Temporal profile of body temperature in acute ischemic stroke: relation to stroke severity and outcome

Background Pyrexia after stroke (temperature ≥37.5°C) is associated with poor prognosis, but information on timing of body temperature changes and relationship to stroke severity and subtypes varies. Methods We recruited patients with acute ischemic stroke, measured stroke severity, stroke subtype and recorded four-hourly tympanic (body) temperature readings from admission to 120 hours after stroke. We sought causes of pyrexia and measured functional outcome at 90 days. We systematically summarised all relevant previous studies. Results Amongst 44 patients (21 males, mean age 72 years SD 11) with median National Institute of Health Stroke Score (NIHSS) 7 (range 0–28), 14 had total anterior circulation strokes (TACS). On admission all patients, both TACS and non-TACS, were normothermic (median 36.3°C vs 36.5°C, p=0.382 respectively) at median 4 hours (interquartile range, IQR, 2–8) after stroke; admission temperature and NIHSS were not associated (r2=0.0, p=0.353). Peak temperature, occurring at 35.5 (IQR 19.0 to 53.8) hours after stroke, was higher in TACS (37.7°C) than non-TACS (37.1°C, p<0.001) and was associated with admission NIHSS (r2=0.20, p=0.002). Poor outcome (modified Rankin Scale ≥3) at 90 days was associated with higher admission (36.6°C vs. 36.2°C p=0.031) and peak (37.4°C vs. 37.0°C, p=0.016) temperatures. Sixteen (36%) patients became pyrexial, in seven (44%) of whom we found no cause other than the stroke. Conclusions Normothermia is usual within the first 4 hours of stroke. Peak temperature occurs at 1.5 to 2 days after stroke, and is related to stroke severity/subtype and more closely associated with poor outcome than admission temperature. Temperature-outcome associations after stroke are complex, but normothermia on admission should not preclude randomisation of patients into trials of therapeutic hypothermia.


Background
Elevated body temperature (pyrexia) is said to be common after stroke [1,2] and is associated with poor outcome [2][3][4][5][6][7][8][9][10]. However, the timing of pyrexia after stroke varied between studies (Table 1). Pyrexia occurring early might reflect the systemic response to ischemic brain damage [11] whereas pyrexia occurring later might be due to complications of severe stroke, such as infection and deep venous thrombosis (DVT). However data on timing, causes and associations between body temperature and outcome vary.
Amongst the 18 previous studies of temperature after stroke (Tables 1 and 2), four performed only one temperature recording [3,6,12,20], some measured temperature at admission but relatively late (e.g. 24 hours) [11] or at unspecified times after stroke [2,4,7]. Where serial readings were obtained, this was only up to three days after stroke [8,9,13,15,18,22]. Six studies included haemorrhagic and ischemic stroke [2][3][4]6,12,19]. Definitions of pyrexia, temperature measurement methods, reporting of associations with stroke severity and of causes of pyrexia varied widely [2,7,10,14,16,17,21]. Thus, while some studies found associations between higher admission temperature and poor outcome (Table 2) [3,6,12], others found that peak temperature occurring several days after stroke, not admission temperature, was associated with poor outcome [7,16,17,19]. The reasons for these differences are unclear but important to understand to limit brain damage and improve outcome after stroke. We obtained measurements of body temperature every four hours from admission to five days after acute ischemic stroke as part of a study to examine serial changes in the ischaemic lesion on MR imaging. We used these data to clarify the temporal profile, associations with stroke severity, subtype and functional outcome and proportion with an alternative explanation for pyrexia.

Methods
We prospectively recruited patients >18 years old who presented with potentially disabling ischemic stroke. We excluded patients with intracerebral haemorrhage (ICH), coma, serious intercurrent illness, and technical or clinical incompatibility with MR scanning. Ethical approval was granted by the Scottish Multi-Centre Research Ethics Committee (06/MRE00/119) and we obtained informed, written consent from all patients or their relatives.
At admission, we recorded National Institute of Health Stroke Score (NIHSS) and the Oxford Community Stroke Project (OCSP) stroke subtype classification [23]. All patients underwent MR imaging to diagnose the ischemic stroke. We recorded temperature using a First Temp Genius W tympanic thermometer immediately on Table 2 Previous studies of body temperature and outcome after stokekey results

First author and year
Key findings Azzimondi 1995 [2] High fever (≥37.9°C) <7 d is independent risk factor for poor prognosis. Fever occurred in 43% of stroke pts <7 d.
Onset of fever occurred in first 2 days in 64% of febrile patients.

