Does near-to-patient testing contribute to the diagnosis of streptococcal pharyngitis in adults?

Does near-to-patient testing contribute to the diagnosis of streptococcal pharyngitis in adults? Per Hjortdahl' and Hasse Melbye2 'Department of General Practice, University of Oslo, Postbox 1130 Blindem, N-0318 Oslo, Norway?Institute of Community Medicine, University of Troms0, Breivika, N-9000 Troms0, Norway Hjortdahl P, Melbye H. Does near-to-patient testing contribute to the diagnosis of streptococcal pharyngitis in adults? Scand J Prim Health Care 1994;12:70-6. Objective - To assess the efficacy of near-to-patient laboratory testing in diagnosing group A beta haemolytic streptococci throat infection in adults, alone and in conjunction with the doctor's clinical judgement. Setting and subjects - 174 primary care patients with a sore throat, of which 59 (34%) were identified as having group A streptococci. Main outcome measures - The erythrocyte sedimentation rate, leucocyte count, and C-reactive protein, and a specific streptococcal immunological assay were evaluated separately and in conjunction with the doctor's clinical diagnosis. The presence of GAS throat infection, the reference standard, was defined as those patients presenting with a chief complaint of sore throat and having a positive GAS culture or a four-fold or more increase or fall of AST or ADNase B serum levels during a four-week observation period. Results - The discriminatory ability of the sedimentation rate was not satisfactory and added little useful clinical information. Leucocytes and C-reactive protein both yielded clinically significant information and had similar test characteristics. The immunological test had the best characteristics of the tests evaluated. Conclusion - Near-to-patient testing, except the erythrocyte sedimentation rate, may, in addition to the clinical evaluation, contribute to the differential diagnosis of streptococcal pharyngitis in adults. Key words: pharyngitis, sore throat, near-to-patient testing, erythrocyte sedimentation rate, leucocyte count, C-reactive protein, rapid immunological assays, primary care. Per Hjortdahl, MD, Department of General Practice, University of Oslo, P.O. Box 1130 Blindern, N-0318 Oslo, Norway. Sore throat is a common reason for consultation in primary health care. The differentiation between throat infections caused by group A beta haemolytic streptococci (GAS) and those due to other infectious agents has major practical implications. It is generally accepted that GAS infections should be treated with antibiotics, while other, uncomplicated throat infections should re- ceive symptomatic treatment alone ( 1). The differential diagnosis of throat infections constitutes a major challenge to the clinician. By evaluating symptoms and signs it is difficult to establish an aetiologic diagnosis; this is best done by microbiological culture (2). Culture, however, is often of limited practical value because the results are first available after 24 hours. Many doctors find

this waiting period unacceptable and base their treatment decision on the clinical diagnosis alone (3). New technical developments and easy-to-perform immunological tests have renewed interest in the general practitioner s laboratory and nearto-patient testing (4). The efficacy of these procedures is often evaluated isolated from clinical reality (5,6). In practice the patient is first seen by the doctor, and, on the basis of the history and examinations, a working hypothesis, or provisional diagnosis, is made, which in turn may be modified by test results (7). Near-to-patient testing should thus also be evaluated in terms of added information to that of the doctor s provisional diagnosis. The aim of the present study was to evaluate the efficacy of the non-specific laboratory infectious disease parameters commonly used in primary care; the erythrocyte sedimentation rate (ESR), leucocyte count (WBC), C-reactive protein (CRP), and a specific streptococcal immunological assay performed at the doctor s office, in diagnosing GAS throat infections in adults. These tests were also evaluated in conjunction with the doctor s clinical diagnosis. Material and methods The study took place at the emergency department in Troms0, a town of 50000 inhabitants in northern Norway. During the study period, October 1988 to May 1989, 40 different doctors treated the patients and 7 nurses performed the laboratory procedures. The doctors had between half a year and 20 years of clinical experience. Patients, 18 years or older, attending the ward between 4 and 9 p.m. with the chief complaint of a sore throat, were included. After consent was obtained, blood samples were drawn. Bacterial samples were taken by first rotating a sterile dacron-coated plastic swab and then a carbon impregnated swab, firmly over red, inflamed areas of the tonsils, uvula and pharynx. The carbon swab was immediately placed in a transport medium and processed within 24 hours at the Department of Microbiology, University Hospital of Tromso. It was inoculated on sheep-blood agar plates, incubated anaerobically at 37 C, and read after 24 and 48 hours. The dacron swab was ana- Does near-to-patient testing contribute 7 1 lysed at the clinic using the PathoDx test kit (Diagnostic Products Corporation, California, USA). This is an immune assay with antigen-binding