THE LANCET         Vol 338   Saturday 21 September 1991   No 8769
 

Chronic fatigue syndrome: clinical condition associated with immune activation
ALAN L. LANDA Y CAROL JESSOP EVEL YNE T. LENNETfE  JAY A. LEVY

There is much conflicting immunological and viral data about the causes of chronic fatigue syndrome (CFS); some findings support the notion that CFS may be due to one or more immune disorders that have resulted from exposure to an infectious agent. In the present study, flow cytometry and several different monoclonal antibodies recognising T, B, and natural killer (NK) cell populations as well as activation and cell adhesion antigens were used to study 147 individuals with CFS. Compared with healthy controls, a reduced CD8 suppressor cell population and increased activation markers (CD38, HLA-DR) on CD8 cells were found. The differences were significant (p = 0,01) in patients with major symptoms of the disease. These immunological indices were not observed in 80 healthy individuals, in 22 contacts of CFS patients, or in 43 patients with other diseases. No correlation of these findings in CFS patients with any known human viruses could be detected by serology. The findings suggest that immune activation is associated with many cases of CFS.
                                Lancet 1991; 338: 707-12.

Introduction
Chronic fatigue syndrome (CFS) has been reponed with increased frequency in the USA, UK, Australia, and other parts of the world.
1 The disorder is characterised by a debilitating fatigue lasting longer than six months with chronic and recurrent low-grade fever, pharyngitis, adenopathy, myalgia, arthralgia, sleep disorders, as well as difficulties in cognition and temperament.
2 In many patients, the syndrome begins with an acute "flu-like" illness. Despite these clinical fmdings, some physicians question whether there is such a syndrome and have related the observations to depression or stress.
3 The Centers for Disease Control (CDC) have publish_d an epidemiological case defmition for CFS, which uses major and minor clinical and laboratory criteria.'

A viral cause of CFS has been suspected because several viral infections are characterised by a chronic post-infection fatigue and because the onset of CFS often resembles an acute viral illness.l.2 However, in other viral infections, in contrast to CFS, the symptoms do not generally persist after several weeks. Initial studies showed that CFS was associated with high concentrations of antibodies to Epstein-Barr virus (EBV).s.6 However, subsequent studies suggested that high antibody titres to EBV were not found in all CFS patients, and that polyclonal activation of B cells was a common finding with antibodies to several viruses, especially herpesviruses. 7.8   Some investigators have reponed enteroviral RNA in muscle tissue of people with CFS: but a role for Coxsackie B virus in CFS has not been supported by serological studies.1O

Lately, serological and polymerase chain reaction methods have pointed to an association of a human T -lymphotropic virus-like agent with the syndrome." Nevenheless, no conclusive evidence of a common causative agent in CFS has been presented.

Immunological disorders such as those seen in viral infections have also been described in CFS-eg, decteased function of natural killer (NK) cells and macrophages. reduced mitogenic response of lymphocytes, B-cell subset changes, and activation of CD8 cells. 12-16   Moreover, infected animals and patients with or recovering from various viral infections often show transient immune abnormalities and chronic fatiguey.18   These findings further support the notion that CFS involves immune disorders due most likely to exposure to an infectious agent.

To see whether we could resolve some of these conflicting data we evaluated certain virological and immunological indices in a clinically well defined cohort of patients with CFS and compared them with control populations.

ADDRESSES: Department of Immunology/Microbiology. RushPresbyterian-St. luke's Medical Center. Chicago. Illinois (A. L Landay. PhD); Virolab, Berkeley. California (E T. Lennette. PhD); and Department of Medicine (C. Jessop, MD, Prof J. A. Levy, MD) and Cancer Research Institute (Prof J. A. Levy), University of California, School of Medicine, San Francisco, California. USA. Correspondence to Prof J. A. Levy. University of California. San Francisco. Room S1280. 3rd and Parnassus. San Francisco. California 94143. USA.