Reith 1996 [3]
Admission body temp is independently related to stroke severity, lesion size, mortality and outcome. [unclear how measured "outcome"; didn't separate AIS from ICH] Castillo 1998 [9] The relationship between the degree of hyperthermia and stroke outcome/FIV is strongest when it begins within 24 h of symptom onset.

Georgilis 1999 [4]
Fever in stroke is assoc with ↑age, ↑severity, more invasive techniques, worse outcome. When fever present without focus of infection, it tends to occur earlier.

Boysen 2001 [8]
Temp < 6 h post stroke onset has no prognostic influence on 3 month mRS. More severe strokes have higher temperature in first 48 h. [Also looked at ICH]. 7 d fatality rate higher in patients with lower body temp on admission.

Kammersgaard 2002 [6]
For all strokes, a 1°difference in admission body temperature gives 30% increase in relative risk of 5 yr mortality.
No association between admission temp and survival in pts still alive at 3 months.

Audebert 2004 [13]
Larger stroke volume and greater NIHSS assoc with higher temp, CRP and WCC. Successful thrombolysis attenuates inflammatory response

Idicula 2008 [16]
Body temp before thrombolysis was not assoc with 3 month outcome, but high temp thereafter was.

Millan 2008 [17]
Body temp ≥37 at 24 h but not at baseline was assoc with lack of recanalisation, greater hyperdensity volume and worse functional outcome, regardless of stroke severity and time to treatment

Saini 2009 [7]
Hyperthermia assoc with poor outcome. Delayed hyperthermia is more strongly assoc with poor outcomes than early hyperthermia. No association between baseline hyperthermia and outcome.
den Hertog (PAIS) 2009 & 2011 [18,19] Baseline body temp was not related to improvement. Increased body temp at 24 h was associated with low likelihood of improvement.

Naess 2010 [20]
High body temperature was associated with favorable short-term outcome in those who were thrombolysed vs. those not thrombolysed Phipps 2011 [21] High "fever burden" (combination of fever height and duration) was associated with death or with referral to hospice arrival at hospital, and four hourly thereafter up to 120 hours. Recordings were taken from the uppermost ear in most patients, and from both ears in 11 patients to test side-to-side variation [24]. Tympanic thermometry is considered reliable for serial readings [25], shows the least variation with age compared with axillary, rectal and oral thermometry [26], and is widely used in clinical practice. Pyrexia was defined as temperature ≥37.5°C [10]. We searched for causes of pyrexia by collecting data on evidence of infection (evident infectious agent, leukocytes in body fluid, signs on imaging), DVT and surgical intervention as well as paracetamol and antibiotic use. Our policy is to prescribe paracetamol to patients who develop pyrexia (or for other indications such as pain relief) and antibiotics when positive evidence of infection and an infectious agent is diagnosed. We recorded modified Rankin Scale score (mRS) at three months after stroke, blind to temperature measurements. We compared admission, peak and final temperatures and calculated area under the temperature/time curve (AUC) [27] which we standardised (AUC [s] ) by assuming all patients had a temperature of 36.5°C between stroke and the admission recording, and dividing AUC by the time of final temperature reading [27]. We tested the association between temperature profile and measures of stroke severity firstly as determined by NIHSS, secondly by comparing total anterior circulation strokes (TACS) vs. less severe stroke subtypes (non-TACS) and thirdly good (mRS ≤2) vs. poor (mRS≥3) 90 day outcome. We adjusted analysis for age and admission NIHSS where appropriate. We used Student's t-test for parametric and Mann Whitney U tests and Spearman correlation for non-parametric data.
There was no difference between ipsi-and contralateral tympanic temperature readings obtained simultaneously (36.6°C vs 36.6°C, 95% confidence interval (CI) -0.19 to 0.16, p=0.821) in 11 patients (five TACS). The mean number of temperature readings per patient was 14 (SD 6.9). No patients died during the observation period.