antibodies linked to latex particles, where the antigen is visualised by agglutination. The one hour sedimentation rate was measured according to the modified Westergren method, using closed vacuum tubes. The leucocyte counts were performed on a Linson CX 320 automatic cell counter. Serum samples were frozen and sent to the Clinical laboratory at Hammerfest hospital where CRP levels were measured using a RA loo0 analyser and reagents from Orion Diagnostics, Finland. AST and ADNase B serological tests were done at the Department of Microbiology, University Hospital of Tromso. The duration of untreated illness was registered as the time from the first symptoms of throat infection to the consultation. After the clinical evaluation, but before getting the laboratory results, the doctor indicated on a ten centimetre visual analogue scale (VAS) his or her degree of certainty about the clinical diagnosis. The scale ranged from; <<Feel quite sure this is not a GAS infection,, to; <<Feel quite sure this is a GAS infection.>> The VAS score was used as a probability estimate in the analysis. The doctors also indicated dichotomously their clinical diagnosis of a GAS infection or not. All patients were asked to return after four weeks for follow-up serology. The presence of GAS throat infection, our reference standard, was defined as those patients presenting with a chief complaint of sore throat and having a positive GAS culture or a four-fold or more increase or fall (one patient) of AST or ADNase B serum levels during the four-week observation period. Sensitivity, specificity, predictive values, and likelihood ratios (LR) were calculated for different thresholds of the tests (8). Receiver Operating Characteristics (ROC) curves were used in the presentation of the results (9). Univariable analyses were performed for each test evaluated. Discrete variables were tested with chi-square statistics and continuous variables with Student s t-test. Stepwise, forward multivariate logistic regression analysis was performed to determine the factors having an independent effect on the diagnosis of streptococcal infection. The presence or absence of GAS infection was used as the dependent vari- Scand J Prim Heoltli Care 1994; 12

72 P. Hjortdalil, H. Melbye able and the doctor's probability estimate and the different test levels as independent variables. The weights achieved by the tests when added to the doctor's probability estimate are presented as chi-square values. P-values equal to or less than.05 were considered statistically significant. The study was approved by the regional committee on medical research ethics. Results Fourteen of the 188 patients fulfilling the inclusion criteria refused to participate, mainly due to reluctance to undergo blood sampling. Among the 174 patients included, 59 had a GAS throat infection, 49 identified by positive culture and 10 only by significant changes in serum titter values of AST or ADNase B. The age of the patients ranged from 18 to 71 years, mean 28. No significant differences were found in age or gender between patients with GAS and non-gas throat infection. The mean duration of untreated illness before seeking medical care was six days. The mean symptom duration in patients with GAS infection was four days, while in patients with non-gas infection it was seven days (p<.oi). Among the 59 patients RcrJence of GAS loo 1 with GAS infection, 41 (69%) came to the doctor within the first three days of illness, 49 (83%) within the first five days. Clinical evairiation The doctors' certainty of aetiology in the 171 cases when registered, related to the incidence of GAS infections, is shown in Figure 1. When the doctor used the lower quartile of the VAS, indicating a fair degree of clinical certainty that the patient did not have a GAS infection, he was right in 53 of 67 cases (true negative rate 79%). Using the upper quartile of the VAS, indicating a fair degree of clinical certainty that the patient did have a GAS infection, he was right in 21 of 33 cases (true positive rate 64%). By clinical evaluation alone, the doctors correctly identified 38 of the 59 GAS infections (sensitivity.64) while 33 of the 115 non-gas were wrongly judged as GAS (p<o.oool) (Table I). Laboratory tests The level of the ESR related to the duration of illness is shown in Figure 2. The mean level in the non-gas group was 15 mm and in the GAS group 20 mm (p=.65). The WBC related to duration of illness is shown in Figure 3. The mean number of leucocytes in non-gas infected patients was 8.8 x 109/L and in GAS infected patients 12.1 x 109/L (p c.01). The CRP response is ESR mm "1... o'. ESR d A S o J - * '..,.,,,,,., 0.3.6-1.1 162.1 2.b.3.5 3.b4.5 465.5 5645 b61.5 7b-85 1.693 9610 Quiu a mn it is not GAS Quiu ccnain it is GAS The doaor's clinical amnry abui acoolog) =marked on the VAS Fig. 1, Prevalence of group A streptococcal infection (GAS) related to the doctors' clinical certainty.about the aetiology. Figures in brackets are number of observations at each interval. N=171. The dotted line indicates agreement between the doctors' clinical certainty and the prevalence of GAS. S w d J Prim Heulrh Caw 1993: 12 0 1 2 3 4 5 6 7 2 8 Days since Tun symptoms Num&.3or&mvsuca.. GA-~AS. VIJ mi 1% M 3n M M 614 krg. 2. 'l'he level of the sedimentation rate (hsk) related to the duration of untreated throat infection. N=174.

