SEARCH

SEARCH BY CITATION

Keywords:

  • erythropoietin;
  • erythrocytosis;
  • thrombo cythaemia;
  • haemoglobin;
  • thrombosis

Abstract

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. References

Summary.  Serum erythropoietin (Epo) values were estimated in samples from 125 patients with erythrocytosis to examine the specificity and sensitivity of reduced and raised values in the diagnosis of polycythaemia vera (PV) and secondary erythrocytosis (SE) respectively. Additionally, Epo values were estimated in samples from 49 patients with primary thrombocythaemia (PT) to determine whether Epo values were altered. We found high specificity (92%) and moderate sensitivity (64%) of low serum Epo values (below the reference range) in the diagnosis of PV, and also poor sensitivity (47%) of raised Epo values in the diagnosis of SE. Raised Epo values were not observed in PV patients with Hb > 14·0 g/dl and were only observed in one PV patient with a relatively low Hb recovering from a gastro-intestinal haemorrhage. Raised Epo values occurred in some patients with apparent erythrocytosis (AE) and idiopathic erythrocytosis (IE), mainly at normal (rather than raised) Hb values (< 16 g/dl). Low Epo values occurred in a few AE, IE and SE patients at higher Hb values (> 16 g/dl). Low Epo values were less specific for PV when the Hb was raised, while raised Epo values were less specific for SE when the Hb was not raised. Approximately one third of patients with PT had a low (below the reference range) Epo value, this being associated with a high normal Hb (> 14 g/dl, P < 0·001) and showing a trend towards association with absence of treatment. The high normal Hb values were in turn associated with an increased incidence of thrombotic events (P < 0·05). These findings could influence the future investigation and management of PT patients.

A low (below the reference range) serum erythropoietin (Epo) value is becoming increasingly recognized as a useful criterion in the diagnosis of polycythaemia vera (PV) (Pearson & Messinezy, 1996). The low value usually persists even when treatment has lowered the haematocrit and Hb (Birgegård & Wide, 1992). The sensitivity and specificity of this criterion are, however, not absolute. An examination of serum Epo values in other erythrocytoses and attempts to define factors which may affect these values are therefore relevant. Low serum Epo values are also known to occur in a number of patients with primary thrombocythaemia (Carneskog et al, 1998), in which the potential implications need examination.

This study reports serum Epo values in treated and untreated patients with PV or other erythrocytoses and also in patients with primary thrombocythaemia (PT). The relationship between serum Epo and Hb value at the time of Epo estimation was examined in these patients.

Patients and methods

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. References

One hundred and twenty-five unselected, treated and untreated adult patients (95 men; 30 women) attending routine haematology clinics in the last 3 years and known to have an erythrocytosis at diagnosis [packed cell volume (PCV) > 0·51 in men, > 0·48 in women] were divided by standard diagnostic criteria (Pearson, 1998) into the following groups: 42 patients with PV (the Epo value was not used in the diagnosis of these patients), 17 patients with secondary erythrocytosis (SE) (absolute increase in red cell mass due to an underlying pathology), 31 patients with idiopathic erythrocytosis (IE) (absolute increase in red cell mass and no known cause of secondary erythrocytosis or diagnostic features of PV) and 35 patients with apparent erythrocytosis (AE) (raised PCV but red cell mass within their normal range). A further 49 patients (16 men; 33 women) with normal PCV, persistently raised platelet count and no cause of secondary thrombocytosis or evidence of other myeloproliferative disorder were diagnosed with PT and were also studied.

Of the total 174 patients, 97 had only one blood sample examined, while 52 had two samples (taken randomly at separate clinic visits) and the remaining 25 patients had 3–6 samples taken. The majority of samples were from patients currently receiving treatment for their erythrocytosis or PT (181 samples), while the remainder were from untreated patients (121 samples).

