Professional Documents
Culture Documents
404
Meta-analysis of fever and stroke mortality
with mortality. All resulting titles were scanned ity. We used fixed effects model as there was no
and relevant articles were analyzed in detail. We significant heterogeneity across the studies
also sought for potential studies for inclusion by (P = 0.28, I2 = 21.2%). The combined estimate
searching bibliographies of identified studies. We of the five studies yielded an odds ratio (OR) of
attempted to contact authors for clarification of 2.20 (95% CI 1.59–3.03, P < 0.00001) (Fig. 1).
unclear or insufficient information. Two authors We also conducted meta-analysis of all six
independently assessed the studies for inclusion in studies using a fixed effect model, to assess the
the review and any disagreement was resolved by impact of including the sixth study. As expected,
discussion. Studies were eligible in case they met the heterogeneity across the studies became statis-
the following inclusion criteria: (i) they had tically significant (P = 0.0002, I2 = 79.7%). Fur-
patients with acute ischemic stroke with mortality ther, meta-analysis with random effects model
within 30 days or in-hospital mortality as outcome, yielded an OR of 1.86 (95% CI 1.10–3.15,
(ii) had performed multivariable analysis (iii) had P = 0.02) (Fig. 2).
adjusted for age and stroke severity (as our aim In both cases, a statistically significant increase
was to determine the association between fever and (by about two times) in odds of death in patients
outcome, independent of age and stroke severity). with fever was observed.
We separately extracted adjusted odds ratios and
standard errors from logistic regression analysis of
Discussion
published studies and used generic inverse variance
method to combine the results. Chi-square and I2 Our meta-analysis shows that fever occurring
were used to examine heterogeneity. Statistical within the first 24 h of onset of ischemic stroke
analysis was done using Cochrane Collaboration is associated with almost twice the risk of short
software, Revman 4.2.8 (24). term mortality. This association seems to be
independent of age, stroke severity or co-morbid-
ity (hypertension, diabetes, hyperlipidemia)
Results
because the odds ratios were obtained from the
We found six cohort studies (7, 8, 12, 15, 18, 22) results of logistic regression analysis of the five
eligible for inclusion in the meta-analysis. The studies that were adjusted for confounding fac-
remaining nine studies (6, 9, 11, 13, 14, 16, 19–21) tors. The results are consistent across the included
were excluded as they did not provide odds ratios studies as the test of heterogeneity has a statisti-
from logistic regression or did not provide mortal- cally non-significant chi square test as well as the
ity data. I2 value of 21.2% (Fig. 1). This meta-analysis
Six cohort studies selected as per inclusion builds on previous one by adding new studies and
criteria had reported odds ratios and standard combining the adjusted odds ratios. Greer et al. in
errors for mortality. These studies included a total a recent meta-analysis of 14,431 patients with
of 2986 patients (Table 1). Table 2 details excluded stroke and other brain injuries found that fever is
studies. Table 3 details the variables included in consistently associated with worse outcomes
multivariate analysis in various studies. across multiple outcome measures (25). They
To estimate the size of effect of fever on short- found that patients with hemorrhagic or ischemic
term mortality, we combined five studies and stroke or traumatic brain injury who had fever or
excluded the study (18) reporting one year mortal- raised body temperature had a prolongation of
Total No of
number of patients
Study patients with fever Inclusion criteria Mortality Cause of fever Comments
405
Prasad & Krishnan
Figure 1. Forest plot of adjusted odds ratios of all studies except study by McWalter et al. (fixed effect model).