                                708    THE LANCET

                    Subjects and methods
Subjects
A total of 147 consecutive patients (30 males, 117 females; median age 38 years, range 15--80) who presented with CFS were studied from September, 1989, to November, 1990. Either at the time of evaluation or according to medical history, all patients met the CDC epidemiological case definition for CFS: including symptoms shown in table I.  As recommended by the CDC criteria, other conditions that could cause fatigue excluded individuals from the study.  The extent of physical activity was assessed in the CFS patients by a self-administered questionnaire that asked the individuals to rate their activity on a scale of I (least active) to 10. Additionally, patients were given other questions about their clinical symptoms. (The questionnaire is available from us on rcquest).  Cognitive ability and muscle strength were assessed by history and by mental status and physical examinations. In most cases, neither the physical examination nor laboratory tests were helpful in assessing muscle function. Patients had been ill for one to five years. Most of the patients were caucasian; 20% were Asian. Racial background did not affect the results. Randomly seleCted control subjects consisted of healthy individuals living in San Francisco working at the UCSF Medical Center (n=50), and subjects who were being seen for routine physical examination (n = 30). Immunophenotypic data on these two control groups were not significantly different, so they were combined for this study as one control group (30 males, 50 females; median age 32 years, range 20-55). Racial distribution was similar to the patient group.

Additional control populations that were evaluated included patients with acute viral-like illness (n = 15); patients with documented depression (n = 10); spouses and family members of
CFS patients (n = 11); medicalmedical personel in  contact with €FS patients (n = 11); individuals with prolonged fatigue without  other clinical criteria for CFS (n = 6); and patients with systemic lupus erythematosus (SLE) (n = 12) as defined by the Amerietm Rheumatology Association. These non-CFS subjects had been referred by physicians in the San Francisco and Chicago areas. Blood samples obtained from all individuals studied were coded and thc data were correlated to clinical status after the results had been tabulated. The studies were approved by the Committee for Human Research of the University of California, San Francisco (UCSF).

Blood samples
Anticoagulated blood (ethylenediamine tetra-acetic acid) was collected for flow cytometric studies, white blood cell counts, and differential counts. An additional serum sample was taken for viral serological studies.

Viral serology
Antibodies to various viruses (including those proposed as "candidate" agents for CFS) in randomly selected sera from 40 healthy laboratory personnel controls and 63 CFS patients were assayed quantitatively. The CFS patients were subsequently found to represent 23 individuals in group A and 27 individuals in group B as defined in Results. Antibody to adenovirus, Coxsackie B4, human herpesvirus 6 (HHV-6), human immunodeficiency virus (HIV), human T-cell lymphotropic viruses I/ll (HTL V I/ll), rubeola, papovavirus, and human spumavirus were assayed by indirect immunofluorescence;'9  antibody response to cytomegalovirus (CMV) was evaluated by immune adherence haemagglutination;19  and antibody to EBV early antigen (EA), viral capsid antigen (VCA), and nuclear antigen (EBNA) were measured both by indirect immunofluorescence and by anticomplement indirect immunofluorescence.19

Sample preparation and flow cytometric analysis
Lymphocyte and monocyte populations were analysed by flow cytometry with dual colour direct immunofluorescence after whole-blood lysis.2O  The panels of fluorescein isothiocyanate (FITC) or phycoerythrin (PE)  monoclonal antibodies used are as follows:
                                VOL 338: SEPT 21,1991
 

Cell subset         Antibodv specificity
CD3                     Total  T
CD4                     T helper/inducer (T H/I)
CD3 CD8*            CD8 'I' suppressor/cytotoxic T S/C)
 
 

CD8 CDll b           Suppressor  T cell
CD8 CD28             Cytotoxic T cell
CD8 CD57             T S/C subset
CD8 Leu8              T S/C subset
CDI6 CD56             Natural killer (NK)
CD20                     B cell
CD5 CD20              B cell subset
CDl4                       Monocyte
 

