Infectious Mononucleosis in an Outpatient Population: Diagnostic Utility of 2 Automated Hematology Analyzers and the Sensitivity and Specificity of Hoagland's Criteria in Heterophile-Positive Patients

The classic criteria for diagnosing infectious mononucleosis developed by Hoagland include a lymphocytosis comprising ≥50% of the white blood cell (WBC) differential with atypical lymphocytes accounting for ≥10% of the total WBC count.1–3 Previous studies have questioned the percentage of patients with serologic evidence of infectious mononucleosis who actually fulfill Hoagland's criteria.4,5 Modern hematology analyzers are capable of flagging atypical lymphocytes as well as other qualitative and quantitative abnormalities present in the hematology profile.

In an earlier study of the Coulter S-Plus IV, a combination of flags correctly identified 59 of 60 cases of infectious mononucleosis as requiring further review.6 However, to our knowledge no recent studies have focused on the diagnostic capability of newer analyzers to flag atypical lymphocytes or suggest a diagnosis of infectious mononucleosis. In addition, the sensitivities and specificities of individual analyzer flags have not been examined in this regard. Accordingly, the purposes of the current study were as follows: to determine the sensitivity and specificity of 2 modern hematology analyzers in flagging heterophile-positive patients; to determine the correlation between a instrument-generated lymphocyte count and a lymphocyte count performed by a skilled technologist analyzing specimens from heterophile-positive patients; to determine if heterophile-positive, instrument-flagged specimens contain a larger number or a different spectrum of atypical lymphocytes; to document the overall sensitivity and specificity of Hoagland's morphologic criteria in identifying heterophile-positive patients in an outpatient population with a clinical diagnosis of infectious mononucleosis; and, finally, to examine if there are individual morphologic features that might represent sensitive or specific findings to aid in the diagnosis of suspected infectious mononucleosis.

Peripheral blood samples were collected from 181 patients with clinical diagnoses of mononucleosis who subsequently tested positive for the heterophile antibody using the Monosticon Dri-Dot test (Organon-Technika, Scarboro, Ontario, Canada). The matching control group included 181 patients who also had presumptive clinical diagnoses of mononucleosis, but who tested negative for the heterophile antibody using the Monosticon Dri-Dot test. Blood was drawn using disposable multisample needles (Becton Dickinson Vacutainer Systems, Franklin Lakes, NJ) and commercially prepared tubes (Becton Dickinson Vacutainer Systems, Rutherford, NJ). Blood samples were analyzed within 1 to 3 hours of collection on either a Coulter STKS (Coulter Electronics, Inc, Hialeah, Fla) or a Sysmex NE-8000 (TOA Medical Electronics Company, Ltd, Kobe, Japan) analyzer, each calibrated according to the manufacturers' specifications. The software version used with the NE-8000 was 6/20, whereas version 2B was used with the STKS. Histogram patterns were plotted for all samples, and any operator alert flags were noted.

A corresponding Wright-Giemsa–stained blood smear prepared at the time of blood collection was subsequently examined by a single skilled technologist who had no prior knowledge of the analyzer flagging results. Each blood film had an initial 200-cell WBC differential performed. A separate 200-cell lymphocyte differential was subsequently performed by examining 200 consecutive lymphocytes on each slide utilizing a microscope eye piece equipped with a micrometer. For the purposes of the current study, lymphocytes were considered atypical or reactive if they showed the characteristic morphologic changes described by Downey and others or were otherwise >15 μm in diameter.2,7–9 

A specific morphologic search was made on each slide for the presence of smudge cells or cloverleaf lymphocyte nuclei, because these morphologic changes have been described previously as possibly aids in the diagnosis of infectious mononucleosis.3,8 The percentage of cloverleaf nuclei was tabulated, and smudge cells were graded as present or absent.

