血涂片诊断

The new england journal of medicine

current concepts

Diagnosis from the Blood Smear

Barbara J. Bain, F.R.A.C.P., F.R.C.Path.

From the Department of Haematology, St.

Mary’s Hospital, London. Send reprint re-

quests to Dr. Bain at St. Mary’s Hospital,

Praed St., London W2 1NY, United King-

dom, or at [email protected].

N Engl J Med 2005;353:498-507.

Copyright 2005 Massachusetts Medical Society.

A related slide show is available at

www.nejm.orgn examination of the blood smear (or film) may be requestedby physicians or initiated by laboratory staff. With the development of sophis-ticated automated blood-cell analyzers, the proportion of blood-count sam-ples that require a blood smear has steadily diminished and in many clinical settings isnow 10 to 15 percent or less. Nevertheless, the blood smear remains a crucial diagnos-tic aid. The proportion of requests for a complete blood count that generate a bloodsmear is determined by local policies and sometimes by financial and regulatory as wellas medical considerations. For maximal information to be derived from a blood smear,the examination should be performed by an experienced and skilled person, either a lab-

oratory scientist or a medically qualified hematologist or pathologist. In Europe, only

laboratory-trained staff members generally “read” a blood smear, whereas in the Unit-

ed States, physicians have often done this. Increasingly, regulatory controls limit the

role of physicians who are not laboratory-certified. Nevertheless, it is important for

physicians to know what pathologists or laboratory hematologists are looking for and

should be looking for in a smear. In comparison with the procedure for an automated

blood count, the examination of a blood smear is a labor-intensive and therefore rela-

tively expensive investigation and must be used judiciously.

A physician-initiated request for a blood smear is usually a response to perceived

clinical features or to an abnormality shown in a previous complete blood count. A lab-

oratory-initiated request for a blood smear is usually the result of an abnormality in the

complete blood count or a response to “flags” produced by an automated instrument.

Less often, it is a response to clinical details given with the request for a complete blood

count when the physician has not specifically requested examination of a smear. For

example, a laboratory might have a policy of always examining a blood smear if the clin-

ical details indicate lymphadenopathy or splenomegaly. The International Society for

Laboratory Hematology has published consensus criteria (available at www.islh.org)

for the laboratory-initiated review of blood smears on the basis of the results of the au-

tomated blood count. The indications for smear review differ according to the age and

sex of the patient, whether the request is an initial or a subsequent one, and whether

there has been a clinically significant change from a previous validated result (referred

to as a failed delta check). All laboratories should have a protocol for the examination

of a laboratory-initiated blood smear, which can reasonably be based on the criteria of

the International Society for Laboratory Hematology. Regulatory groups should permit

a

these differ somewhat from the reasons why laboratory workers initiate a blood-smear

examination. Sometimes it is possible for a definitive diagnosis to be made from a blood

smear. More often, the smear is an important tool in the provision of a differential di-

498n engl j med 353;5www.nejm.orgaugust 4, 2005

current concepts

agnosis and the indication of further necessary tests.

The blood smear can have an important part in the

speedy diagnosis of certain specific infections. Oth-

erwise, its major roles are in the differential diag-

nosis of anemia and thrombocytopenia and in the

identification and characterization of leukemia and

*In acute disseminated intravascular coagulation, red-cell fragments may be

absent.inations of blood smears are usually performed inresponse to clinical features or to a previously ab-normal complete blood count. The presence of un-explained jaundice, particularly if unconjugated hy-perbilirubinemia is also present, is an additionalreason for a blood-smear examination. Laborato-ry-initiated examinations of blood smears for pa-tients with anemia are usually the result of a labo-ratory policy according to which a blood smear isordered whenever the hemoglobin concentrationis unexpectedly low. This policy should be encour-aged, since the consideration of the blood smearand the red-cell indices is a logical first step in theinvestigation of any unexplained anemia.1 Initiat-ing a smear as a reflex test also means that a further

blood sample does not have to be taken for this

purpose.