Does near-to-patient testing contribute 73 Table I. Diagnostic value of tests performed to differentiate group A streptococci (GAS) from other causes of acute throat infection. Among the 174 patients in whom CRP, leucocytes and ESR were measured, 59 had GAS infection, while among the 164 patients in whom the PathoDx was performed, 57 had similar aetiology Overall clinical evaluation; CRP; mg/l Leucocytes: x 10y/L ESR; mm PathoDx; >20 >40 >60 >I0 >I2 714 >14 720 >40 shown in Figure 4. The mean CRP concentration in the non-gas patients was 19.5 mg/l and in the GAS patients 50.4 mg/l (p<.ooi). Among the 164 patients tested with the immunological assay, 56 belonged to the GAS group. Forty of these were correctly identified by PathoDx (sensitivity.71), while 100 of the 107 non-gas patients were thus identified (specificity.93) (Table I). Sensitivity Specificity Likelihood Positive ratio predictive value 0.64 0.71 2.2 0.52 0.73 0.47 0.32 0.7 1 0.47 0.29 0.65 0.47 0.05 0.7 1 0.67 0.82 0.91 0.77 0.84 0.90 0.68 0.78 0.94 0.93 2.2 2.5 3.4 3.0 3.0 3.0 2.0 1.o 0.6 8.7 0.53 0.56 0.63 0.61 0.61 0.6 I 0.49 0.5 1 0.23 0.82 The sensitivity, specificity, likelihood ratio, and positive predictive value for different threshold levels of CRP, ESR, and WBC are given in Table I and Figure 5. Clinical evaluation and laboratory tests The significance of each test added separately to the clinical probability estimate, using the logistic (3cp: m@ 40 71 Fig. 3. The level of leucocytes (WBC) related to the duration of untreated throat infection. N= 174. Fig. 4. The level of C-reactive protein (CRP) related to the duration of untreated throat infection. N=174. Srnnd J Prim Healrh Care 1994; I2

~~~~ 74 P. Hjortdahl. H. Melbye Table II. The significance of adding laboratory tests to the doctor s probability estimate of streptococcal throat infections in 174 adult patients with a sore throat, calculated by logistic regression. PathoDx was entered as a dichotomous variable, the other tests as continuous variables Specificity 0.9 0.7 0.5 0.3 0.1 Chi-square Probability ESR 3.2 0.07 CRP 14.3 >.0005 WBC count 20.2 >.OOo I PathoDx 32.1 >.OOo 1 regression analysis, is shown in Table 11. The leucocyte count, CRP and PathoDx each contributed significantly to the differential diagnosis, with PathoDx yielding the greatest independent value. Discussion Microbiological identification of GAS is the reference standard most often used when evaluating the aetiology of sore throat. This method has its own potential errors, however, since it sometimes misses streptococcal infections (2). In the present study, 10 patients with a negative culture, but with a four-fold. or greater change in AST or ADNase B serum titre levels were included in the reference standard. This increased the prevalence of GAS from 28% to 34% in the study population. Somewhat more patients with GAS were therefore included in this study than is otherwise common. It is our opinion, however, that the validity of the reference standard has thus been improved. Clinical elvaluation When the doctors felt clinically sure that the patient did not have a streptococcal infection, using the lower quartile of the VAS in Figure I, they tended to underestimate the prevalence of GAS, while overestimating it in the rest of the cases. This is in line with the study of Poses et a1 (lo), though the degree of overestimation is less in the present study. This is probably due to the different prevalence of GAS in the two study populations. 