Activation markers
CD4 CD25           Activated T H/I
CD4 HLA-DR       Activated T H/I
CD8 CD38           Activated T S/C
CD8 HLA-DR       Activated T S/C
CDS CD25           Aclivataed T S/C
CD16 CD25         Activated NK
CDI6 HLA-DR      Activated NK
CDl4 CD25         Activated monocyte
 

Cell adhesion markers
CDlla            Adhesion molecule integrin family (LFAI-x)
CD18            Adhesion molecule integrin family (LFAI-_)
CD44            Homing recepror
CD54            Intraccllular adhesion molecule (ICAM-I)
      By convention. the FITC conjugated anlibody IS listed first and the
     PE conjugaled antibody second.

A single laser flow cytometer ('FACScan', Becton Dickinson Immuncytometry Systems, San. Jose" California),. which can discriminate Forward and side scatter as well as two fluorochromes, was used with the 'Consort 30' computer software.

Briefly, I 00 ul of whole blood was placed into tubes, monoclonal antibody added, and incubated for 15 min at room temperature. Red cells were lysed with a commercial lysing reagent ('FACsLyse', Becton Dickinson) for IO min, spun down, and washed with phosphate buffered saline containing 3% fetal calf serum and 0.1 % sodium azide. Cells were resuspended in the wash buffer and fixed with a fmal concentration of I% paraformaldehyde.  Lymphocyte gates were confirmed by the HLE FITC (CD45), LeuM3 PE (CDI4) antibody combination ('Leukogate' Becton Dickinson).  Evaluation of the cell adhesion antigens on lymphocyte, monocyte, and neutrophil populations was done by indirect immunofluorescence staining.

        Data analysis and statistics

The Mann Whitney U test was used to evaluate differences
among CDS subsets between the groups of patients, since these

        TABLE I-SYMPTOMS OF PATIENTS WITH CFS
 
     Symptoms                             Group A      Group B    Group C     Total
                                                  (n=67)          (n= 59)     (n=21)          (n=147)
Exhaustion/fatigue                        100               5             5                 70
Post exertional weakness             100             90             24               85
Arthralgia/myalgia                        100             94             25                85
Muscle weakness                          95              30             0                  56
Severe cognitive dysfunction*       92               20             0                  50
Abdominal/gastrointestinal pain     90              88             0                  76
Nausea                                         90              13              0                  46
Neuroirritability                            90              68              5                  69
Sleep disorder t                            90              42              0                  58
Twitching/myoclonus                    90              34              10                56
Frequent headache                        81              51              14                59
Balance problems                         75              78               5                 66
Depression (t)                               73              32               0                 46
Chills                                            55             20               0                  33
Sore throat                                    33             10               5                  20
Lymph-node pain                           30 14 0 I 19
        Data are percent of patients having symptoms at time of evaluatIon (see text) .
    *Short-lerm memory loss. encoding. stimulus recognition
    t Hypersomnia and hyposomnla
    (T) In most cases. onset of depression occured 6 -10 mo after onset of Illness
        See Results for definItIon of patient groups
 

            VOL 338: SEPT 21,1991     THE LANCET    709
I
markers are known to have a non-Gaussian distribution.21'   The Scheffe test was used to analyse CD8 cell subsets and NK cell populations when comparing control groups to healthy individuals. The McNemar chi square test was used to evaluate viral serology data and for probability analysis.

        Results

Clinical evaluation of CFS patients
All the patients had a history of fatigue for more than six months and post-exertional weakness substantially worse than previously observed.  Many complained of muscle weakness, myalgias, neurological disorders, frequent headaches, and other symptoms shown in table I.  At the time of evaluation, several patients had improved in clinical symptoms. About half had a medical history of a "flu-like" illness at the onset of this clinical condition. Only 8 patients had depression before the onset of illness, but depression developed in many patients after two years of illness. A full report on the clinical aspects of the patients will be published elsewhere.