Modern automated instruments impart valuable

information about the nature of anemia. They pro-

vide not only a red-cell count, mean cell volume,

mean cell hemoglobin (a measure of the average

amount of hemoglobin in an individual red cell),

and the mean cell hemoglobin concentration (a

measure of the average concentration of hemoglo-

bin in a cell) but also newer variables that give in-

formation that previously could be derived only

from a blood smear. These variables usually include

the red-cell–distribution width, which correlates on

a blood smear with anisocytosis, and they may also

include the hemoglobin-distribution width and the

percentages of hypochromic and hyperchromic

cells, which correlate with anisochromasia, hypo-

chromia, and hyperchromia. A variety of histograms

and scatterplots give a visual representation of red-

cell characteristics. It may be possible to detect in-

creased numbers of hyperchromic cells (spherocytes

or irregularly contracted cells), small hyperchromic

cells (microspherocytes), hypochromic microcyt-

ic cells, large normochromic cells (normally hemo-

globinized macrocytes), and hypochromic macro-

cytes (either reticulocytes or dysplastic red cells).Despite this wealth of information, there arestill morphologic abnormalities that are critical inthe differential diagnosis of anemia and that canbe determined only from a blood smear. Particu-larly important is the detection of variations in cellshape and of red-cell inclusions, such as Howell–Jolly bodies (nuclear fragments), Pappenheimerbodies (hemosiderin-containing granules), and ba-sophilic stippling or punctate basophilia (alteredribosomes).hemolytic anemiaIn the hemolytic anemias, red-cell shape is of con-siderable diagnostic importance. Some types ofhemolytic anemia yield such a distinctive bloodsmear that the smear is often sufficient for diagno-sis. This is true of hereditary elliptocytosis (whichis only infrequently associated with anemia) (Fig.1A) (a slide show is included in the SupplementaryAppendix, available with the full text of this articleat www.nejm.org), hereditary pyropoikilocytosis(Fig. 1B), and Southeast Asian ovalocytosis, a dis-tinctive type of inherited hemolytic anemia that iscommon in some parts of Southeast Asia and is

499n engl j med 353;5www.nejm.orgaugust 4, 2005

The new england journal of medicine

now also seen in Europe and North America as a

result of immigration (Fig. 1C). The presence of

spherocytes is not diagnostically specific, since this

may result from hereditary spherocytosis, autoim-

mune hemolytic anemia, or alloimmune hemolytic

anemia (e.g., hemolytic disease of the newborn or

a delayed transfusion reaction). Nevertheless, con-

sideration of the clinical features, together with the

results of a direct antiglobulin test, in patients with

spherocytes will generally indicate the correct diag-

nosis.

Microspherocytes (i.e., cells that are both hy-

perchromic and significantly reduced in size and

therefore in diameter) may be present in low num-

bers in patients with a spherocytic hemolytic ane-

mia but are also characteristic of burns and of mi-

croangiopathic hemolytic anemia. The detection of

a microangiopathic hemolytic anemia (Fig. 1D) isof considerable clinical significance, since this typeof anemia may indicate pregnancy-associated hy-pertension, disseminated cancer, chronic dissemi-nated intravascular coagulation, the hemolytic–ure-mic syndrome, or thrombotic thrombocytopenicpurpura; the latter two conditions both require ur-gent diagnosis so that appropriate management canbe initiated. In microangiopathic hemolytic anemia,examination of the blood smear is also importantto validate the platelet count, since red-cell frag-ments and platelets may be of similar size. Most au-tomated instruments cannot make this distinction.A minority of automated instruments that measureboth the size and the refractive index of small parti-cles in the blood sample can make this distinctionand can be used to exclude red-cell fragmentation;however, although the fragment “flag” on such in-struments is sensitive, it is not specific. Hence, a

500n engl j med 353;5www.nejm.orgaugust 4, 2005

current concepts

blood smear is still advised for validation.2 Blood-hemolysis is most often seen in glucose-6-phos-smear features similar to those seen in microan-phate dehydrogenase (G6PD) deficiency but cangiopathic hemolytic anemia are also a feature ofalso occur with other defects in the pentose shuntmechanical hemolytic anemia, such as that associ-or in glutathione synthesis and when oxidant expo-ated with a leaking prosthetic valve, and provide im-sure overwhelms normal protective mechanisms.portant evidence of this cause of hemolytic anemia.Oxidant damage may be exogenous, as in exposure