5% in that of Poses et a1 and 34% in ours. Scmd J Prim Healtli Cure 1994: 12 Fig. 5. ROC curves showing the diagnostic properties of PathoDx, C-reactive protein (CRP) at thresholds 10, 20,40, 60, 80 and 100 mg/l, erythrocyte sedimentation rate (ESR) at thresholds 20, 40, 60 mm and leucocyte count (WBC) at thresholds 7, 8, 9, 10, 11, 12, 13, 14 x 109/L, in the clinical diagnosis of streptococcal throat infections in 174 adult patients with a sore throat. The overall clinical evaluation, with sensitivity.64 and specificity.71, was better than that found in most other studies (11,12). Still, one out of three clinical diagnoses were erroneous, and, as demonstrated in Figure 1, a substantial overestimation of streptococci was present by clinical evaluation alone, with the potentially unnecessary use of antibiotics. Sedimentation rate The discriminatory ability of the sedimentation rate between GAS and non-gas infections was not statistically significant, and the ESR levels during the course of untreated GAS and non-gas infections overlapped (Figure 2). Evaluating sensitivity and specificity of the ESR at different thresholds by the ROC curves (Figure 5), it is evident that the ESR is the least useful of the tests thus appraised. It yields little or no clinically useful information in adults with acute throat infections, and is best left out of the diagnostic evaluation of these patients.

Does near-to-patient testing contribute 75 Leucocyte count A rapid rise in the leukocytes takes place during the first 24 hours of throat infection, with a subsequent fall during the next few days (Figure 3). There was a significant difference in the WBC count between the two patient groups. From Table I and Figure 3 it appears that a leucocyte count of 10 x 109/L is a clinically useful threshold in this differential diagnosis. Among patients coming within the first 24 hours of illness, however, the mean WBC value was high also in the non-gas group. A majority of these patients can be expected to have viral infections (13). With viral infections a rapid leucocytosis with an ensuing relative leucopenia has been reported (14). In addition, the more severely sick patients can be expected to seek medical help more quickly, also in the non-gas group. C-reactive protein The serum CRP level in healthy adults is low, usually 5 mg/l or less (15). A rapid and marked increase was seen within the first day of streptococcal infection, while lower mean values were found at longer durations of untreated illness (Figure 4). In the non-gas patients, a different CRP response pattern is evident with a slow and moderate increase in mean values during the first four days. followed by a fairly rapid decline in mean values toward normal levels within the next few days of untreated illness. If the sore throat has lasted less than a week, a CRP threshold of 40 mg/l appears to have reasonable discriminatory properties, while if the patient has been sick for a week or longer, 20 mg/l is more appropriate. The significantly shorter duration of illness in the GAS-group than in the non-gas patients has contributed to the differences in mean values for both CRP and WBC. CRP and WBC have similar test characteristics and both yield clinically significant information (Figure 5). The leucocyte count in the present study was done with an automated cell counter. Manual counting is less accurate and reliable (16), which may limit the utility of this test in general practice. CRP has until recently been of little practical value as an infectious disease parameter in primary care. This is changing, partly at least due to the recent development of semi-quantitative CRP immunometric test kits, which are fairly accurate and rapid and which can be performed without auxiliary instrumentation (17). Immunological testing The direct immunological identification of streptococci showed a sensitivity of.71, which is lower than has been demonstrated in previous studies (18, 19). Including patients with a negative culture, but significant serological changes in the reference standard, has contributed to the low sensitivity, since the sensitivity of PathoDx in these patients was only.40. Furthermore, latex agglutination tests that were used in this study may be difficult to interpret at low antigen levels. The sensitivity demonstrated may thus be related to limited experience in performing and