Statistical analysis was used to compare results obtained from the heterophile-positive and heterophile-negative populations. Items that were compared included patient demographics and the distribution of hemoglobin, WBC counts, and platelet counts. Values for the automated instrument-generated WBC differential versus the 200-cell microscopic manual differential were compared for the heterophile-positive patients. The overall flagging profile and the distribution of individual flags for each analyzer were noted. The sensitivity and specificity values of the Sysmex and Coulter analyzers in flagging heterophile-positive patients with either the variant lymphocyte or blast flags were determined. A comparison was also made of the frequency distribution of the various morphologic types of atypical lymphocytes expressed as a percentage of the total lymphocyte count in the heterophile-positive and -negative patient populations. Logistic regression analysis was performed using the blast flag and the variant lymphocyte flag as dependent variables, looking at each of the following factors as independent variables: the presence of atypical lymphocytes expressed as a percentage of the total WBC count; the presence or absence of individual morphologic types of atypical lymphocytes, including Downey types I, II, and III; and the presence of atypical lymphocytes >15 μm in diameter.

The number of patients who met both or either of Hoagland's criteria for the presumed diagnosis of infectious mononucleosis was calculated, as were the sensitivity and specificity values of the criteria taken individually or together. The number of patients in the heterophile-positive and -negative groups who had an atypical lymphocytosis of ≥20% and ≥40%, respectively (expressed as a percentage of the total WBC), was tabulated. For patients in the heterophile-positive group who did not satisfy either of Hoagland's criteria, the presence or absence of a normal WBC count, an absolute lymphocytosis, or an atypical lymphocyte count ≥12% of the total lymphocyte count was determined.

Finally, a comparison was made of the sensitivity and specificity of smudge cells or cloverleaf lymphocyte nuclei in predicting heterophile-positive infectious mononucleosis.

The heterophile-positive population included 79 males and 102 females, ranging in age from 6 to 61 years, with a mean age of 21 years. The control population included 71 males and 110 females, ranging in age from 4 to 78 years, with a mean age of 22 years.

Quantitative data for the heterophile-positive and heterophile-negative groups are shown in Table 1. When comparing standard hematologic parameters between the 2 groups, statistical testing showed a significant difference for the total WBC count and the platelet count.

The frequency of individual quantitative abnormalities in the heterophile-positive and -negative groups is shown in Table 2. Lymphocytosis was the most common abnormality detected, followed by leukocytosis, monocytosis, thrombocytopenia, anemia, and leukopenia. Statistically significant differences were noted between the 2 groups for leukocytosis, lymphocytosis, thrombocytopenia, and the category of any quantitative change.

Comparison of the Coulter and Sysmex automated 5-part differentials with the 200-cell microscopic leukocyte differential in the heterophile-positive group of patients showed statistically significant differences for the monocyte and basophil counts with the Coulter analyzer and with the monocyte count for the Sysmex analyzer; the analyzer-generated monocyte counts were consistently higher.

Qualitative instrument flagging results for both the Coulter and Sysmex analyzers are listed in Table 3. The flags commonly associated with the presence of heterophile-positive status were the blast flag and the variant lymphocyte flag. In the Coulter group of 105 heterophile-positive patients, the variant lymphocyte flag was present in 76 (72.4%) of 105 patients, the blast flag was present in 43 patients (41%), both flags were present in 42 patients (40%), and the left shift/immature granulocyte flag was present in 3 patients (2.9%). The presence of each of these categories of flags was statistically significant when comparing the heterophile-positive group with the heterophile-negative group. For the Coulter analyzer, the sensitivity and specificity values in predicting positive heterophile status for the blast flag were 41% and 97.1%, respectively; for the variant lymphocyte flag, 72.4% and 79.1%, respectively; and for both flags, 40% and 98.1%, respectively.

In the heterophile-positive group of 76 patient samples run on the Sysmex analyzer, the variant lymphocyte flag was positive in 12 (15.8%), and the blast flag and the left shift/immature granulocyte flag were each positive in 33 (43.4%). Both the variant lymphocyte flag and the blast flag were positive in 8 patient samples (10.5%). The WBC differential voted out in all 33 patients who had a positive blast flag and left shift/immature granulocyte flag, as this represents a function of the individual algorithms of this analyzer. The presence of either the blast flag or the left shift/immature granulocyte flag was statistically significant when comparing the heterophile-positive and heterophile-negative groups. The sensitivity and specificity values for the blast flag in predicting heterophile-positive status with the Sysmex analyzer were 43.4% and 88.6%, respectively; for the variant lymphocyte flag, 15.8% and 90.8%, respectively; and for both flags, 10.5% and 96%, respectively. Considering the entire heterophile-positive patient population for the presence of either qualitative or quantitative flags, 156 of 181 patients manifested an instrument-generated flag, producing an overall flagging rate of 86.2. The corresponding false-negative rate was 13.8%. However, of the 25 patients who had neither a qualitative nor a quantitative flag registered by an analyzer, 7 had an atypical lymphocytosis making up ≥10% of the overall WBC count.

