholesterol levels via inhibition of the 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA)-reductase. Experimental and animal studies demonstrated additional anti-inflammatory and immunmodulatory properties [8, 9]. These pleiotropic effects were regarded to contribute to a lower probability of pneumonia and sepsis in patients treated with statins recently reported in large observational studies, especially in patients with cardiovascular disease and diabetes [10-12]. Besides prevention of pneumonia, a recent study also showed an association of statins with improved survival after pneumonia . However, data on a possible influence of statins on prevention of pneumonia in acute stroke patients is lacking.
Therefore, the aim of our study was to analyze the impact of statin treatment on the frequency of poststroke pneumonia and on outcome after acute ischemic stroke, including in-hospital mortality as well as mortality and favorable outcome (modified Rankin Scale ≤2) after three-months.
Study population and outcomes
All patients with acute ischemic stroke receiving tissue plasminogen activator (tPA) within 4·5 hours after symptom onset at our institution were registered in an ongoing, prospective thrombolysis register. The presented analysis included data collected from January 2005 to December 2010.
For each stroke patient demographic data (age, gender, living conditions), presence of dysphagia, stroke risk factors and concomitant diseases (hypertension, atrial fibrillation, diabetes mellitus, hypercholesterolemia, smoking, previous stroke, coronary artery disease) were recorded. Moreover, frequency of prestroke disability according to the modified Rankin scale (mRS) was assessed. Dichotomization was done at a score of 2 indicating prestroke independence. Fasting low-density lipoprotein (LDL) cholesterol levels were measured one-day after stroke unit admission and were categorized as elevated if above 130 mg/dl (3·36 mmol/l) . Stroke severity on admission was analyzed by certified raters using the National Institute of Health Stroke Scale (NIHSS). Stroke etiology was specified according to modified criteria of the Trial of Org 10172 in Acute Stroke Treatment (TOAST) . In contrast to the original classification, etiology of large-artery arteriosclerosis was also assumed in patients with a high degree of plaques present in ultrasound and appropriate clinical and imaging presentation. Neurologic improvement after tPA treatment was defined as an improvement on the NIHSS ≥ 8 points or a score of 0 or 1, respectively .
Prior statin use was identified retrospectively from medical records. Because approximately 75% of the events of poststroke pneumonia occur within three-days after hospital admission , and because abrupt discontinuation of chronic statin therapy leads to a overshoot reversal of pleiotropic effects , statin treatment was affirmed if administered before stroke onset and continued during at least the first three-days after hospital admission (statin-group). Statin type and dosage was recorded. Statin dosage was subcategorized into three groups according to their lipid lowering effect. For simvastatin, pravastatin and fluvastatin, doses <40 mg were regarded low, doses ≥40 mg and <80 mg were considered medium and ≥80 mg high. For atorvastatin, doses <40 mg were defined as medium dose and ≥40 mg as high dose.
The primary outcome of our study was the frequency of pneumonia during the hospital stay. Pneumonia was diagnosed according to standardized criteria of the US Centers for Disease Control and Prevention. At least one finding of the first and at least two of the second category had to be present in addition to pulmonary infiltrates in chest X-rays: (I) leukocytosis (>12·000/mm3) or leukopenia (<4·000/mm3), fever (>38·0°C), new unexplainable mental disturbance in patients aged ≥70; (II) new or progressive cough, dyspnea, tachypnea, abnormal respiratory examination, new or altered purulent sputum, impaired gas exchange .
Further outcome measures were in-hospital mortality, degree of disability as determined by the mRS as well as mortality after three-months. The mRS score at three-months was dichotomized into favorable (mRS ≤ 2) and unfavorable and was obtained either via standardized telephone interview of the patient or a close relative by a certified mRS rater or via written questionnaire .
In univariate analysis, statistical significance was determined with Pearson's χ2-test for categorical variables and with the Wilcoxon-Mann-Whitney-U-test for continuous variables.
A multivariable logistic regression analysis was performed to calculate odds ratios (OR) and corresponding 95% confidence intervals (CI) for the occurrence of pneumonia after stroke. Variables for adjustment were predefined and selected from literature review and included dysphagia, older age, higher stroke severity determined by NIHSS on admission, male sex, diabetes as risk factors for pneumonia [5, 20, 21]. Age and NIHSS were entered into the model as continuous variables. To test the sensitivity of the model, alternative multivariable regression models were calculated, replacing prior statin treatment ceteris paribus for all variables that were found to be significantly different in the statin and nonstatin group, as well as for all factors significantly associated with poststroke pneumonia in the univariate analysis. In this analysis, the variable length of hospital stay was dichotomized at seven-days, which was the median length of stay in our cohort. Finally, we conducted a forward stepwise regression analysis containing all of these variables.
