Acute Phase Response Detection and Quantitation at the Point of Care in Older Adults With Acute Bacterial Infections

Results. There were 144 patients and 144 controls at a respective mean 6 standard deviation age of 71.3 6 20.7 and 70.6 6 20.2 years. A highly significant difference was noted between patients and controls in all laboratory markers, both conventional and those obtained by the slide test. By using hs-CRP, we correctly predicted the individual group as control or bacterial infection. When analyzed by means of a receiver-operated characteristic (ROC) curve, hs-CRP was again most reliable, with the slide leukocyte test being superior to the WBCC.

T HE prompt identification of acute bacterial infections in older adults might have significant prognostic and therapeutic implications.It is also important to recognize those patients who react with an overwhelming inflammatory response to the invading microorganism.It was suggested that this hyperinflammatory response is frequently in itself deleterious to the patient (1).
We have previously reported that a simple slide can be used for the identification of the above-mentioned acute phase response (2,3).The availability of this approach via telemedicine may enable the determination of the intensity of the inflammatory response even in residents of nursing homes where modern laboratory services are not constantly available.
In the present study we used an improved method of our previously described system in a model of acute bacterial infections in older adults.We examined the reliability of this system as compared to accepted conventional markers of the acute phase response.

METHODS
Written informed consent was obtained from all participants.The participating units included the Emergency Department and the Department of Internal Medicine ''D'' at the Tel Aviv Sourasky Medical Center, and the Emergency Department and the Departments of Internal Medicine and Geriatrics at the Shaare Zedek Medical Center in Jerusalem, Israel.We included older adults presenting with signs and symptoms consistent with an acute bacterial infection.
The diagnosis of an acute bacterial infection was based on the typical clinical presentation in patients with erysipelas or cellulitis, the presence of lobar pneumonia by chest x-ray, positive blood cultures in patients with sepsis, and urine cultures in individuals with urinary tract infections and pyelonephritis.The diagnosis of pyelonephritis was based on the presence of fever with or without chills, urinary findings, and flank tenderness.Patients were recruited in the two medical centers situated in Tel Aviv and Jerusalem by following a strict unified protocol that ensured the same recruitment criteria.Although we can not be sure that patients with atypical pneumonia were included, it is the practice of most clinicians to assume that adults with fever, cough, and a pulmonary infiltrate should be treated as patients with bacterial infections until otherwise proven.Patients were excluded if they had received antibiotic treatment prior to the presentation, if they had an underlying inflammatory condition (such as, among others, arthritis or inflammatory bowel disease), or if they were on steroidal or nonsteroidal anti-inflammatory medications (except for aspirin in a dosage of up to 325 mg/d).These exclusions were needed to make sure that the inflammatory reactions result from the acute infectious disease and not from an underlying disease.

Controls
Age-and sex-matched controls were included if they had no overt infectious and inflammatory disease or condition in the 6-month period prior to their recruitment.They were recruited from a group of retired employees of both Medical Centers that participated in the study.

Laboratory Methods
Blood count was performed by using the Coulter STKS (Beckman Coulter, Nyon, Switzerland) electronic analyzer.The erythrocyte sedimentation rate (ESR) was determined by the method of Westergren, fibrinogen by the method of Clauss and using a Sysmax 6000 (Sysmex Corporation, Hyaga, Japan) analyzer, and the high sensitivity C-reactive protein (hs-CRP) concentrations were determined by using the Boering BN II Nephelometer (DADE Boering, Marburg, Germany) analyzer.

Slide Test
Venous blood was gently mixed with sodium citrate 3.8% at a ratio of 3 volumes of whole blood to one volume of citrate.The blood was applied to a glass slide kept at an angle of 308 as described (4).After several minutes they were analyzed with an image analyzer (5) consisting of a Pentium Win 95 equipped with a Matrox Meteor (Matrox Ltd., Montreal, Canada) color frame grabber, a color CCD camera, and a microscope which was operated at 3200 magnification, resulting in an image resolution of 0.4 micron per pixel.The analysis was performed by a technician who was completely blinded to any clinical or laboratory data of the patients or the controls.Two variables were obtained, namely, the number of white blood cells per square millimeter (slide leukocytes) and the erythrocyte percentage (EP), the latter being a marker of the humoral acute phase response (6).The EP is practically the area covered on the slides by the erythrocytes.For example, if there is no aggregation, 100% of the slide is covered by the cells.When they begin to aggregate, free spaces are created between the aggregates and this percentage is reduced.The validity of the slide test has been confirmed in previous studies (7)(8)(9).

Statistical Analysis
All continuous variables were summarized and displayed as mean 6 standard deviation (SD), and all the categorical data were summarized and displayed as the number and percentage of participants in each group.Because the hs-CRP has a non-normal distribution, we thus used a logarithmic transformation; the results are expressed as back-transformed geometric means and SDs.For all continuous variables, a comparison between the participants with bacterial infection and the healthy controls was performed using Student's t test for independent samples.We used discriminant analysis to determine the ability of the different inflammatory variables to correctly classify the participants according to group (bacterial infection or control).To further evaluate the performance of classification schemes and to compare the different inflammatory variables (classifiers), we used a receiver-operated characteristic (ROC) curve analysis.We calculated the area under the curve to compare the classifiers and the asymptotic statistical significance to reject the hypothesis that the curve is similar to the reference line, which is a random classifier.The SPSS statistical package was used to perform all statistical evaluations (SPSS Inc., Chicago, IL).