A total of 302 samples from the 174 patients were analysed. Hb and serum Epo levels were estimated in each sample. The diagnostic sensitivity and specificity of serum Epo were derived from each patient's initial blood sample only.

The Hb values were assigned to five intervals (10·1–12·0, 12·1–14·0, 14·1–16·0, 16·1–18·0, > 18·0 g/dl) to allow examination of the Epo data corresponding to each Hb interval. If a patient was found to have more than one Epo estimation within a given Hb interval, only the chronologically first Epo value was retained for initial analysis and charting of results. Thus, results from a total of 233 samples were charted according to the individual Hb value and corresponding serum Epo estimation from each sample.

All the later Epo results beyond the chronologically first were analysed subsequently in order to examine the reproducibility of Epo values obtained for individual patients within a given Hb interval.

Serum erythropoietin estimation

Serum Epo estimations were made on blood samples taken between 09:00 and 12:00 h by enzyme-linked immunosorbent assay (ELISA) using Quantikine IVD Erythropoietin ELISA, R & D Systems Inc., Abingdon, Oxon, UK (the within-assay precision at a level of 5·8 mIU/ml being quoted as 4·95%). Our reference range was obtained using single blood samples from 36 normal members of laboratory staff. This was used to subdivide Epo values into low, normal or raised categories depending on whether each value was below, within or above this reference range.

Results

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. References

The reference range for serum Epo (see above) was 2·9–15·1 mIU/ml.

Sensitivity and specificity of low Epo in the diagnosis of PV

Forty-two patients with PV were studied. The first blood sample taken from each patient for this study showed a low Epo (< 2·9 mIU/ml) in 27 patients (diagnostic sensitivity 64%). The sensitivity was not influenced by whether patients had received treatment or not. The diagnostic specificity of a low Epo in the diagnosis of PV versus the other erythrocytoses (SE, IE, AE) was 92%.

Sensitivity and specificity of raised Epo in diagnosis of SE

Seventeen patients diagnosed as SE were studied (they comprised six patients with hypoxaemic lung disease, four with hypoxaemic congenital heart disease, three with high oxygen affinity Hb, two with sleep apnoea syndrome and two with renal disease). The diagnostic sensitivity of a raised Epo value (> 15·1 mIU/ml) in SE (eight patients) was only 47%. The diagnostic specificity of a raised Epo in the diagnosis of SE versus PV was high at 98%. Some raised Epo values occurred in AE and IE and were not as uncommon as in PV (see below).

Epo variation with Hb values in the erythrocytoses

The incidence of low, normal and raised Epo values at different Hb intervals is shown in Fig 1 for PV, Fig 2 for AE, Fig 3 for IE and Fig 4 for SE patients.

image

Figure 1. Serum erythropoietin (Epo) values and Hb in polycythaemia vera (PV).

Download figure to PowerPoint

image

Figure 2. Serum erythropoietin (Epo) values and Hb in apparent erythrocytosis (AE).

Download figure to PowerPoint

image

Figure 3. Serum erythropoietin (Epo) values and Hb in idiopathic erythrocytosis (IE).

Download figure to PowerPoint

image

Figure 4. Serum erythropoietin (Epo) values and Hb in secondary erythrocytosis (SE).

Download figure to PowerPoint

Forty-two patients with PV, 35 with AE, 31 with IE and 17 with SE had, respectively, 56, 50, 46 and 25 Epo values plotted according to the Hb interval in which these values occurred (for each patient only the first Epo value in each Hb interval was plotted). Figures 1 and 4 illustrate the prominent number of samples with low Epo values in PV and raised Epo values in SE. There were only three low Epo values in AE, one in IE and four in SE (each Epo value derived from a different patient); these all occurred at Hb values above 16 g/dl and almost all were in untreated patients. Raised Epo in AE and IE was more common than low values; 15 raised values were plotted in AE and 9 in IE (each Epo value derived from a different patient). The raised values in AE and IE occurred mainly when the Hb was below 16 g/dl. Only two raised Epo values (derived from one patient) were observed in PV. This patient had a Hb of 10·8 g/dl and 13·6 g/dl at the time of the two raised Epo values and was recovering from a gastrointestinal haemorrhage. Thus, the few low Epo values in erythrocytoses other than PV, and raised Epo values in erythrocytoses other than SE, seem to be chiefly associated with raised and normal Hb values respectively.