To determine the impact of continued statin treatment on favorable outcome and mortality after three-months, logistic regression analysis was performed including age, gender, stroke severity (NIHSS on admission), occurrence of pneumonia, history of stroke and atrial fibrillation . Due to the low event rate, multivariable adjustment regarding in-hospital mortality was done including age, stroke severity and occurrence of pneumonia.
All tests were two-tailed and statistical significance was determined at an alpha level of 0·05. Statistical analyses were performed using spss, version 18·0 for Windows.
Overall, 487 patients were registered in our institutional thrombolysis register between 2005 and 2010. Information about prior statin use or diagnosis of pneumonia was not available in six patients. The remaining 481 patients were included in our analysis.
Statin pretreatment continued for at least three-days after the index stroke was found in 83 patients (17%, statin group). Patients with and without continued statin treatment did not differ regarding age, sex, stroke severity on admission, presence of dysphagia, stroke etiology, frequency of admission from nursing home and prestroke independence, smoking habit, length of hospital stay, achievement of major neurologic improvement after tPA treatment and frequency of symptomatic intracranial hemorrhage. Hypertension, diabetes, hypercholesterolemia, coronary artery disease, history of stroke and elevated LDL-cholesterol levels were more common in the statin group (Table 1). The majority of the patients received simvastatin (66%). Other statins used in our cohort were atorvastatin (12%), pravastatin and fluvastatin. Approximately half of the patients with statin use received medium doses (n = 42), whereas 40% had low doses (n = 33) and 10% had high doses (n = 8).
Table 1. Baseline characteristics of patients with and without continued statin treatment
|Age, years, median (IQR)||74 (66–83)||75 (70–82)||0·29|
|Male sex, % (n)||49·7 (198)||51·8 (43)||0·73|
|NIHSS on admission, median (IQR)||11 (6–18)||10 (6–16)||0·13|
|Dysphagia, % (n)||19·1 (76)||19·3 (16)||0·97|
|Admission from a nursing home, % (n)||11·1 (44)||8·4 (7)||0·48|
|Prestroke independence (mRS ≤ 2), % (n)||67·5 (268)||60·2 (50)||0·20|
|Length of hospital stay, median (IQR)||7 (6–11)||7 (5–10)||0·38|
|Stroke risk factors|| || || |
|Hypertension, % (n)||80·4 (320)||95·2 (79)||<0·01|
|Atrial fibrillation, % (n)||42·6 (169)||38·6 (32)||0·51|
|Diabetes mellitus, % (n)||21·9 (87)||36·1 (30)||<0·01|
|Hypercholesterolemia, % (n)||38·9 (148)||73·2 (60)||<0·01|
|Smoking, % (n)||23·7 (94)||22·9 (19)||0·88|
|Coronary artery disease, % (n)||15·8 (63)||39·8 (33)||<0·01|
|Previous stroke, % (n)||20·9 (83)||34·9 (29)||<0·01|
|Elevated LDL cholesterol, % (n)||28·7 (96)||10·6 (7)||<0·01|
|Stroke Etiology|| || ||0·61|
|Cardioembolic, % (n)||45·0 (179)||36·1 (30)|| |
|Large-artery arteriosclerosis, % (n)||40·7 (162)||49·4 (41)|| |
|Small vessel occlusion, % (n)||2·8 (11)||3·6 (3)|| |
|Other defined cause, % (n)||3·0 (12)||2·4 (2)|| |
|Undefined cause, % (n)||8·5 (34)||8·4 (7)|| |
|Neurologic Improvementa||14·8 (40)||20·4 (43)||0·11|
|Symptomatic intracranial hemorrhageb||4·8 (19)||6·0 (5)||0·63|
Frequency of pneumonia was 11% (n = 54). Univariate comparison revealed patients with pneumonia to be older (median age 82 vs. 74, P < 0·01) and to have more severe strokes (median initial NIHSS 19 vs. 10, P < 0·01). Presence of dysphagia (P < 0·01), atrial fibrillation (P = 0·03), hypertension (P = 0·02) and diabetes (P = 0·05) was more common in patients with pneumonia, whereas prestroke independence was less common (P = 0·02). Median length of hospital stay was significantly longer in patients who developed pneumonia (10 days vs. 7 days, P < 0·01). Major neurologic improvement after tPA treatment occurred significantly less often in patients with pneumonia (P < 0·01). Symptomatic intracranial hemorrhage was observed significantly more often in patients with pneumonia (P = 0·03).