Discussion
In patients with ischemic stroke, tympanic temperature was not elevated on admission even in patients with more severe strokes (TACS), and admission temperature did not correlate with admission NIHSS. Instead, we found that peak temperature, occurring at around 1.5 to 2 days after stroke and overall temperature, as expressed by AUC were associated with admission stroke severity as measured by NIHSS and TACS subtype. Peak temperatures were higher, occurred later and temperature elevation lasted longer in more severe than less severe strokes perhaps indicating a more prolonged, greater inflammatory response to the volume of infarcted tissue. Patients with poor functional outcome (mRS≥3) at 90 days had higher admission and peak temperatures than patients with good outcome (mRS≤2), although all admission temperatures were <37°C. We were not able to find a source of infection in 69% of pyrexial patients, and no alternative cause of pyrexia, other than the stroke itself, in 44%. This prospective study of detailed four-hourly tympanic temperature measurements up to five days after ischemic stroke helps explain the varying results of previous studies (Tables 1 and 2) some of which found associations between admission temperature, stroke severity and outcome [3,4,6,12,13], while others did not [8,14,16,19]. This variation can be attributed to some studies being retrospective, or not specifying the timing of temperature recording after stroke, sampling temperature only once, measuring "admission" temperature relatively late after stroke, only measuring serial temperature up to 72 hours, the different severities of stroke included in each study or including patients with both haemorrhagic and ischemic stroke. Consistent with our results, others showed that patients with more severe strokes experience higher peak temperatures [15], and that elevated temperature at 24 hours [14,17,19], 48 hours [8] or 7 days [7] after stroke was more closely linked to poor outcome than admission readings. Our detailed longitudinal findings also demonstrate the higher, later and longer duration of temperature elevation in more severe than less severe stroke, which perhaps has not been appreciated previously.
Our study has limitations. The small sample size was constrained by selection of patients for an MR imaging study, but on the other hand it allowed very detailed temperature monitoring. However, the range of stroke severity was consistent with those that would be considered for trials of therapeutic hypothermia. A larger sample size would allow more adjustment for potential confounders, and clarification of the significance of any differences in timing of peak temperature readings, and comparison of patients whose tympanic temperature  may have been affected by antibiotics and antipyretics, although our results show that pyrexia was just as common in patients who were prescribed paracetamol as in those who were not, in contrast to others' results [15]. However paracetamol may have influenced the profile of temperature change. Although our data on temperature profile and stroke severity are consistent with five previous studies, we cannot exclude the possibility of an association between admission temperature and stroke severity. The study strengths are the detailed four-hourly tympanic temperature measurements for 120 hours after ischemic stroke and the detailed comparison with stroke subtype, severity and outcome. The duration of temperature recording ensured that the peak temperature was captured in both TACS and non-TACS.
Two other points raise questions for further study. Admission temperature in the patients with more severe strokes, ie TACS, was not higher than in patients with milder strokes, as might have been expected. Perhaps, by analogy with other serious acute illness, the temperature in severe stroke may reflect severe illness [28]. Secondly, why do patients with a poor outcome have a marginally higher (but still normothermic) admission temperature, and while severe stroke is associated with poor outcome, severe stroke is not associated with admission temperature? This might be explained at least in part by a possible cascade of events suggested in experimental data. Increased temperature opens the blood-brain barrier [29] which, in acute ischemia, would lead to increased extracellular oedema, more infarct swelling, more restricted capillary flow in the ischemic tissue, less chance of reperfusion, all contributing to increasing ischemic damage, swelling and leading to a larger infarct around 48 hours [30], consolidating the potential for tissue damage that was suggested by the severe stroke symptoms at presentation, and consequently leading to the poor outcome at 90 days. Longer duration of temperature monitoring should be considered in future research to improve understanding of temperature profiles after stroke.
This interpretation, if true, raises some implications for therapeutic hypothermia trials. Firstly, and paradoxically, patients may benefit the most from hypothermia if it prevents the tissue cascade outlined above from causing more tissue damage and should certainly not be excluded from trials; thus cooling should be initiated as early as possible and not influenced by the patient's admission temperature. However cooling may only prevent worsening of damage, not actively salvage tissue that is already at risk, so combinations of therapies to salvage (e.g. thrombolysis) as well as to restrict progressive new damage (e.g. hypothermia, if it works) may be required. Secondly, if the main effect of hypothermia is to reduce blood-brain barrier opening thus preventing the cascade that leads to larger stroke lesions, then hypothermia could still be valuable if started many hours after the stroke when the possibility of salvaging at risk tissue was lost but there was still some "future secondary damage" to prevent. If correct, then hypothermia should reduce subacute infarct oedema and mass effect. Thirdly, if correct, then therapeutic thrombolysis and hypothermia should work synergistically to produce greater benefit than either alone. However hypothermia may also have some disadvantages. In addition to increased risk of secondary infection and the need to manage unpleasant side effects like shivering, lower temperatures might delay thrombus lysis which pyrexia might accelerate. The balance of risk and benefit presented by these possibilities requires testing in future therapeutic trials of hypothermia after stroke.