A comparison of the percentage of atypical lymphocytes, as well as the various morphologic types of atypical lymphocytes as enumerated in the special 200-cell lymphocyte differential, is provided in Table 4 for both heterophile-positive and heterophile-negative patients. The mean percentage of atypical lymphocytes was 40.9% in the heterophile-positive group versus 9.15% in the heterophile-negative group. In the heterophile-positive patients, Downey type II lymphocytes were most common (range, 0% to 80%; mean, 26.1%). Testing showed that there was a statistically significant difference between the heterophile-positive and heterophile-negative groups of patients for the percentages of each morphologic type of atypical lymphocyte.

Results of logistic regression analysis relating the presence of blast and atypical lymphocyte flags taken as dependent variables compared to various lymphocyte-related independent variables are presented in Table 5. Blast flag positivity identified by the Coulter analyzer showed an association with the presence of Downey type I, II, and III lymphocytes, as well as with the overall percentage of atypical lymphocytes. With the Coulter analyzer, atypical lymphocyte flag positivity was associated with the presence of Downey type I and II lymphocytes and with the overall percentage of atypical lymphocytes. Blast flag positivity identified by the Sysmex analyzer demonstrated an association with the presence of lymphocytes >15 μm and with the overall percentage of atypical lymphocytes, whereas atypical lymphocyte flag positivity was associated with the presence of Downey type I lymphocytes and lymphocytes >15 μm.

After applying Hoagland's criteria to the entire heterophile-positive population of 181 patients, we found that 120 patients (66.3%) had a lymphocytosis of ≥50% of the WBC differential, and 135 (74.6%) had an atypical lymphocytosis of ≥10% of the total WBC count. In the heterophile-negative population, 28 patients (15.5%) had a lymphocytosis of ≥50%, and 14 (7.7%) had atypical lymphocytosis of ≥10%. Thus, a lymphocytosis ≥50% had a sensitivity of 66.3% with a specificity of 84.5% for diagnosing heterophile positivity, whereas an atypical lymphocytosis ≥10% had a sensitivity of 74.6% and a specificity of 92.3%. If we required both criteria to be positive, the sensitivity was 61.3% and the specificity was 95%. In the heterophile-positive group, 56.4% and 25.4% of patients had an atypical lymphocytosis >20% and >40% of the total WBC, respectively, while the figures were 2.2% and 0%, respectively, for the heterophile-negative patients. In total, 36 (19.9%) patients in the heterophile-positive group demonstrated neither of Hoagland's criteria. Within this group, 27 (75%) had a normal WBC count, 33 (91.7%) did not have an absolute lymphocytosis, and 16 (44.4%) had an atypical lymphocyte count ≤12%, expressed as a percentage of the total lymphocyte count.

Cloverleaf nuclei were detected in 28 patients (15.5%) in the heterophile-positive group as part of the special 200-cell lymphocyte differential, as opposed to 3 patients with cloverleaf nuclei in the control group (Figure 1). The percentage of lymphocytes with cloverleaf nuclei was small, ranging from 1% to 4%, with a median of 1%. The sensitivity of the presence of cloverleaf nuclei for predicting heterophile-positive status was 15.5%, and the specificity was 98.3%.

Fifty-five (30.4%) patients in the heterophile-positive group demonstrated smudge cells, as opposed to 7 (3.9%) in the control group (Figure 2). The sensitivity of the presence of smudge cells for predicting heterophile-positive status was 30.4%, and the specificity was 96.1%.