406
Meta-analysis of fever and stroke mortality
Review: Fever is associated with doubling of odds of short term mortality in ischemic stroke: an updated meta-analysis
Comparison: 01 With fever versus without fever
Outcome: 01 mortality
Study Odds ratio (random) Weight Odds ratio (random)
or sub-category log [Odds Ratio] (SE) 95% Cl % 95% Cl
Figure 2. Forest plot of adjusted odds ratio of all studies (random effects model).
fact that ten studies (6–15) found significantly post-ischemic brain. It has been found that hypo-
higher mortality rates among patients with fever as thermia may interrupt this cycle, and conversely,
compared to patients without fever, whereas seven hyperthermia would enhance or reactivate this
studies (16–22) did not find fever to be associated process (29). The explanation for the possible
with higher mortality in stroke. This may be due to neuroprotective effect of hypothermia in acute
one or more of the following reasons: stroke is not completely understood but most
studies suggested that it might be due to slowing
1. Selection bias: If patient population among the down of the apoptosis and necrosis in the penum-
studies varies in time since onset, it introduces bra. In ischemic stoke, the time window for
different degrees of survivor bias in different reperfusion to be established is likely to be widened
studies. For example, if a study has a time by such a measure, but no such benefit is expected
window of 1 week since onset, those presenting for hemorrhagic stroke (8). Georgilis et al. (14)
towards the end of a week represent patients have studied the causes of fever in acute stroke.
ÔsurvivingÕ the risk of mortality in the first few Fever occurred in 37.6% of 330 patients; more
days. often in hemorrhagic stroke with mass effect,
2. The definition of fever varies from study to transtentorial herniation and intraventicular
study. Fever was defined as temperature >37.4 blood as also in large ischemic strokes. Patients
degrees centigrade in two studies (7, 15), 37.5C with fever had lower scores on admission on the
in three and (12, 18, 22), 37.9C in one (8). Glasgow Coma Scale. In their study 22.7% of
3. The most common reason is the inadequate patients with fever had an infectious cause; the
sample size to detect the association adjusted development of fever was associated with prior use
for other prognostic factors like age, GCS score, of an invasive technique especially urinary cathe-
etc. This is reflected in statistically non-signifi- terization (14). This means that majority of
cant but wide CI in some of the studies. patients with stroke develop fever due to pro-
4. Some studies may not have conducted an inflammatory cytokines and not due to infectious
adjusted (regression) analysis. A crude analysis cause.
may overestimate or underestimate the associ- To summarize, our updated meta-analysis shows
ation that can be found in a regression analysis. that fever is associated with doubling of odds of
short-term mortality in acute ischemic stroke and
the association is independent of age and stroke
Pathophysiologic rationale
severity
What is the pathophysiologic basis for this asso-
ciation? Neuronal damage produced by ischemia
References
generates neurotoxic substances, such as superox-
ide, nitric oxide, or cytokines, which could induce a 1. Carlberg B, Asplund K, Hagg E. Course of blood pressure
rise in temperature after the ischemic episode. in different subsets of patients after acute stroke. Cere-
brovasc Dis 1991;1:281–7.
Interleukin-1b induces fever and is expressed in the
407
Prasad & Krishnan
2. Busto R, Dietrich WD, Globus MY, Valdes I, Scheinberg 16. Castillo J, Davalos A, Marrugat J, Noya M. Timing for
P, Ginsberg MD. Small differences in intraischaemic brain fever-related brain damage in acute ischemic stroke. Stroke
temperature critically determine the extent of ischaemic 1998;29:2455–60.
injury. J Cereb Blood Flow Metab 1987;7:729–38. 17. Sharma JC, Ross IN. Antipyretic therapy in acute stroke.
3. Brandstater ME. Prognostication in stroke rehabilitation. Lancet 1998;352:740–1.
In: Chino N, Melvin JL, eds. Functional evaluation of 18. McWalter R, McMahon A, Fraser H, Bruce V, Hendrick
stroke patients. Tokyo: Springer-Verlag, 1996;93–102. S. Does body temperature on admission predict long-term
4. Kim Y, Busto R, Dietrich WD, Kraydieh S, Ginsberg MD. outcome after an acute stroke? Cerebrovasc Dis 1998;
Delayed postischemic hyperthermia in awake rats worsens 8(Suppl 4):11.
the histopathological outcome of transient focal cerebral 19. Boysen G, Christensen H. Stroke severity determines body
ischemia. Stroke 1996;27:2274–81. temperature in acute stroke. Stroke 2001;32:413.