Epo values, Hb and incidence of thrombosis in PT

Forty-nine patients with PT were studied. Sixteen patients (33%) had a low Epo value (< 2·9 mIU/ml) in the first blood sample taken from them for this study. There was no evidence of ‘masked PV’ in these patients as they had normal PCV, mean cell volume (MCV) and ferritin values. No association was found between the Epo values and presence of raised white count or splenomegaly. Eleven of the 16 patients with a low Epo had a history of thrombosis or ischaemia, compared with only 16 of 33 patients without a low Epo. Although there appeared to be an association between low Epo values and thrombosis in this group of PT patients, this was not statistically significant.

The incidence of low, normal and raised Epo values at different Hb intervals in PT is shown in Fig 5. A total of 56 Epo values were plotted according to the Hb interval in which these values occurred (for each patient only the first Epo value in each Hb interval was plotted). Of the 19 low Epo values plotted, 4 (21%) occurred at Hb ≤ 14 g/dl and 15 (79%) occurred at Hb > 14 g/dl. Of the 37 normal or raised Epo values plotted, 29 occurred at Hb ≤ 14·0 g/dl and 8 at Hb > 14·0 g/dl. The proportion of patients with Hb > 14·0 g/dl was significantly higher in patients with low Epo than in those with normal or raised Epo (P < 0·001). All five Epo values in patients with Hb > 16 g/dl were low (these patients all had PCV values within the normal reference range and no evidence of erythrocytosis). Low Epo values therefore appeared to be more common at the higher Hb intervals in PT.

image

Figure 5.  Serum erythropoietin (Epo) values and Hb in primary thrombocythaemia (PT).

Download figure to PowerPoint

Among the 46 patients with PT in whom clinical information was available, 76% (22 of 29 patients) of those with a history of thrombosis had Hb values greater than 14·0 g/dl, while only 47% (8 of 17 patients) of those without a history of thrombosis had Hb values in that range. The Hb value nearest the time of thrombosis was used in this analysis. It suggests an association between higher Hb values in PT and incidence of thrombosis (P < 0·05).

Of the 56 Epo values plotted in PT, 23 were from patients undergoing cytoreductive treatment for PT at the time of study (19 hydroxyurea, 2 α-interferon, 2 intermittent low-dose busulphan) and 33 were untreated. Low Epo values were found in 5 (22%) of the treated group and 14 (42%) of the untreated group, although this difference was not found to be significant.

Variation of median Epo values with Hb

Epo values obtained in each Hb interval for each of the erythrocytoses (PV, AE, IE, SE) and for PT are plotted in Fig 6. For each of the total of 174 patients, only the first Epo value in each Hb interval was plotted (total of 233 values). The median values of Epo for each disease at each Hb interval were compared. As expected, low Epo values (median below the reference range) predominated in PV at all Hb intervals. Low Epo values occurred next most frequently in PT, but the median values were only low at the higher Hb intervals in this disease. The other erythrocytoses (AE, IE, SE) showed some reduction in the median Epo values at the higher Hb intervals only (although these median values did not fall below the reference range).

image

Figure 6. Serum erythropoietin (Epo) values in erythrocytoses and in primary thrombocythaemia (PT) at different Hb intervals. The shaded area indicates the normal reference range for Epo.

Download figure to PowerPoint

Reproducibility of Epo values

The Epo interpretations (high, normal or low) per patient per single Hb interval were classed as either ‘same’ or ‘different’ in those patients for whom two or more samples were available within the same Hb interval (51 patients). Table I shows good Epo value reproducibility from repeated sampling within a single Hb interval in individual patients (the one patient with six samples within the same Hb interval was omitted from this table).