After the clinical information had been analysed, we found that the patients could be placed into three groups according to their symptoms.  Group A consisted of 67 patients whose illness was so severe that they had less than 25% of their normal daily activity and also had multiple symptoms; group B were 59 individuals with reduced physical activity who continued to have moderate symptoms; and Group C consisted of 21 patients who initially had many symptoms and an incapacitating illness like patients in group A, but at the time of evaluation they had substantially improved for at least two months and regained 75% or more of their nonnal physical activity with only mild symptoms.  Healthy controls showed none of the symptoms given in table I on a persistent basis.

Viral serology
Healthy controls and CFS patients were generally similar with respect to the prevalence and titres of antibodies to CMV, EBV-VCA, EBNA, rubeola, adenovirus type 2, and the papovavirus (BK virus) (table II).  Evaluation of CFS patients by clinical status as defined in table I also showed no differences in these titres compared with controls.  No antibodies to HTL V - I/ll and HIV were found in any subject.  Antibody titres to HHV -6 in the CFS patients were twice as high (p = 0'05) as those in controls.  The prevalence of antibodies to Coxsackie B4 virus was also significantly

            TABLE II:   SURVEY OF ANTIVIRAL ANTIBODIES IN CFS PATIENTS
                             AND CONTROL SUBJECTS
                                         Geometric mean titre (% positive)
               Virus                 Healthy controls         CFS patients
                                                (n=40)                     (n=63)
    Cytomegalovirus                 42 (40)                     67 (46)
    Epstein-Barr
        VCA                             314 (95)                     393 (95)
        EA                                 25 (15)                      40* (51)
    EBNA                                206 (95)                     96^ (95)
    Rubeola                             473 (98)                    577 (89)
    HHV-6                               104 (98)                    201 t (100)
    Adenovirus                        285 (95)                    155 (81)
    Coxsackie B4                    1l3 (65)                    134 (90)
    Papovavirus BK                  83 (55)                      61 (44)
    HTLV-IIII                             (0)                             (0)
    HIV-I/HIV-2                          (0)                             (0)
    Human spumavirus              (0)                             (0)
    'p _ 0-001 (chi square); tp _ 0-05 (chi square); tp_ 0-01 (chi square)
 

        TABLE III:   PERIPHERAL BLOOD PHENOTYPIC PROFILE IN CFS
                            PATIENTS AND HEALTHY INDIVIDUALS
                                                             I       I                                Healthy
                                                              CFS                                  controls
        Cell subset                                  (n=147)                                  (n=80)

Surface markers
% CD3 cell (total T cell)                         71 (8)                                  74 (8)
% CD4 cell (T helper/inducer cell)          46(9)                                    48 (9)
CD4 cell nwnber (per ul)                      834 (315)                              889 (273)
% CD8 cell (T suppressor/cytotoxic)      24 (8)                                    22 (4)
CD8 cell nwnber (per ul)                      468 (227)                              488 (220)
% CD20 cell (B cells)                              II (16)                                  10 (3)
% CD16/CD56 (NK cell)                          13 (4)                                   14 (8)
CD4/CD8 ratio*                                      17 (0-5)                                 18 (0-4)
% CD4 cells expressing:
            CD25                                         I33 (7)                                   I34 (6)
            HLA-DR                                      12 (6)                                     10 (7)
% CD16 cells expressing:
            CD25                                            4 (1)                                     3 (2)
            HLA-DR                                         3 (I)                                     3 (I)
% CD14 cells expressing:
            CD25                                              <I                                      < 1
    Data given as mean (SD)
    *88% of CFS patients had ratios in normal range (1-2-5), 5% had decreased CDS cells (ratios>
    2-5). and 7% had increased CDS cells ratio « 1-0)

TEXT CON'T
higher in the CFS group than in controls (90% vs 65%, p=O'OOI), but the geometric mean titres (GMTs) were similar. Moreover, CFS patients had a significantly higher prevalence of antibodies to
EBV -EA than did controls (90% vs 65%, p=O.OOl), but the geometric mean titres  (GMT's)  were
similar.  Moreover, CFS patients had a significantly higher prevalence of antibodies to  EBV-EA
than did controls (51% vrs 15%' p=0.001)  This finding was specific for EA since the VCA and
EBNA seropositivity rates for both groups were identical.