A blood smear is particularly important in theto oxidant chemicals or drugs (most often dapsone),diagnosis of acute hemolysis induced by oxidantor endogenous, as in Wilson’s disease.3

damage. The characteristic feature is the presenceG6PD deficiency affects millions of personsof keratocytes, or “bite” cells (Fig. 2A and the Sup-worldwide. A blood smear is important for the di-plementary Appendix), “blister” cells (Fig. 2B), andagnosis of this condition, for two reasons. First, itirregularly contracted cells (Fig. 2B); the latter mustis available far more rapidly than are the results ofbe distinguished from spherocytes (Fig. 2C) be-a G6PD assay and, when considered together withcause of the quite different diagnostic significance.the patient’s ethnic origin and clinical history, per-These irregularly contracted cells share with sphero-mits a provisional diagnosis. Second, a blood smearcytes the lack of central pallor but differ in thatcan suggest the diagnosis of G6PD deficiency eventhey have an irregular outline. Oxidant-inducedif a G6PD assay is normal. Normal G6PD activity

n engl j med 353;5www.nejm.orgaugust 4, 2005

501

The new england journal of medicine

may be found after acute hemolysis in G6PD-defi-

cient persons in two situations. First, men of Afri-

can or African-American ancestry who are hemizy-

gous for the A¡ alloenzyme (which is present at

normal levels in reticulocytes) can have normal

G6PD levels because the reticulocyte count is high

after hemolysis. Second, female carriers — for ex-

ample, those who are hemizygous for the common

Mediterranean variant of G6PD — can have a nor-

mal assay result after an acute hemolytic episode

because the abnormal cells have lysed preferential-

ly, leaving mainly the cells expressing the normal

allele in the circulation. In both of these circum-

stances, the observation of a typical blood smear in

an appropriate clinical setting is an indication to

repeat the assay once the acute hemolytic episode

is over.

Other features may aid in the differential diag-

nosis of hemolytic anemia. For example, the pres-

ence of red-cell agglutinates usually indicates the

presence of a cold agglutinin, and erythrophagocy-

tosis is often a feature of paroxysmal cold hemo-

globinuria (Fig. 2D).

macrocytic anemiather investigation and a trial of treatment are need-ed. Liver disease and excess ethanol consumptionare common causes of macrocytosis, with the bloodsmear usually showing round rather than oval mac-rocytes and lacking hypersegmented neutrophils;target cells and stomatocytes may also be present.In elderly patients, the myelodysplastic syn-dromes are an important cause of macrocytosis.Blood-smear features that may point to the diagno-sis include hypogranular or hypolobulated neu-trophils (Fig. 3B), blast cells (Fig. 3B), giant or hy-pogranular platelets, Pappenheimer bodies (Fig.3C), and the presence of a minor population of hy-pochromic microcytic cells, leading to a dimorphicsmear (Fig. 3C). Macrocytic anemia resulting fromcongenital dyserythropoietic anemia also yields acharacteristic blood smear, with striking poikilocy-tosis (Fig. 3D). When macrocytosis is the result ofhemolysis or recent blood loss, the blood smearshows polychromasia, which results from an in-creased reticulocyte count.microcytic anemia

The blood smear is of great importance in the dif-

ferential diagnosis of macrocytic anemias. For pa-

tients in whom there is a deficiency of vitamin B12

or folic acid, the blood smear shows not only mac-

rocytes but also oval macrocytes and hypersegment-

ed neutrophils (Fig. 3A and the Supplementary Ap-

pendix). When the anemia is more severe, there may

be marked poikilocytosis, with teardrop poikilo-

cytes and red-cell fragments. Although these defi-

ciency states are now usually recognized on the ba-

sis of assays of vitamin B12 and folic acid, the blood

smear remains important for two reasons. First, it

permits a speedy provisional diagnosis, and initia-

tion of appropriate treatment in severely anemic

patients while assay results are pending. Second,

occasionally there are patients with a clinically sig-

nificant vitamin B12 deficiency despite a normal as-

say result. This discrepancy occurs because much

of the vitamin B12 that is measured in the assay is

bound to haptocorrin, whereas the functional vita-

min B12, which is bound to transcobalamin, con-

tributes much less to the assay of total B12.

