The Reference Value of Erythrocyte Sedimentation Rate for Differential Diagnosis of Rheumatic Fever Among Bangladeshi Children

The aim of the present study is to determine the reference value of erythrocyte sedimentation rate for differential diagnosis of rheumatic fever in the National Center for Control of Rheumatic Fever and Heart Diseases, Dhaka, Bangladesh among patients with signs and symptoms which may be related to acute rheumatic fever. All medical records for the patients aged 5 to 20 years who attended the outpatient department of the hospital between July, 1994 and November, 1995 were reviewed. Fifty-three of 337 such patients had acute rheumatic fever defined by the updated Jones criteria. The performance of erythrocyte sedimentation rate test was evaluated by sensitivity, specificity, positive predictive value, and receiver operating characteristic curve. The findings of this study suggest that the lower limit for a positive test should be considered at 30 mm (Westergren 1 h) in this hospital. J Epidemiol, 1996 ; 6 : 109-113.

out2 .Therefore, it is essential to establish the reference range of erythrocyte sedimentation rate for the specific population 3).
The National Center for Control of Rheumatic Fever and Heart Diseases (hereinafter referred to as Rheumatic Fever Center) is a tertiary level referral hospital in Dhaka, Bangladesh.
It has been in operation since 1988.Many patients with signs and symptoms which may be related to acute RF are referred to the Rheumatic Fever Center.When the final decision is made whether the patient meets the Jones criteria for RF, erythrocyte sedimentation rate plays a relatively important role.To date, however, the reference value has not been established and varies among physicians of the Rheumatic Fever Center.The aim of the present study is to determine the reference value of erythrocyte sedimentation rate for differential diagnosis of RF in the Rheumatic Fever Center among patients with signs and symptoms which may be related to acute RF.

Subjects :
After approval of our study protocol by the Review Board of the Rheumatic Fever Center, all medical records for the patients aged 5 to 20 years who attended the outpatient department between July, 1994 and November, 1995 were reviewed.
Among these patients, those who presented with acute manifestation(s) defined by the presence of any of the signs and symptoms that might be related to rheumatic fever -carditis, arthritis, arthralgia, erythema marginatum, subcutaneous nodule, chorea, and fever -were identified.

Measurement of Erythrocyte Sedimentation Rate :
Erythrocyte sedimentation rate was measured by the method of Westergren (1 hour) and performed within two hours of venepuncture in room temperature, 18 to 251C, as recommended by the International Committee for Standardization in Haematology2).The quality of the test measurement was periodically checked by a staff physician.
Gold Standard for the diagnosis of Rheumatic Fever : Although the Jones criteria are not sufficiently sensitive and specific4), the patients who met the updated criteria1) were identified as "true" patients.Presence of two major criteria (carditis, polyarthritis, chorea, erythema marginatum, and subcutaneous nodule), or one major and two minor criteria (arthralgia, fever, positive C-reactive protein test, and prolonged P-R interval on electrocardiogram), indicated a high probability of acute RF if supported by the evidence of preceding group A beta-hemolytic streptococcal infection (high antistreptolysin 0 titer and/or positive throat culture for group A beta-hemolytic streptococci).To make a final diagnosis, all patient information were considered jointly by two staff physicians.Thus, 53 out of 337 children were confirmed to be "true" patients with acute RE Validation Analyses : The performance of erythrocyte sedimentation rate test was evaluated by sensitivity, specificity, and positive predictive value which were all expressed as proportions defined in the standard manner5).The computational formulas are mentioned at the footnote of the Table .The cut-off point that includes all "true" patients, i.e., the value for 100% sensitivity was defined as the reference value in spite of the fact that this implied lower specificity.Although it is obviously desirable to have a screening that is both highly sensitive and highly specific, a compromise has to be sought between the two.The receiver operating characteristic curve takes into account the sensitivity and specificity together, and indicates the best cut-off point5).Generally, it is at or near the "shoulder" of the curve , the intersection of the curve and the diagonal line drawn between upper-left and lower-right vertices, unless there are clinical reasons for minimizing either false negatives (1-sensitivity) or false positives (1-specificity).
Therefore, the erythrocyte sedimentation rate value at the "shoulder" of the receiver operating characteristic curve was defined as the second reference.
The frequency distribution of the values for erythrocyte sedimentation rate (x) was skewed to the right but found to follow approximately log-normal (log,ox) distribution.The antilogarithms of mean and `mean + 1 standard deviation', in logarithms, were obtained to get the corresponding actual erythrocyte sedimentation rate values.The latter was defined as the third or statistical reference, i.e., the value which was exceeded xwsby about 16% of observations.

RESULTS
The mean age of the subjects (47% boys) was 11.4 years, and 80% were 14 years old or younger.The mean age of the "true" patients with RF was 12 .6 years and that of the non-RF patients was 11.2 years.The proportion of boys was 66% in "true" patients and 43% in the rest .
The parameters indicating performance of erythrocyte sedi- *the clinical diagnosis of rheumatic fever defined by the updated Jones criteria , which incorporated positive C-reactive protein test as an evidence of acute phase response, was considered as the "gold standard" **the lower limit for a positive test , mm (Westergren 1 h) ***sensitivity = a/(a+c) , specificity = d/(b+d) and PV, (positive predictive value) = a/(a+b) ; where, a indicates number of rheumatic fever cases who had positive test ; b, number of non-rheumatic subjects who had positive test ; c , number of rheumatic fever cases who had negative test ; and d, number of non-rheumatic subjects who had negative test  mm, but at this cut-off point the specificity was fairly low (0.67).Above this cut-off point, the specificity increased at the expense of sensitivity.
Figure 1 shows the frequency distribution of subjects at different cut-off points of erythrocyte sedimentation rate.The estimated statistical reference value (antilogarithm of `mean + 1 standard deviation' in logarithm) of erythrocyte sedimentation rate was 60 mm.As shown in the Figure 2, the receiver operating characteristic curve indicated 60 mm at its "shoulder" (the second reference) which is exactly but coincidentally the same as the statistical reference.