Peripheral white blood cell analyses (table III)

CFS patients and healthy control subjects did not differ significantly with respect to the following indices: white blood cell count (6-10x 1O(3)/ul);  total lymphocyte, monocyte, and neutrophil populations (irrespective of whether the CFS patients were considered as an entire group or as
separate groups) (data not shown); phenotype of major lymphocyte (T, B, NK cell) populations;
percentage and absolute number of CD4 and CD8 T cells; mean CD4/CD8 ratio; presence of
activation antigens (CD25 and HLA-DR) on CD4 T cells, NK cells, or monocytes; a B-cell subset
(CD5/CD20), which we tested because it has been found to be increased in autoimmune disease;22
and percentage positive or fluorescence intensity of cell adhesion antigens (CDlla CD18, CD44, CD54) (data not shown). We also assessed CD4/CD8 ratios because previous studies have suggested that CFS patients have alterations in the distribution and ratios of these cells! Most patients had a normal range (table III).

CDB cell subset analysis
Previous cell surface marker studies in acute viral infections have shown elevation of CD8 cells that express activation antigens (CD38, HLA-DR) 23-25   With respect to the herpesviruses, these cell numbers return to normal two to four weeks after infection.  We evaluated various cell surface
antigens expressed on CD8 T cells from the 147 CFS patients and compared them (in total and as
separate groups) with the 80 healthy controls.

Three cell surface markers gave noteworthy results. In the total CFS patients evaluated, the population of CD8 cells expressing CD 11 b was decreased, but not significantly, compared with the nonnal controls (19 [SD16] vs 25 [10] )  indicating a decrea  se in the phenotypic suppressor CD8 T-cell compartment. Since the total CD8 cell count did not- - - -
 
      710   THE LANCET  VOL 338: SEPT 21. 1991

change, there was a concomitant increase in the phenotypic cytotoxric (CD8 CD28 CD II b )
population. This result was confirmed in a preliminary study in which there was an increase in the
CD8 CD28 population in these patients (data ot shown). C ompared with controls, CFS patients showed an increase in CD38 (47, [20] vrs 35 [12] ) and HLA-DR (22 [8] vrs 14 [6] )  expression.  An aditionaI activation antigen, CD25, as well as the CD 57 and Leu8, antigens , were xpressed on the CD8 celIs at a level similar to conatrols.

When the same CD8 cell subset markers were considered in group A patients, the differences
compared vith healthy controls and group C patients were saristically significant (P=0'0l; (figure). By contrast, evaluation of the CD8 cell antigen expression among the group C patients showed no significant differcnces from healthy controls (figure).  With respect ro CD8 cell abnormalities, group B patients were not significantly different from control subjects. Further evaluation of group C patients showed that only 10% of patients had tWO or more sigcificanrly abnormal results among the CD8 CDllb", CD8 CD38, or CD8 HLA-DR subsets, whereas among the group A patients 85% bad two or more abnormal results.  These data point to a high probability (90%) (p=O-OI) of having active CFS if an individual bas two or more of the CD8 cell subset alterations (McNemar chi square test with improvement fraction).

To control for possible changes in cell-surface antigen apression due: to the whole-blood lysis procedure, a subset of patients was evaluated after 'Ficoll-Hypaque' isolation of peripheral blood
mononucleau cells. No differences were seen between the whole blood lysis proceedure and
ficol.. paque purified cells (daIa, notshown).