Similarly, acute folic acid deficiency sometimes

develops in patients even though the total red-cell

folate level remains normal. The observation of a

blood smear that is typical of megaloblastic ane-

mia despite normal assays is an indication that fur-

502The blood smear is generally less important in thedifferential diagnosis of the microcytic than themacrocytic anemias. Red-cell indices and serumferritin levels, sometimes supplemented by mark-ers of inflammation, that are interpreted in the con-text of clinical features, permit the diagnosis of themajority of cases. However, it is important to notethat the presence of Pappenheimer bodies and red-cell dimorphism in the sideroblastic anemias andof basophilic stippling in cases of lead poisoning(Fig. 4A and the Supplementary Appendix) and insome types of thalassemia is diagnostically sig-nificant.hemoglobinopathy and thalassemiaA blood smear is useful in the diagnosis and differ-ential diagnosis of sickle cell disease, particularly ifthere is an urgent need for diagnosis and if the re-sults of hemoglobin electrophoresis or high-per-formance liquid chromatography are not instantlyavailable. Patients with sickle cell anemia (in whichthere is homozygosity for hemoglobin S) have ane-mia, but those with compound heterozygosity forhemoglobin S and hemoglobin C may have a nor-mal hemoglobin level, and the condition thus maybe confused with sickle cell trait if a blood smear isnot examined. Consideration of the blood-smearfeatures, of the hemoglobin level, and of the resultsof a sickle cell solubility test usually permits an ac-n engl j med 353;5www.nejm.orgaugust 4, 2005

current concepts

curate diagnosis4,5 (Fig. 4B and 4C). The bloodcondition (Fig. 4D); sometimes there is coexist-

smear of a compound heterozygote usually showstarget cells, irregularly contracted cells, and boat-

shaped cells but few classic sickle cells; typical he-

moglobin SC poikilocytes (formed only when he-moglobin S and hemoglobin C are both present)are often seen. Sometimes the blood smear of acompound heterozygote shows only target cellstients with thrombocytopenia, both to confirm theand irregularly contracted cells and cannot be dis-thrombocytopenia and to look for the underlyingtinguished from the smear in hemoglobin C homo-cause. Falsely low platelet counts may be the resultzygosity; a positive sickle cell solubility test per-of small clots, platelet clumping (Fig. 5A and themits these conditions to be distinguished in anSupplementary Appendix), platelet satellitism (Fig.emergency situation (e.g., preoperatively). A blood5B), or abnormally large platelets. Fibrin strandssmear is also important in the diagnosis of an un-(Fig. 5C) indicate that thrombocytopenia is likelystable hemoglobin, with irregularly contractedto be factitious. Underlying causes that may be re-cells and macrocytosis being characteristic of thisvealed by the blood smear include the May–Hegglin

n engl j med 353;5www.nejm.orgaugust 4, 2005

503

The new england journal of medicine

anomaly (Fig. 5D), microangiopathic thrombopa-

thies, and leukemias and lymphomas. High plate-

let counts should be confirmed microscopically

with a blood smear; falsely high counts may be the

result of other particles (red-cell fragments, frag-

ments of leukemic cells, or fungi) being counted as

platelets.6-9 Examination of the blood smear is also

important in patients with thrombocytosis to look

for evidence of a myeloproliferative disorder, such

as giant platelets, or an increase in the basophil

count; the latter is not reliably detected by auto-

mated counters. A sudden, unexpected improve-

ment in the platelet count also should be confirmed

by blood-smear examination, since such an im-

provement may be factitious7 (Fig. 5E).

there is unexplained leukocytosis, lymphocytosis,or monocytosis or when the flagging system of anautomated instrument suggests the presence ofblast cells. Depending on the instrument and thepractice of the local laboratory, a flag for atypical orvariant lymphocytes may also be an indication forexamination of a blood smear, since this flag issometimes indicative of the presence of blast cells.Low rather than high counts likewise are an indi-cation for a smear, since they may be indicative ofaplastic anemia, acute leukemia, hairy-cell leuke-

504n engl j med 353;5www.nejm.orgaugust 4, 2005

current concepts

mia, or infiltration of nonhematopoietic malignantproduces a highly improbable result. Such results

cells into the bone marrow. The role of the bloodmay be factitious, resulting from the accidentalsmear in the diagnosis of leukemia and lymphomafreezing or heating of the blood, from hyperlipide-is to suggest a likely diagnosis or range of diag-mia, or from the presence of cold agglutinins, a cry-noses, to indicate which additional tests should beoglobulin (Fig. 6C), bacteria, or fungi. Factitiousperformed, and to provide a morphologic contextresults also may stem from unusual characteristicswithout which immunophenotyping and other so-of the blood cells or the plasma, such as a pseudo-phisticated investigations cannot be interpreted.neutropenia caused by a myeloperoxidase deficien-For two conditions, Burkitt’s lymphoma (Fig. 6Acy that occurs when the automated instrument em-and the Supplementary Appendix) and acute pro-ploys a peroxidase reaction for the identificationmyelocytic leukemia (Fig. 6B), a blood smear is ofof neutrophils, eosinophils, and monocytes. Falselyparticular importance because it facilitates rapid

low counts also may result from neutrophil or plate-tiate a blood smear if an automated instrumentdiagnosis that can be very important to the patient