Table I.  Reproducibility of serum erythropoietin (Epo) interpretations in patients having two, three or four samples.
 Number of samples per patient per Hb interval
223344
Interpretation of Epo resultSameDifferentSameDifferentSameDifferent
 Patient numbers     
Polycythaemia vera 4310
Idiopathic erythrocytosis1120310
Apparent erythrocytosis12070
Secondary erythrocytosis 10
Primary thrombocythaemia 50

The actual Epo values in the few patients (nine) in whom Epo value interpretations changed on repeated sampling within a single Hb interval are illustrated in Fig 7 (3 PV, 6 IE).

image

Figure 7. Patients with different Serum erythropoietin (Epo) interpretations on repeat sampling (each vertical line is an individual patient). The shaded area indicates the normal reference range for Epo.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. References

Current methods for measuring Epo are not always adequately sensitive at low Epo concentrations (Lindstedt & Lundberg, 1998). The ELISA method used in this study is promising in this respect and also in its reproducibility when comparing different samples in individual patients (Table I, Fig 7).

Our data show high specificity of low (below the reference range) Epo values for the diagnosis of PV (92%) compared with the other erythrocytoses (AE, IE and SE). False positives were rare, although a small number of low Epo values were found in AE patients (3 out of 35), in IE (1 out of 31) patients and in SE patients (4 out of 7); all these false positive low values occurred in samples having Hb values above 16 g/dl (Figs 2, 3 and 4), and almost all were in untreated patients. Thus, low Epo values are less specific for PV at the higher Hb levels in our study. Carneskog et al (1999) and Shih et al (1997) have reported a similar incidence of low Epo values in AE and IE patients, respectively, but did not relate this to Hb. It is possible that these AE and IE patients might develop overt PV in the future and this possibility needs to be addressed during follow-up. Examination of the clinical and laboratory data in the AE and IE patients with low Epo did not suggest that a diagnosis of PV had been missed. Our data imply that treated patients with SE, AE and IE, having overall lower Hb levels than untreated patients, are less likely to have low serum Epo values that may misleadingly suggest a diagnosis of PV. Low serum Epo estimations obtained in untreated patients with SE, AE and IE should therefore be repeated after treatment.

The sensitivity of low Epo in the diagnosis of PV was lower in this study than in that of Birgegård & Wide (1992) but similar to our previous study (Messinezy et al, 1995) and similarly unrelated to whether or not the patients had received treatment. Our current study in a different group of patients with PV and using a different method for Epo estimation showed 36% false negative results (diagnostic sensitivity 64%). A more robust diagnostic marker for PV, perhaps based on clonality or molecular genetics, is clearly still needed.

Our study confirms the known relatively poor sensitivity (47%) of raised Epo values in the diagnosis of SE (Cotes et al, 1986). We studied insufficient patients to distinguish between different causes of SE in terms of Epo values. When raised Epo values in SE were tested versus PV, the specificity was very high (98%). A raised Epo value therefore makes the diagnosis of PV very unlikely. However Figs 2 and 3 show that raised Epo values do occur in AE and IE (mainly at Hb < 16 g/dl). Missing a diagnosis of SE is always a possibility in such patients especially because of the large numbers of potential causes of SE; careful exclusion of underlying SE seems particularly important when characterizing an erythrocytosis and also during follow-up.

The 33% incidence of low Epo in PT was similar to although a little lower than found in other studies (Viala et al, 1993; Najean et al, 1995; Carneskog et al, 1998). Low Epo values were more common (but not significantly so) in patients who had not received cytoreductive treatment. There was no evidence that a raised PCV (erythrocytosis) was missed or masked (as by iron deficiency) in our patients, although blood volume measurements were not done (the PCV values being normal). Because raised platelet counts are commonly a feature of PV, the possibility that PV and PT are overlapping diseases still needs to be considered. A case for routine blood volume estimations in patients with PT and high/normal PCV values (especially in the presence of splenomegaly) could be made.