Patients with prior statin use were less likely to develop poststroke pneumonia (5% vs. 13%, P = 0·04). Statin treatment was shown to significantly predict absence of pneumonia after adjustment for predefined factors age, stroke severity (NIHSS on admission), dysphagia, sex and diabetes in a multivariable regression analysis (OR 0·31; CI 95% 0·10–0·94, Table 2). Variables that were found to be significantly different between the statin and the nonstatin group or associated with pneumonia in the univariate comparisons were alternatively forced into the multivariable model instead of statin use, but no significant effect of hypertension, hypercholesterolemia, coronary artery disease, previous stroke, elevated LDL-cholesterol levels and atrial fibrillation were detected, respectively. Longer length of hospital stay (LOS, P < 0·01), neurologic improvement after tPA therapy (P = 0·05) and symptomatic intracranial hemorrhage (P = 0·04) yielded a significant association with pneumonia.
Table 2. Multivariable analysis regarding poststroke pneumonia
|NIHSS, per point||1·12||1·06–1·18||<0·01|
|Age, per year||1·05||1·02–1·08||<0·01|
Finally, we also forced all variables with a possible association with pneumonia into a forward stepwise multivariable regression analysis and found a significant association of statin use (OR 0·32; 95% CI 0·10–0·98), dysphagia (P < 0·01), higher stroke severity (P < 0·01), older age (P < 0·01), longer LOS (P < 0·01) and symptomatic intracranial hemorrhage (P = 0·02, Table 3).
Table 3. Forward stepwise regression analysis regarding poststroke pneumonia
|NIHSS, per point||1·12||1·06–1·18||<0·001|
|Age, per year||1·05||1·02–1·08||0·001|
|Length of hospital stay ≥seven days||3·45||1·64–7·29||0·001|
|Symptomatic intracranial hemorrhagea||3·98||1·28–12·38||0·017|
Frequency of in-hospital mortality was 8%. There was no statistically significant difference between the statin and the nonstatin group (6% vs. 8%, P = 0·53). Multivariable regression analysis revealed higher stroke severity to predictin-hospital mortality significantly (OR per point on NIHSS 1·16; 95% CI 1·10–1·23).
Follow-up rate was 98%. Overall, 97 patients (21%) died during the follow-up period of three months. As shown in Table 4, three-month mortality was lower in the statin group (13% vs. 22%), albeit this did not reach statistical significance in our sample (P = 0·08). After multivariable adjustment for age, gender, stroke severity, occurrence of pneumonia, history of stroke and atrial fibrillation no significant impact of statin treatment was found (OR 0·64; 95% CI 0·30–1·37). Independent predictors of mortality after three-months were older age (OR per year 1·09; 95% CI 1·05–1·12), higher stroke severity (OR per point on NIHSS 1·14; 95% CI 1·10–1·19) and occurrence of pneumonia during the hospital stay (OR 3·20; 95% CI 1·59–6·45). Presence of favorable outcome after three-months did not differ significantly between patients with and without continued statin treatment (48% vs. 44%, P = 0·56). Older age (OR per year 0·95; 95% CI 0·93–0·97), higher stroke severity (OR per point on NIHSS 0·85; 95% CI 0·82–0·88), occurrence of pneumonia (OR 0·04; 95% CI 0·01–0·31) and history of stroke (OR 0·42; 95% CI 0·24–0·75) were negatively associated with favorable outcome in a multivariable regression analysis.
Table 4. Outcome measures in patients with and without continued statin treatment
|Pneumonia, % (n)||12·6 (50)||4·8 (4)||0·35 (0·24–1·00)||0·04||0·31 (0·10–0·94)||0·04|
|In-hospital mortality, % (n)||8·0 (32)||6·0 (5)||0·73 (0·28–1·94)||0·53||0·88 (0·32–2·46)||0·81|
|Favorable outcomea, % (n)||44·1 (171)||47·6 (39)||1·15 (0·71–1·85)||0·56||1·22 (0·68–2·20)||0·50|
|Three-month mortality, % (n)||22·2 (86)||13·4 (11)||0·54 (0·28–1·07)||0·08||0·64 (0·30–1·37)||0·25|
Pneumonia is one of the most serious medical complications after acute stroke as it is strongly associated with worse clinical outcome and mortality [2-4]. The results of the present analysis suggest a preventive effect of continued statin treatment against poststroke pneumonia. In addition, older age, higher stroke severity and presence of dysphagia were confirmed to be predictors of poststroke pneumonia, as shown in previous studies [2, 20, 21].
Our findings might be supported by in vitro and animal studies in which statins were shown to attenuate immune responses and inflammation [9, 22, 23]. A growing number of studies support the notion of a possible impact of statins on the development of pneumonia initially concluded from the results of large case-control studies, in particular in patients with atherothrombotic diseases [10-12].