RESULTS
Following the exclusion of 14 individuals, we included a total of 144 patients (68 women and 76 men with a mean 6 SD age of 71.3 6 20.7 years) with acute bacterial infections.Controls included 144 apparently healthy individuals with no infectious or inflammatory conditions aged 70.6 6 20.2 years with a sex distribution identical to that of the patients.The frequencies of the different diagnoses in the group of patients with acute bacterial infections are reported in Table 1.A highly significant difference was noted between patients and controls in all laboratory markers, both conventional and those obtained by our image analyzer (Table 2).The ability of the different variables to correctly predict the group as control or bacterial infection is reported in Table 3; hs-CRP was superior to the others, followed by the number of polymorphonuclear leukocytes, WBCC, ESR, EP, and slide leukocyte count.However, it should be noted that the absolute difference between the hs-CRP and EP (markers of the humoral acute phase response) was 12.3%, whereas that between the WBCC by the electronic cell analyzer and our slide test was only 0.6%.
We evaluated our results further by using ROC analysis (Table 4).By this method, hs-CRP was most reliable, followed by the number of polymorphonuclear leukocytes, slide leukocytes, ESR, WBCC, and EP.In this model, a better ROC curve was noted for the slide leukocytes than for the WBCC by the electronic analyzer.The absolute difference between both markers of the humoral acute phase response (hs-CRP and EP) was 13 in favor of hs-CRP.Typical pictures obtained from several patients and a control individual are shown in Figure 1.

DISCUSSION
The presence and intensity of an acute phase response are easily identified by simple biomarkers, such as hs-CRP, WBCC, and ESR.Whereas paramedical personnel can perform the ESR, quantitative hs-CRP and WBCC require laboratory facilities that are not readily available, particularly in small nursing homes and in centers situated in the periphery.There is thus a need to develop an alternate approach that is readily available.We have previously described such an approach which is based on a count of the number of white blood cells in an unstained peripheral blood slide (10,11) and on the degree of erythrocyte adhesiveness and aggregation (12)(13)(14)(15).This approach permits a dual analysis, namely, that of the cellular acute phase response (number of leukocytes) and the humoral one (expressed as the degree of erythrocyte adhesiveness and aggregation).
The main finding of the present study is that, in older adults, the slide leukocyte count provided a similar diagnostic yield to the WBCC by electronic analyzer in differentiating between persons with acute bacterial infections and matched controls.Similar results were obtained when comparing the slide EP to the Westergren ESR.The best biomarker was quantitative hs-CRP, but the absolute difference between the results of hs-CRP and slide biomarkers was not substantial.This difference was 6.6% between hs-CRP and slide leukocytes (Table 3), and the difference between the ROC curves was even smaller (5%).
A potential advantage for a given individual is that the picture obtained by the slide test can be easily transmitted to a remote location where it can be analyzed and the results reported to the user (10,11).Further, the establishment of an Inflammation Data Center ( 16) allows an automatic matching of appropriate controls.In fact, data from more than 3000 apparently healthy individuals are already   available at this center and are used to provide matched controls.This is of particular importance considering the possibility that many apparently healthy individuals may suffer from low grade inflammation (17).Thus, when faced with acute infection, it is important to distinguish between the baseline degree of inflammation and the acute response.
The findings described in this study show that our approach is probably as accurate as an approach using the electronic analyzer WBCC.The possibility of using the slide leukocyte count for follow-up purposes is currently under investigation.Although the results obtained are inferior to those of quantitative hs-CRP, the absolute differences are relatively small.
We did not include individuals with viral infections in the present study.This has been evaluated previously (18)(19)(20) with a similar diagnostic yield for the slide test as compared to conventional biomarkers.

Conclusion
The slide test described provides a reasonable alternate approach when used for the detection and quantitation of the acute phase response in older adults with acute bacterial infections.This approach should be further analyzed in view of its potential to become a telemedicine-based real-time biomarker at the site of care provision.

Figure 1 .
Figure 1.Typical picture obtained from a patient with gram-negative sepsis (left) and an appropriate control (right).

Table 1 .
Frequency of the Different Diagnoses in the Bacterial Group

Table 2 .
Minimum, Maximum, Mean, and Standard Deviations (SD) of Different Inflammatory Parameters in Both Groups and Statistical Significance Between the Means Note: ESR ¼ erythrocyte sedimentation rate; hs-CRP ¼ high sensitivity C-reactive protein; WBCC ¼ white blood cell count; PMN ¼ polymorphonuclear leukocytes; EP ¼ erythrocyte percentage.

Table 3 .
Percent of Correct Classification of the Different Variables Predicting the Control or Bacterial Group

Table 4 .
Receiver-Operated Characteristic (ROC) Curve Analysis of the Different Variables Predicting the Controls and the Bacterial Groups