While Epo values are known to be inversely related to Hb in normals, one would not expect low (below normal) Epo values to result from Hb levels in the higher part of the normal range, as found in a large minority of our PT patients. The significance of the frequent low Epo values found in patients with PT is uncertain. Our demonstration of an association between low Epo values and higher Hb values (> 14 g/dl, P < 0·001) deserves further scrutiny in larger numbers of PT patients. The possibility that PT patients whose Hb is in the higher normal range are at increased risk of thrombotic incidents has been suggested by this study and could have therapeutic implications. While we did not find a statistically significant correlation between low Epo values and thrombosis in PT, there was a trend and this would be worth re-examinination in a larger group of patients. Examination of clonality, growth of endogenous erythroid colonies or growth patterns related to varying Epo concentrations and bone marrow histology among other investigations could further subdivide the PT group and may show additional correlations with the Epo values in the future.

Acknowledgment

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. References

We are grateful to Monica Nestor for preparing the manuscript.

References

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. References
  • Birgegård, G. & Wide, L. (1992) Serum erythropoietin in the diagnosis of polycythaemia and after phlebotomy treatment. British Journal of Haematology, 81, 603606.
  • Carneskog, J., Kutti, J., Wadenvik, H., Lundberg, P.-A. & Lindstedt, G. (1998) Plasma erythropoietin by high-detectability immunoradiometric assay in untreated and treated patients with polycythaemia vera and essential thrombocythaemia. European Journal of Haematology, 60, 278282.
  • Carneskog, J., Safai-Kutti, S., Suurküla, M., Wadenvik, H., Bake, B., Lindstedt, G. & Kutti, J. (1999) The red cell mass, plasma erythropoietin and spleen size in apparent polycythaemia. European Journal of Haematology, 62, 4348.
  • Cotes, P.M., Doré, C.J., Liu Yin, J.A., Lewis, S.M., Messinezy, M., Pearson, T.C. & Reid, C. (1986) Determination of serum immunoreactive erythropoietin in the investigation of erythrocytosis. New England Journal of Medicine, 315, 283287.
  • Lindstedt, G. & Lundberg, P.-A. (1998) Are current methods of measurement of erythropoietin (EPO) in human plasma or serum adequate for the diagnosis of polycythaemia vera and the assessment of Epo deficiency? Scandinavian Journal of Clinical Laboratory Investigation, 58, 441458.
  • Messinezy, M., Westwood, N.B., Woodcock, S.P., Strong, R.M. & Pearson, T.C. (1995) Low serum erythropoietin – a strong diagnostic criterion of primary polycythaemia even at normal haemoglobin levels. Clinical and Laboratory Haematology, 17, 217220.
  • Najean, Y., Schlageter, M.H., Toubert, M.E. & Rain, J.D. (1995) Erythropoietin concentration in the serum from patients with primary thrombocythaemia. European Journal of Haematology, 55, 272273.
  • Pearson, T.C. (1998) Diagnosis and classification of erythrocytoses and thrombocytoses. Bailliere's Clinical Haematology, 11, 695720.
  • Pearson, T.C. & Messinezy, M. (1996) The diagnostic criteria of polycythaemia rubra vera. Leukemia and Lymphoma, 22, 8793.
  • Shih, L.Y., Lee, C.T. & Ou, Y.C. (1997) Prediction of clinical course in patients with idiopathic erythrocytosis by endogenous erythroid colony assay but not by serum erythropoietin levels. Experimental Hematology, 25, 288292.
  • Viala, J.J., Ville, D., Sebban, C., Assouline, D., Devaux, Y. & Hanss, M. (1993) Plasma erythropoietin in essential thrombocythaemia. Nouvelle Revue Francaise D'hematologie, 35, 423424.