n engl j med 353;5www.nejm.orgaugust 4, 2005

505

The new england journal of medicine

(Table 2). As an example, the detection of features

of unexpected hyposplenism (Fig. 6D) may suggest

a congenital absence of the spleen, splenic atrophy,

deposition of amyloid in the spleen, infiltration of

neoplastic cells (e.g., in leukemia, lymphoma, or

carcinoma) in the spleen, previous splenic infarc-

tion, or even a splenectomy of which the patient

was unaware — in each case putting the patient at

risk for complications of hyposplenism. Converse-

ly, the failure to observe expected hyposplenism in

a blood smear from a patient who has undergone

splenectomy for the treatment of autoimmune

thrombocytopenic purpura may indicate that there

is functioning residual splenic tissue, either from

splenosis or from accessory spleens, that may be

responsible for a relapse of the disease.

or the only evidence of a specific diagnosis, such asmyelodysplastic syndrome, leukemia, lymphoma,or hemolytic anemia. It is important that, if possi-ble, such blood smears be stored over the long term,just as a tissue that provides a histologic diagnosisis stored over the long term. In practice, such stor-age is easily achieved if a patient has also had a bonemarrow aspirate (since a blood smear should al-ways be stored with an aspirate), but it is harder toachieve if the blood smear alone has provided thediagnosis. Individual laboratories should have amechanism to make possible the retention of such

506n engl j med 353;5www.nejm.orgaugust 4, 2005

current concepts

smears or an image derived from them. Some lab-

oratories retain all smears that have been reviewed

by a laboratory hematologist or pathologist; this

can create a storage problem, and it is likely that,

increasingly, digital images of important abnormal

highlighted by the recent introduction of photo-graphs of blood smears as a regular feature in both

the journal Blood

10 and the British Journal of Haema-

tology, by ongoing efforts to develop image-recog-

nition technology for the automated examination

of blood smears, and by the development of tele-ry staff should make and examine a blood smearhematology to permit the remote interpretation orwhenever the results of the complete blood count11,12indicate that a blood smear is essential for the vali-

dation or the further elucidation of a detected ab-

normality. If error is to be avoided, sophisticatedmodern investigations of hematologic disorders

should be interpreted in the light of peripheral-bloodsmear remains an important diagnostic tool. Physi-features as well as the clinical context.cians should request a blood smear when there areI am indebted to Dr. Bernadette Garvey of Toronto and Dr. LoAnneclinical indications for it. Members of the laborato-Peterson of Chicago for their helpful comments on the manuscript.1.Case Records of the Massachusetts Gen-

eral Hospital (Case 30-2004). N Engl J Med

2004;351:1333-41.

2.Lesesve JF, Salignac S, Alla F, et al. Com-

parative evaluation of schistocyte counting

by an automated method and by microscop-

ic determination. Am J Clin Pathol 2004;121:

739-45.

3.Bain BJ. Heinz body haemolytic anaemia

in Wilson’s disease. Br J Haematol 1999;104:

647.

4.Diggs LW, Bell A. Intraerythrocytic he-

moglobin crystals in sickle cell-hemoglobin

C disease. Blood 1965;25:218-23.

5.Bain BJ. Blood smear features of sicklecell-haemoglobin C disease. Br J Haematol1993;83:516-8.6.Latif S, Veillon DM, Brown D, et al.Spurious automated platelet count: enu-meration of yeast forms as platelets by theCell-DYN 4000. Am J Clin Pathol 2003;120:882-5.7.Arnold JA, Jowri Z, Bain BJ. Candida gla-brata in a blood smear. Br J Haematol 1999;104:1.8.van der Meer W, MacKenzie MA, Din-nissen JW, de Keijzer MH. Pseudoplatelets:a retrospective study of their incidence andinterference with platelet counting. J ClinPathol 2003;56:772-4.9.Kakkar N. Spurious rise in the automat-ed platelet count because of bacteria. J ClinPathol 2004;57:1096-7.10.Shattil SJ. A (blood) smear campaign.Blood 2003;101:2453.

11.Abramson N. Inside blood: a picture (inthe microscope) is worth a thousand words.Blood 2004;103:367-8.12.Luethi U, Risch L, Korte W, Bader M,Huber R. Telehematology: critical determi-nants for successful implementation. Blood2004;103:486-8.Copyright 2005 Massachusetts Medical Society.

n engl j med 353;5www.nejm.orgaugust 4, 2005507