    CD8 Cell Markers in other control $ubjects

To evaIuate the specificity of the CD8 cell aIteration in CFS patients, the control subjects with ocher clinical conditions were evaIuated with the same panel of mococIonal reaqents.  Among subjects who bad an an accute viral like illness, ( cold or flu-like illness),  there was an increase
in both percentage and absolute numbers of NK cells (CD16; CD56).  These NK ceIIs cxpressed
CDS, CD38, CDllb, but not HLA-DR as judged by multiparameter flow cymtometric studies (data
not shown).  In several individuals this initial natural kiIIer response was followed by an increase in activated CD8/CDllb-Tcells; (CD38 HLA\-DR).  In all these subjects, recovery one to two weeks later was accompanied by a return to normal or all these immunological indices. All other CD8
cell markers were normal in these mdividuals.
   Evaluation of CD8 cell subsets in control subjects with a diagnosis of depression sbowed no  significant differences compared with healthy controls. Furthermore,  family members  and
contacts of CFS patients bad normal CD8 cell subsets as did individuals presenting with farigue
other than that associated with CFS (table IV)  Laboratory findings among CFS patients have shown low level autoantibodies which may reflect an underlying autoimmune disease. (36)
Evaluation of patients with SLE showed only moderately increased expression of the CD 38 marker
on CD8 cells (table iv; other cell surface markers were within the normanl range.

                            Discussion

We have not  found any serological evidence of m aetiological association with CFS of  the various human viruses that we tested Although seroprevalence to Cuxsackie B4 virus was higher in CFS patients thm in controls, the GMT was similar.  Moreover, other evidence bas not supported a role of this virus in CFS.(10)    Anti-EA\ antibody prevalence was also significantly increasced in the CFS patients who in this Study were not preseIected for EBV serology. Additionally, like other investigators, (6,8), we found a relatively high GMT for EBV-EA and HHV-6 antibodies; and we also recorded a significantly lower GMT for EBV-EBNA in our cohort of CFS patients.  'These findings
most probably reflect T-celI  disorders in these patients. (29)
   Immunological testing of CFS patients has resulted in confliciting reports about abnormalities in
T-cell subsets, mitogenic responses, and cyrokine production.(28,31).  Also,

    TABLE iv: CD8 CELL SURFACE MARKERS AND NK CELLS IN CONTROL POPULATIONS
         OMMITTED this table
 
        VOL 338: SEPT 21,1991  THE LANCET  711

changes in B-cell immunity associated with autoantibody production, high antiviral antibody titres, circulating immune complexes, and increased numbers of CD5/CD20 B cells(26),  have been noted,
as have decreased function of NK cells and changes in NK cell numbers(14.16.29).  Finally, monocytes from CFS patients have been reported to have decreased HLA-DR cell-surface antigen expression and depressed phagocytic activity (32).

    Our study centred on the use of flow cytometry to defme immunological alterations in patients
with CFS. Our CFS patients had a female to male ratio of 3.9/1 and a characteristic median age for CFS of 38 consistent with other studies!  That most of our CFS patients had a normal percentage and number of CD4 and CD8 cells, also accords with others (10).  When all CFS patients were considered, we found a state of immune activation specifically among the CD8 lymphocyte population. Moreover, the suppressor subset of CD8 (CD11b) was reduced in many patients, significantly so in patients with multiple symptoms and severe incapacitating illness (group A). These results agree with those reported by Klimas et al (16),  who demonstrated in 30 patients an
elevation in CD8 cells expressing related markers (HLA-DR and CDw26).  In that study, however,
patients were preselected for high EBV titres.
 
Unlike in CFS, an elevation of cells expressing CDllb or CD57, usually NK cells, is a common finding in acute viral illness(33).  Our data confirm this observation; within one to two weeks, the immunological indices in our subjects with acute viral illness returned to normal. Our fmdings suggest that in CFS patients, the NK cell number is normal, the CD8 CD11b population is reduced, and the CD38 and HLA-DR markers remain persistently raised.