5. Dietrich W, Busto R, Valde¢s I, Loor Y. Effects of nor- 20. Sharma JC, Fletcher S, Vassallo M, Ross I. What influ-
mothermic versus mild hyperthermic forebrain ischemia in ences outcome of stroke-pyrexia or dysphagia? Int J Clin
rats. Stroke 1990;21:1318–25. Pract 2001;55:17–20.
6. Kammersgaard LP, Jørgensen HS, Rungby JA et al. 21. Wong AA, Davis JP, Schluter PJ, Henderson RD, OÕSul-
Admission body temperature predicts long-term mortality livan JD, Read SJ. The effect of admission physiological
after acute stroke: The Copenhagen Stroke Study. Stroke variables on 30 day outcome after stroke. J Clin Neurosci
2002;33:1759–62. 2005;12:905–10.
7. Reith J, Jorgensen HS, Pedersen PM et al. Body temper- 22. Prasad K, Krishnan PR. Fever is associated with doubling
ature in acute stroke: relation to stroke severity, infarct of odds of short term mortality in ischemic stroke. Int J
size, mortality and outcome. Lancet 1996;347:422–5. Stroke 2008;3(Suppl 1):418.
8. Azzimondi G, Bassein L, Nonino F, Fiorani L, Vignatelli L. 23. Hajat C, Hajat S, Sharma P. Effects of poststroke pyrexia
Fever in acute stroke worsens prognosis: a prospective on stroke outcome: a meta-analysis of studies in patients.
study. Stroke 1995;26:2040–3. Stroke 2000;31:410–4.
9. Castillo J, Martinez F, Leira R, Prieto JM, Lema M, Noya 24. Review Manager (RevMan) [Computer program]. Version
M. Mortality and morbidity of acute cerebral infarction 4.2 for Windows. Copenhagen: The Nordic Cochrane
related to temperature and basal analytic parameters. Centre, The Cochrane Collaboration, 2003.
Cerebrovasc Dis 1994;4:56–71. 25. Greer DM, Funk SE, Reaven NL, Ouzounelli M, Uman
10. Terent A, Andersson B. Prognosis for patients with cere- GC. Impact of fever on outcome in patients with stroke
brovascular stroke and transient ischaemic attacks. Ups J and neurologic injury: a comprehensive meta-analysis.
Med Sci 1981;86:63–74. Stroke 2008;39:3029–35.
11. Huo CN, Zhao YX. Relationship between fever and mor- 26. Dippel DWJ, van Breda EJ et al. Effect of paracetamol
tality in patients with acute ischemic cerebrovascular (acetaminophen) on body temperature in acute ischemic
stroke [in Chinese]. Chin J Nurs 1997;32:640–2. stroke: a double-blind, randomized Phase II Clinical Trial.
12. Hanchaiphiboolkul S. Body temperature and mortality in Stroke 2001;32:1607–12.
acute cerebral infarction. J Med Assoc Thai 2005;88:26– 27. den Hertog HM, van der Worp HB, van Gemert HM et al.
31. The Paracetamol (Acetaminophen) In Stroke (PAIS) trial:
13. Roy MK, Ray A. Effect of body temperature on mor- a multicentre, randomised, placebo-controlled, phase III
tality of acute stroke. J Assoc Physicians India 2004;52: trial. Lancet Neurol 2009;8:434–40.
959–61. 28. Den Hertog HM, van der Worp HB, Tseng MC, Dippel
14. Georgilis K, Plomaritoglou A, Dafni U, Bassiakos Y, DW. Temperature-lowering therapy for acute stroke.
Vemmos K. Aetiology of fever in patients with acute stroke. Stroke 2009;40:e481–2.
J Int Med 1999;246:203–9. 29. Coimbra C, Drake M, Boris-Moller F, Wieloch T. Long-
15. Jorgensen HS, Reith J, Pedersen PM, Nakayama H, Olsen lasting neuroprotective effect of postischemic hypothermia
TS. Body temperature and outcome in stroke patients. and treatment with an anti-inflammatory ⁄ antipyretic drug.
Lancet 1996;348:193. Letter. Stroke 1996;27:1578–85.
408