Meningitis is characterized by vaso-dilatation, hyperemia and increased vascular permeability of blood vessels, all of which manifest as changes in cytology and biochemistry of cerebrospinal fluid (CSF). The release of cytokines and other mediators affect the temperature center at hypothalamus. The tympanic membrane, due to its proximity to the site of inflammation and because of the fact that it shares a common blood supply with the hypothalamus, the thermostat of the body, makes it a potential site for measurement of temperature in such patients.

The aim of the present study was to assess the accuracy of tympanic membrane temperature (TMT) in predicting "core" body temperature, to compare TMT and RT in two groups of febrile pediatric patients, one with meningitis and the other with some other focus of infection and to ascertain if any correlation existed between TMT and various parameters of CSF profile. The hypothesis to be tested was whether local (CNS) inflammatory changes in meningitis are reflected in tympanic thermometry or not.

The study was conducted at the Emergency Division of Kalawati Saran Children’s Hospital between November 1999 to April 2000. A total of 130 subjects with fever were enrolled at initial triage. The ‘Cases" included 60 babies with an acute febrile illness of less than 3 days duration, having signs and symptoms of meningitis and a suggestive CSF profile, according to standard cut off values(2). No attempt was made to further classify etiology of menin-gitis. Sixty febrile patients without features of meningitis were taken as "Controls". Children less than six months of age, (as the orientation of ear probe in them may be difficult due to its size, hence resulting in false readings), those with suppurative/serous otitis media (as it may slightly elevate the ear temperature albeit by a clinically insignificant arnount)(3), otitis externa and moderate to large amounts of wax were excluded from the study. Those having anal fissures or receiving enemas and those who had ingested anti-pyretics within six hours of measurement were also excluded. There were 10 exclusions, 6 in "Cases" and 4 in "Controls".

Ambient temperature was measured by using a mercury column thermometer and kept within a range of 20°C-24°C. The patients were subjected to measurement of TMT and RT. A cut off of 37.6º C for each site was used to define a febrile patient(4). CSF analysis for cytology was done by the principal author while biochemical profile was assessed in laboratory.

RT was obtained by using a digital electronic thermometer with probe inserted 2 cm into the rectum, till an audible beep signaled calibration with body temperature. Rectal thermometer was pre-calibrated against a controlled temperature water bath over a range of 35º-41°C and demonstrated an accuracy of ±0.1ºC. TMT was measured by using Braun’s Thermoscan Instant Thero-meter IRT 1020 fabricated by Thermoscan Inc., San Diego, California U.S.A., which undergoes self calibration tests before every reading. The instrument consists of an ear probe, which uses pyro-electric sensors to detect infrared rays emitted from the tympanic membrane. After otoscopic examination (conducted by AS), the probe was gently introduced into the ear canal and oriented towards the contralateral eye, while traction (ear tug) was applied to the pinna. In children less than one year of age, traction was applied posteriorly while in older children it was in postero-superior direction(5). The trigger was depressed for two seconds. The plastic probe cover was changed with each measurement. Measurements were repeated as soon as possible after the instrument’s electronic circuitry had reset. Attempts were made to take all readings in a patient within five minutes and in following order: TMT ® RT. Two readings from each ear were recorded and a mean taken, while for RT, a single reading was taken. Informed consent was taken prior to including subjects in the study.

Statistical analysis of the data was done by computerized software STATISTICA. The association between RT and TMT was studied by Karl Pearson correlation coefficient and was tested for significance by Student’s ‘t’ test. The TMT values between cases and controls were tested by Students ‘t’ test. The relationship between TMT and various parameters of CSF profile (cell count, proteins and sugar levels) was examined by linear multiple regression technique.

There was no statistically significant difference between the mean age in Cases (4.37 ± 2.57, range 0.5-9 years) and in Controls (4.47 ± 2.54, 0.67-9 years). The overall mean age was 4.4 years. There was no statistically significant correlation between age of the patient and temperature readings taken from the two sites in both the groups, the coefficient of correlation (r) in age vs RT in "Cases" and "Controls" being 0.091 and 0.093 respectively while for age vs TMT in "Cases" and "Controls" they were 0.163 and 0.077 respectively. Since a highly significant correlation was observed between TMT of two ears (r = 0.987) a mean of the two was used for further analysis and comparisons. On paired sample analysis, significantly high correlations between RT and TMT were seen in both "Cases" (r = 0.812) and "Controls" (r = 0.994) and these existed over a wide range of age and temperature readings. A straight line (Fig. 1) denotes this in both the groups. The mean TMT in "cases" was 39.3ºC ± 0.83, which was significantly higher than that in "controls" (38.9ºC ± 0.9) (P = 0.04). The mean RT in "cases" was 38.4ºC ± 1.0 while in "controls" it was 38.8ºC ± 0.8 (P = 0.11). The overall ranges of temperature readings were quite wide in both the groups for both RT and TMT. Analysis of ear and rectal temperature amongst the "Cases" (39.3 ± 0.8 vs 38.4 ± 1.0) revealed a temperature difference of 0.8° C ± 0.5º C, which was statistically highly significant (P = 0.001). No such observation was seen in controls (0.1ºC ± 0.1ºC).