    The reason for this immune activation is not known, but we have seen it in individuals who have had symptoms for one to five years. Moreover, the abnormalities have persisted in 6 patients followed for more than a year.  In 2 other cases, an improvement in symptoms was accompanied by a return to normal of some immunological markers (data not shown). Our results with CFS patients whose symptoms have improved (group C) support these observations. The immune disorder in CFS does not seem to reflect an underlying autoimmune disease or depression (table IV).  A possible explanation is that the NK dysfunction,(16) , or the decreased CD11b CD8 suppressor cell population (our findings), leads to a persistent hyperimmune response of the
remaining CD8 cells.  The activation might lead to an outpouring of cellular products and cytokines (eg, interferon, tumour necrosis factor, interleukin-l),  that are characteristically associated with myalgia, fatigue, neurological signs, and other signs and symptoms associated with acute viral
infections(15).  Unless the immune system is brought back into balance, however, this chronic activation affects the individual further and might eventually lead to other clinical illnesses.

    Although not diagnostic of CFS and not observed in all individuals with a history of CFS, the immunophenotypic data presented here indicate that many individuals with symptoms of CFS have
CD8 cell immune activation.  Additionally, many have a reduced level of CD8 cells with a suppressor  phenotype (CD 11 b) and an increase in CD8 cells of the cytotoxic type (CD11b_).
Further functional studies of these CD8 cell subsets may provide insights into CFS pathogenesis. Most noteworthy is the statistical evidence that an individual with two or more of the CD8 cell subset alterations (increased CDllb_, CD38, and HLA-DR) has a high probability (90%) of having active CFS. These findings are consistent with chronic stimulation of the immune system perhaps
by a virus, although evidence for a common agent has not been found.'

We thank Dr Noel Warner, Dr David Buck, and Becton Dickinson Corporation (San Jose, California) for the monoclonal antibody reagents and FACScan flow cytometer and Ms Connie Ethridge of AMAC Corporation for the cell adhesion antibodies. We also thank the Chronic Fatigue Immune Dysfunction Foundation of San Francisco for its support. A major part of this work was conducted while A. L was on sabbatical leave in the laboratory of J. A. L; A. L was supported by the Lloyd Frye Foundation. We thank Dr Philip Loh for the human spumavirus and positive control sera; Ms Sue Fujimara and Ms Karen Farrington for their careful technical assistance; Ms Pat McDonald for help with the patients; Ms Lisa Willielm for her excellent secretarial assistance; and Dr Ira Tager, Dr Jim Wiley, and Dr George Rutherford for helpful discussions.

REFERENCES
1. Straus S.
    History of chronic fatigue syndrome.
    Rev Infecr Dis 1991; 13  (suppll): S2-S7.

2. Komaroff A, Buchwald D.
    Symptoms and signs of chronic fatigue syndrome.
    Rev Infect Disease 1991; 13 (suppll): S8-SII.

3. AltayHT, TonerBB, Brooker H, Abbey SE, Salit IE, Garfmkel P.
    The neuropsychological dimensions of postinfectious neuromyasthenia (chronic fatigue
    Syndrome):  a preliminary report.
    Int J Psycln'alry Med 1990; 20: 141--49.

4. Holmes GP, Kaplan JE, Gantz NM, et al.
    Chronic fatigue syndrome: a working case defmition.
    Ann Intern Med 1988; 108: 387-89.

5. Tobi M, Marag A, Ravid Z, et al.
    Prolonged atypical illness associated with serological evidence of persistent
    Epstein-Barr virus infection.
    Lanai 1982; i: 61-64.

6. Jones JF, Ray CG, Minnich LL, Hick MJ, Kibler R, Lucus DO.
    Evidence for active Epstein_Barr virus infection in patients with persistent unexplained
    illnesses; elevated anti_early antigen antibodies.
    Ann Inlern Med 1985; 102: 1-7.

7. Hellinger WC, Smith TF, Van Scoy RE, Spidwr PG, Forgacs P, Edson RS.
    Chronic fatigue syndrome and diagnostic utility of antibody to Epstein Barr virus early Antigen.
    JAMA 1988; 260: 971-73.

8. Levy JA, Greenspan D, Ferro F, Leonette ET.
    Frequent isolation of HHV-6 from saliva and high seroprevalence of the virus in the population.
    Lanai 1990; 335: 1047-50.