interpreting the test by the laboratory personnel. The test, however, was more accurate than clinical judgement alone, and had by far the best characteristics of the tests evaluated (Table I). Technical improvements that have recently been introduced, such as immune filtration, washing procedures, and quantitative colour reactions, may further increase the accuracy of the immunological method (17). Clinical implications Our findings indicate that the diagnosis of streptococcal pharyngitis can best be done on the basis of the doctor s clinical evaluation. If he or she is uncertain of the clinical diagnosis, most additional help may be obtained from an immunological rapid test. If this procedure is not available, significant differential diagnostic help may be obtained from evaluating the CRP level or, if performed on automatic cell counter, a leucocyte count. In this clinical situation, the ESR is of little or no diagnostic help. References 1. Lundberg C, Nord C-E. Streptococcal throat infections: still a complex clinical problem. Scand J Infect Dis 1988; suppl. 57: 7-1 1. 2. Kellogg JA, Manzella JP. Detection of group A streptococci in the laboratory or physician s office. JAMA 1986; 255: 2638-42. 3. Cochi SL, Fraser DW, Hightower AW, Facklam RR, Broome CV. Diagnosis and treatment of streptococcal pharyngitis: survey of U.S. medical practi- Scmd J Prim Health Care 1994; I2

76 P. Hjortdahl, H. Melbye tioners. In: Schulman ST. ed. Pharyngitis - Management in an era of declining rheumatic fever. New York: Praeger Press: 1984: 73-94. 4. Hilton S. Near patient testing in general practice: a review. Br J Gen Pract 1990 40: 32-6. 5. Broughton P. Laboratory medicine in primary health care. Br J Gen Pract 1990; 40: 2-3. 6. Hlatky MA, Pryor DB, Hmll FE, Califf RM, Mark DB, Rosati RA. Factors affecting sensitivity and specificity of exercise electrocardiography. Multivariable analysis. Am J Med 1984; 77: 64-7 1. 7. Hull FM. The consultation process. In: Sheldon M, Brooke J, Rector A. ed. Decision-making in general practice. London. The MacMillan Press LTD: 1985: 13-26. 8. Radack KL, Rouan G, Hedges J. The likelihood ratio. Arch Pathol Lab Med 1986 110 689-93. 9. Beck JR, Shultz EK. The use of Receiver Operating Characteristic (ROC) curves in test performance evaluation. Arch Pathol Lab Med 1986 110 13-20. 10. Poses RM, Randall DC, Centor RM. Evaluating physicians probabilistic judgments. Med Decis Making 1988; 8: 233-40. 11. Hjortdahl P, Laerum E, Mowinckel P. Clinical assessment of pharyngitis in general practice. Scand J Prim Health Care 1988; 6: 219-23. 12. Poses RM, Cebul RD, Collins M, Fager SS. The accuracy of experienced physicians probability estimates for patients with sore throat. JAMA 1985; 254: 925-9. 13. Glezen WP, Clyde WA Jr, Senior RJ, Sheaffer CI. Denny FW. Group A streptococci, mycoplasmas, and viruses associated with acute pharyngitis. JAMA 1967; 202: 455-60. 14. Douglas RG. Influenza in man. In: Kilboume ED. ed. The influenza viruses and influenza. New York. Academic Press: 1975: 39543. 15. Morley J, Kushner I. Serum C-reactive protein levels in disease. Ann NY Acad Sci 1982; 389: 406-1 8. 16. Landaas S, Juell A. The quality of laboratory analyses in primary health care. Scand J Prim Health Care 1986; 4: 169-73. 17. Hjortdahl P, Landaas S, Urdal P, Steinbakk M, Fuglerud P, Nygaard B. C-reactive protein: a new rapid assay for managing infectious disease in primary health care. Scand J Prim Health Care 1991; 9: 3-10. 18. Hjortdahl P, Laerum E, Gaustad P. How reliable and useful is the latex agglutination test in diagnosing streptococcal throat infection in general practice? Scand J Prim Health Care 1987; 5: 1514. 19. Hoffmann S. An algorithm for a selective use of throat swabs in the diagnosis of group A Streptococcal pharyngo-tonsillitis in general practice. Scand J Prim Health Care 1992; 10295-300. Received November 1992 Accepted November 1993 Scand J Prim Heulth Cure 1994: 12