Fig. 1. Correlation Between RT and TMT in Cases and Controls (r = 0.81)

One of the aims of our study was to assess any association between TMT and various parameters of CSF profile in patients of meningitis. It was observed that there was a gradual increase in mean TMT with increasing number of cells and proteins in CSF and declining levels of CSF sugar. Although all the correlations were statistically significant, maximum correlation was seen with CSF cytology (r = 0.79) and is depicted in Fig 2. The coefficient of correlation (r) between TMT and CSF protein and sugar were 0.74 and –0.73 respectively. RT readings were well correlated with CSF profile (vs cytology = 0.77, vs CSF protein = 0.74, vs CSF sugar = – 0.71).

Fig. 2. Correlation Between TMT and Number of Cells in CSF in Cases (r = 0.79)

After calculating predicted RT and TMT using the multiple regression equation, sensitivity and specificity for detecting fever was determined. The sensitivity, specificity, positive and negative predictive values were best at 39.5º C and are depicted in Table I.

Table I__ Sensitivity and Specificity of Tympanic Membrane Temperature (TMT) to Predict Fever as Compared to Rectal Temperature (RT).
 

Criterion for defining fever

38º C

38.5º C

39º C

39.5º C

40º C

Sensitivity

1.00

1.00

1.00

1.00

1.00

Specificity

0.76

0.80

0.85

0.89

0.88

PPV*

0.78

0.87

0.72

0.71

0.75

NPV*

1.00

1.00

1.00

1.00

1.00

Tympanic thermometry is a relatively new concept and not much research work has been done using this, at least in developing countries. This technology utilizes the advantages of pyroelectric sensors to detect infrared rays emanated from the tympanic membrane, without any direct contact between the ear probe and the tympanic mem-brane. Obtaining an accurate measurement depends on proper orientation of the probe since misdirection towards the canal wall results in lower readings. In previous studies correlation at age <3 months have been not satisfactory(5), and these were explained on anatomic grounds i.e., size of the probe may be too large for their ear canal.

Talo, et al.(7) had demonstrated signifi-cantly large (P = 0.01) correlations between temperatures in the left and right ear. There was good correlation between TMT and RT in both groups. Results of our study are in agreement with previous work. Terndrup, et al.(8) found excellent correlations (r = 0.90) between TMT and RT in 102 patients (age unspecified). In another study(9), multiple comparisons of PAC, RT and TMT were performed in 9 adult ICU patients using PAC temperature as the gold standard. It was seen that PAC and TMT were not significantly different and the performance of TMT and RT was similar in the ability to predict PAC temperature. Shinozaki, et al.(10) found TMT correlating well with PAC thermistor over a range of 34°C to 39°C (r = 0.98). TMT is considered an excellent indicator of core body temperature and can be especially important in dynamic situations where the patient’s core temperature is changing rapidly, since RT, long considered "gold standard", has been shown to lag behind these changes and may underestimate core body temperature. In the present study too, high correlations between TMT and RT were observed.

The observation that TMT was higher in "Cases" implies that there is a definite role of the local (CNS) inflammatory pathologic changes in patients with meningitis. The tympanic membrane, due to its proximity to the site of inflammation, and due to the fact that it shares a common blood supply with the hypothalamus, the thermostat of the body, makes it a potential site for measurement of temperature in these babies. A correlation between CSF profile and TMT was especially sought as increasing severity of inflammation resulted in profuse alterations in CSF profile and reflected in increase in TMT readings. The ambient temperature was kept within a narrow range of 20°C-24°C. In a study on adults, Zehner, et al.(11) observed that TMT has a predictable relationship with RT at normal ambient temperature, however in warm examination areas (35ºC) spuriously elevated oral and TMT may occur. Brinnel, et al.(12) reported a change of 0.0025ºC in TMT per °C change in ambient temperature. In our study, the sensitivity at various temperature cut offs was 100% while there was gradual increase in specificity with rising temperature. Chamberlain, et al.(5) had reported a sensitivity of 80% at 38ºC and 38.5ºC with a specificity of >90%, but had used the maximum rather than the mean of TMT readings. Terndrup, et al.(7) reported increasing specificity with rising temperature from 38ºC to 39.5ºC. Our results are in agreement with previous work done.

On comparing both the techniques, TMT is an easy, rapid and clean procedure with no contact with mucous membrane and is acceptable to parents and children. Added advantages include lack of influence of factors known to spuriously affect oral temperature like (recent fluid intake, mouth breathing, tachypnea) and a broader range of temperature over which TMT are reportedly accurate. TMT can reliably predict "core" temperature over a wide range of readings and TMT is significantly higher in patients with menin-gitis as compared to non-meningitic febrile patients.

Contribution: AS and NKD conceptualized the study while AS collected the data. AS, NKD and MCJ were instrumental in analysis, interpretation and result formulation. AS and NKD approved the final draft and AS wrote the paper.

Funding: None.

Competiting interest: None declared.