Despite these promising reports, data on effects of statins on the occurrence of pneumonia in acute stroke patients are scarce and inconsistent. Recently, Arboix et al. reported lower rates of respiratory events and infectious complications in patients with first-ever stroke and prior statin use . Respiratory events included pulmonary embolism and atelectasis in addition to pneumonia. Unfortunately, no definition of pneumonia was given and data were not adjusted for possible confounders. A Spanish group found no association of previous statin treatment and infectious complications in stroke patients less severely affected than in our cohort . In the MISTICS trial even a greater proportion of infections was found in patients with initiation of statin therapy shortly after stroke onset . However, in contrast to our analysis, this was an interventional clinical trial excluding patients with prior statin use.
An alternative explanation of our observations might be an interaction of tPA and statins. Statins were found to enhance thrombolytic efficacy of tPA in animal studies and successful thrombolysis was shown to decrease systemic inflammatory response [27, 28]. Hence, it might be possible that statins contribute to prevention of pneumonia via improvement of tPA response. However, we did not find frequency of neurologic improvement after tPA treatment significantly more often in statin users and neurologic improvement was not independently associated with pneumonia in the forward stepwise analysis. This might be explained by the fact that improvement on the NIHSS can occur in several items and does not necessarily mirror improvement of dysphagia or immunosuppression in all cases.
In our analyses, longer LOS was found to be independently associated with pneumonia. A longer LOS makes development of pneumonia more likely, but on the other hand pneumonia might lead to longer LOS. The majority of patients develop pneumonia early after stroke, when risk of aspiration is greatest and dysphagia is most pronounced .
It has been argued that statin users may reflect a ‘healthier’ population with a closer medical attention and adherence to treatments. Majumdar et al. analyzed the outcome of patients hospitalized for pneumonia and found no impact of statin pretreatment after adjustment for confounding factors such as smoking, up-to date vaccination and admission from nursing homes . Therefore, the authors concluded that beneficial effects of statins on prevention of infections in observational studies were contorted by some kind of ‘healthy user effect’. However, in our study population, current smoking and frequency of admission from nursing homes did not differ significantly between patients with and without statin intake arguing against a ‘healthy user’ bias. Pneumonia risk factors age, stroke severity and dysphagia were found in equal proportions in statin users and nonstatin users, suggesting a similar baseline probability to develop pneumonia. Nevertheless, there was a higher prevalence of concomitant diseases in statin users which might mirror a better documented medical history and better controlled risk factors and we cannot exclude that some unnoticed or not sufficiently controlled confounders contributed to the observed associations.
Reflecting the huge impact of pneumonia on mortality in stroke patients described in previous studies , one might assume to detect a lower mortality in patients with statin use. In fact, statin treatment was recently shown to decrease mortality in patients hospitalized for pneumonia in a large British cohort study . We also found a lower three-month mortality in statin-treated patients, although this held not true after multivariable adjustment. In addition, statin use was not associated with decreased in-hospital mortality or with more favorable outcome three-months after stroke. This is in accordance with a large multicenter study recently published by Engelter et al., which found no effect of prior statin use on three-month functional outcome in thrombolyzed stroke patients . The effect size of statin use on stroke outcome, compared to that of tPA, as well as the number of patients analyzed in our study might have been too small to delineate a beneficial effect of the observed reduced incidence of pneumonia on long-term outcome.
Strengths of our study include the standardized diagnostic criteria of pneumonia and equal standard of care (i.e. stroke unit treatment) for all patients. The definition regarding continuation of statin therapy after hospital admission is relevant, because the beneficial effects of statins may diminish after withdrawal [17, 31]. Moreover, our results were robust throughout different methods of multivariable adjustment and sensitivity analyses.
However, there are certain limitations that need to be reflected. First, the monocentric and retrospective design mitigates the validity of the results. Second, the results are confined to patients who received thrombolysis. On the other hand, the restriction to this population implicates that all included patients were admitted within 4·5 hours after symptom onset. Therefore, we did not need to adjust for prehospital delay, which increases the chances to develop pneumonia because of unobserved aspiration or immobility .
Third, we are not able to provide information about preexisting lung diseases. This can be relevant as patients with preexisting structural lung disease are prone to infections. Fourth, we cannot present data about stroke volume, which was shown to be a predictor of pneumonia after stroke [4, 32]. Fifth, subgroups of patients receiving other statin type than simvastatin were too small to determine whether a class effect or type effect accounts for our observations. However, beneficial effects in secondary prevention of cardiovascular events as well as pleitropic effects of statins are regarded to be class effects rather than type specific [17, 33].
In conclusion, the results of our study provide evidence that statins before and continued after stroke might reduce the incidence of poststroke pneumonia within the acute care of ischemic stroke patients receiving tPA. This reduction might contribute to the previously described beneficial effects of statins in stroke patients . Further studies are needed to validate the possible impact of statins on prevention of infections in general stroke populations including patients without thrombolytic treatment.
We would like to thank the statistician Dr Uwe Malzahn for advice. Furthermore we thank Janina R. Behrens, Lea Gerischer and all other persons involved in data acquisition.