In this series of 181 heterophile-positive patients, patient demographics and quantitative data were similar to findings reported previously.4,6,7 An absolute lymphocytosis was the most common quantitative abnormality in the heterophile-positive group; an absolute leukocytosis was present in 38.1% of patients overall, and a mild thrombocytopenia was documented in 13.8%.10,11 Of course, the hematologic picture would be expected to vary among any particular grouping of heterophile-positive individuals, since any series of mononucleosis patients would include individuals tested at differing times in relation to the onset of infection. Typically, the lymphocytosis of infectious mononucleosis peaks during the second or third week of illness, whereas patients in the early stage of illness may manifest a relative or absolute neutropenia.2,3 

We found good correlation with both of the analyzers tested for lymphocytes, monocytes, and granulocytes. The increased variation noted between machine and manual differentials in relation to the monocyte count for both analyzers and the basophil count for the Coulter analyzer has been reported previously.12 Although the automated monocyte count for both analyzers was consistently higher than the manual monocyte count, the automated absolute lymphocyte count was consistently lower. This finding has also been documented by other authors and suggests that some of the larger reactive lymphocytes may have overlapping features in relation to analyzer algorithms and, as a result, are counted as monocytes.6 

Although our study did not directly compare the flagging ability of the individual analyzers by using the same samples, it is important to note that when the results of all quantitative and qualitative flags were combined, 156 patients (86.2%) in the heterophile-positive group were identified as meriting further evaluation. Since this investigation was performed using specimens from patients who were referred with a suspicion of infectious mononucleosis, we could not estimate the positive predictive value of 1 or a combination of flags being positive in the general population. However, overall, these results are in keeping with an earlier investigation in which a combination of all analyzer-detected abnormalities would have resulted in a 98% detection rate of suspected mononucleosis patients as requiring additional review.6 The presence of the blast flag appeared to have approximately equivalent sensitivity and specificity for predicting heterophile-positive status in the case of both analyzers. However, the Coulter analyzer variant lymphocyte flag appeared to be significantly more sensitive, although less specific, than the variant lymphocyte flag of the Sysmex analyzer. This difference in flagging sensitivity between the 2 instruments for the variant lymphocyte flag has also been documented by other authors.13 Of the 25 patients who had no qualitative or quantitative analyzer flag, 7 had an atypical lymphocyte count comprising ≥10% of the total WBC count. This suggests that there is room for further improvements in the performance of these flags in detecting abnormal samples. Other authors have documented similar problems with the overall flagging of atypical lymphocytes, although the positive predictive value significantly improved for specimens with >10% atypical lymphocytes.14 

While some authors consider the presence of >2% atypical lymphocytes expressed as a percentage of the total lymphocyte count to be an abnormal finding in healthy individuals, others have reported values of up to 29% as normal.15,16 Probably the most reliable data come from Wood and Frenkel,15 who found that a value of up to ≤12% represented a normal finding. In our study, the mean percentage of atypical lymphocytes expressed as a percentage of the total lymphocyte count in the heterophile-negative patients was 9.15%, suggesting normalcy if this range were to be used. Of the approximately 20% of heterophile-positive patients who satisfied neither of the Hoagland's criteria, 44.4% also had an atypical lymphocytosis if this reference range was used. While the presence of 20% or more atypical lymphocytes expressed as a percentage of the total WBC count has been viewed as a useful criterion for distinguishing infectious mononucleosis from the lymphocytosis seen with cytomegalovirus, acquired toxoplasmosis, viral hepatitis, herpes infections, roseola, rubella, adenovirus infection, and influenza B, the presence of 40% atypical lymphocytes has been considered pathognomonic.17 In considering these criteria, 2.2% and 0% of heterophile-negative patients had greater than 20% and 40% atypical lymphocytes, respectively, whereas the figures were 56.4% and 25.4% for the heterophile-positive individuals. Considerable heterogenicity among types of atypical lymphocytes present has also been considered a diagnostic criteria for infectious mononucleosis as demonstrated in our study.2,3 Typically, the Downey type II lymphocyte has been described as the most common variant noted in infectious mononucleosis, and our results are consistent with this conclusion.16,18 The percentage of Downey type III lymphocytes noted in the our study was of interest because this variant is the most likely to be confused with a malignant process.18 Although few studies on the morphologic distribution of atypical lymphocytes in heterophile-positive patients have been published, in 1 series of patients with atypical lymphocytes from various causes, Downey type III cells comprised 9%.15 This figure is considerably higher than the mean of 3.09% noted in our study.