DISCUSSION
The reference value for "normal" erythrocyte sedimentation rate should be established that includes 95% of the local healthy population 3).In this particular hospital-based population, each subject had some sort of clinical problems ranging from ill-defined to severe manifestations.Therefore, the reference value suggested here can not be generalized.
The erythrocyte sedimentation rate and the C-reactive protein are the best-known and most commonly used tests to measure presence and degree of systemic inflammatory process in patients with RE They are very sensitive but neither is specific for RE The C-reactive protein test is somewhat better since it is independent of anemia1).However, erythrocyte sedimentation rate remains the most popular and widely used test in developing countries like Bangladesh, because it is less expensive, easier to perform, and does not require electrical power supply and capital investment in equipment.
To keep both the false positives and false negatives at zero or minimum level, it is necessary to establish the reference value at which the erythrocyte sedimentation rate becomes 100% or nearly so for sensitivity and specificity.Unfortunately, this was not possible as it happens in case of other clinical data that take on a range of values 5).In these situations, the location of a cut-off point, the point on the continuum between normal and abnormal, is an arbitrary decisions.Thus, the decision must be made in the light of consequence of diagnostic errors6), and the money and time required for largescale screening.As the Rheumatic Fever Center is a tertiary level hospital, the false negative patients may encounter serious consequences due to the heart valve damage.On the other hand, the false positives could be minimized with the aid of other criteria.Therefore, the lower limit for a positive test should be considered at 30 mm for both boys and girls, although the false positive rate is fairly high for this cut-off point.We suppose that, assuming similar patient and laboratory conditions, this reference value could be used also in other tertiary level hospitals in Bangladesh.If we agree to give weight to both sensitivity and specificity and an additional weight to positive predictive value, at least for large-scale screening in tertiary level hospitals, an erythrocyte sedimentation rate test result of 60 mm or above may be considered as positive.We consider it important to have an increment, for this cut-off point, in the positive predictive value from 0.36 to 0.56, and a reduction in the false positive rate from 0.33 to 0.11 (Table, and Figure 2).Nonetheless, it reduces the sensitivity from 1 to 0.77.It is noteworthy that we observed a high predictive value of erythrocyte sedimentation rate due to the high prevalence7) of RF (53/337 = 0.16) because of the type of recruitment in a specially oriented RF hospital.
The frequency distribution of log erythrocyte sedimentation rate shows two apparent peaks (Figure 1).They may represent sampling variations or two groups of subjects : the left with illdefined or milder problems and the right with more severe problems.In this hospital, usually, more severe form of RF patients attend.They are supposed to have higher levels of erythrocyte sedimentation rate as compared to RF patients in general.To the contrary, patients have to pass through a referral chain to attend the Rheumatic Fever Center (being a tertiary level hospital).Therefore, none of our "true" patients are fresh cases of RF and they received some sort of medications before attending this hospital.These two mutually opposing factors might have balanced the resultant level of erythrocyte sedimentation rate, but we can not say what exactly happened.Nevertheless, we recommend the reference value for the Rheumatic Fever Center or other tertiary level hospitals only.In these hospitals, the background of the patients is assumed to be similar.
The erythrocyte sedimentation rate is influenced by anemia, age, sex, and corticosteroid drugs2).In this population 46% had anemia, although mild, defined by the World Health Organization criterion (hemoglobin <l1g/100 ml for children * 6 years of age, <12g for children 7-14 years, <13 g for boys * 15 years, <12 g for girls*15 years)8).An analysis for nonanemic subjects (54%) revealed slightly higher sensitivity and specificity for corresponding cut-off points (data not shown), except for the equal sensitivity at the cut-off point 30, as compared with the whole population.Among the boys, the performance of the test for corresponding cut-off points is fairly similar (Appendix).Separate results for girls are not presented because of the smaller number of "true" patients (18).We realize the drawback of under-represenation of girls in the "true" patients .However, it is not known whether there is any gender difference for occurrence of RF in Bangladeshi children.We observed almost similar results for the age group * 14 years (data not shown).This may be due to the fact that the effect of age on erythrocyte sedimentation rate becomes apparent after 203 (the upper age limit of our subjects).Unfortunately, we do not have information on corticosteroid medication.However, we assume that the number of patients who received corticosteroid drugs, if at all, is minimum.Separate analyses for initial (72%) and recurrent attack patients were not considered, because a single reference value, irrespective of recurrence status, is needed in the hospital.
In conclusion, sensitivity deserves priority to make a differential diagnosis of RF at this hospital, and hence the lower limit for a positive test should be considered at 30 mm.When both sensitivity and specificity are given equal weight, a test result of 60 mm or above should be considered as positive.

Figure 1
Figure 1.Histogram of the frequency distribution of erythrocyte sedimentation rate