The logistic regression data relating the positivity of the blast and atypical lymphocyte flags to various other lymphocyte parameters are also of interest. The fact that we found no consistent patterns of correlation between the number of atypical lymphocytes or the distribution of the various morphologic types of atypical lymphocytes with the activation of the blast or variant lymphocyte flags in the heterophile-positive patients probably relates to the interpretive algorithms used by the individual analyzers.13,14 Thus, the separation of atypical lymphocytes into subtypes, although of academic interest, is not of diagnostic significance, nor is it warranted in routine clinical practice.

The utility of Hoagland's observations in predicting a diagnosis of infectious mononucleosis has been debated in the literature.4,5 Although some authors have questioned the usefulness of the individual criteria, the current study showed that both a lymphocytosis of ≥50% and the presence of ≥10% atypical lymphocytes represent useful findings, with sensitivities approaching 70% or greater and specificities in the range of 85% to 90%. However, if a requirement for the diagnosis of heterophile positivity was for both parameters to be positive, the sensitivity dropped into the 60% range, while the specificity rose to 95%. A slightly smaller study found a sensitivity of 39% with a specificity of 99% if the diagnostic requirement was for both criteria to be met.4 

In evaluating the validity of these data relating to sensitivity and specificity, several factors must be considered. First, the Monosticon Dri-Dot test has a false-negative rate of approximately 4.8%.19 Second, in some patients heterophile antibody production lags behind the appearance of a reactive lymphocytosis, such that retesting may yield a positive result after 1 to 2 weeks. We also know that the incidence of heterophile-negative infectious mononucleosis is approximately 10% overall.9,20 While heterophile-negative infectious mononucleosis is significantly more common in younger patients, by the age of 4 years the incidence is similar to that of young adult patients.9 The mean age of patients in the heterophile-negative control group in our study was 22 years, and no patients were younger than 4 years. However, it is important to remember that while patients in the control group were heterophile-negative with the Monisticon test, a proportion probably had positive Epstein-Barr virus serology compatible with an acute infection.20 On the other hand, our sensitivity and specificity figures for Hoagland's criteria represent a “real world” testing situation, since Epstein-Barr virus serology is not routinely performed when screening for suspected infectious mononucleosis. We plan to report on the incidence of the various viral illnesses that were detected in the subgroup of heterophile-negative control patients who also manifested a significant atypical lymphocytosis elsewhere.

The data relating to the presence of smudge cells or cloverleaf lymphocyte nuclei as potential aids in the diagnosis of infectious mononucleosis are interesting.8 Disintegrated lymphocytes designated as smear cells, smudge cells, or basket cells may also be seen in chronic lymphocytic leukemia and a variety of other reactive conditions, including whooping cough.3,18 Presumably, the smudging detected is related to changes in the lymphocyte membrane, possibly secondary to pathologic activation resulting in increased fragility. Lymphocytes with cloverleaf nuclei have also been described in adult T-cell leukemia, human immunodeficiency virus infection, and scrub typhus disease.21–23 When noted, either parameter would appear to have a high specificity for heterophile-positive status. Unfortunately, overall utility is limited by the relatively low sensitivity, which is in the range of 30% for the presence of smudge cells and 15.5% for the presence of cloverleaf nuclei. Thus, while the detection of either of these morphologic abnormalities may serve as an additional clue to the presence of heterophile-positive status, because of their relative infrequency, they cannot be relied upon to reliably establish the diagnosis.

Modern hematology analyzers are efficient in identifying heterophile-positive patients. Using a combination of all qualitative and quantitative flags, the 2 analyzers tested identified 156 (86.2%) of 181 heterophile-positive cases as meriting further review. This result corresponds to a reasonable false-negative rate of 13.8%. Hoagland's criteria proved valuable for suggesting a diagnosis of heterophile-positive status. The presence of either smudge cells or cloverleaf lymphocyte nuclei was verified as a specific but relatively insensitive clue to the diagnosis of heterophile-positive status.