9. Archard LC, Bowles NE, Behan PO, Bell EJ,DoyleD.
    Post viral fatigue syndrome: persistence of enterovirus RNA in muscle and
    elevated creatinine kinase.
   J R Soc Med 1988; 81: 326-29.

10. Miller NA, Carmichael HA, Hall FC, et al.
    Antibody to Coxsackie B virus in diagnosing postviral fatigue syndrome.
    Br MedJ 1991; 302: 140-43.

II. DeFreitas E, Hilliard B, Cheney P, et a!.
    Rerroviral sequences related to HTL VII in patients with chronic fatigue
    immune dysfunction syndrome.
    Proc Nail Acad Sci 1991; 88: 2922-26.

12. KiblerR, LucasDD, Hicks MJ, Poulos BT, JonesJF.
    Immune function in chronic active Epstein-Barr virus infection.
    J Clin Immuno11985; 5: 46-54.

13. Tosato G, Straus S, Henle W, Pike SE, Biaese RM.
    Characteristic T cell dysfunction in patients with chronic active Epstein_Barr
    virus infection  (chronic infectious mononucleosis).
    J ImmunfJ11985; 134: 3082-88.

14. Caligiuri M, Murray C, Buchwald D, et al.
    Phenotypic and functional deficiency of natural killer cells in patients
    with chronic fatigue syndrome.
    J ImmunfJ11987; 139: 3306-13.

15. Lloyd AR, Wakefield D, Boughton CR, Dwyer JM.
    Immunologic abnonnalities in the chronic fatigue syndrome.
    MedJ Aus11989; 151: 122-24.

16. Klimas NG, Salvato FR, Morgan R, Fletcher MA.
    Immunologic abnormalities in chronic fatigue syndrome.
    J Clin MicromoI 1990; 28: 1403--10.

17. Notkins AL, Mergenhagen SE, Howard RJ.
    Effect of viral infections on the function of the immune system.
    Annu Rev Microbiol 1970; 24: 525-40.

18. Rouse BT, Horohov DW.
    Immunosuppression in viral infections.
    Rev Infeci DiJ 1986; 8: 850-73.
19. Lennette ET, Ward E, Henle G, Henle W.
    Detection 'If antibodies to Epstein-Barr virus capsid antigen by immune
    adherence hemagglutination.
    J Clin Micromo11982; 15: 69-73.

20. Coon J, Landay A, Weinstein R. Advances in flow cytomerry for diagnostic pathology.
    Lab Inves11987; IS: 453--79.

21. Siegel S.
    Nonparametric statistics for the behavioral sciences.
    New York: McGraw-Hill Book Co, 1956.

22. Casali P, Burastero SE, Nakamura M, Inghirani G, Notkins AL.
    Human lymphocytes making rheumatoid factor and antibody to single structured
    DNA belongs to Leu I' B cell subset.
    Science 1987; 236: 77-81.

712
THE LANCET
VOL 338: SEPT 21,1991

23. Carney WP, Rubin RH, Hoffman RA, Hansen WP, Henley K, Hirsch MS.
    Analysis of T lymphocyte subsets in cytomegalovirus mononucleosis.
    J ImmUl1lJ11981; 126: 2114-16.

24. Tomkinson BE, Wagner D, Nelson DL, Sullivan SL.
    Activated lymphocytes during acute Epstein-Barr virus infection.
    J Immunol 1987; 139: 3802-D7.

25. Landay AL, Mackewicz C, Levy JA.
    Flow cytometric evaluation of CD8- cell subsets in HIV-infected individuals.
    Sixth International Conference on AIDS, San Francisco, California, 1990: 141.

26. Buchwald D, Komaroff A.
    Review of the laboratory findings for patients with chronic fatigue syndrome.
    Rev Infect Dis 1991; 13: 512-18.

27. Henle W, Henle G.
    Epstein-Barr virus specific serology in immunologically compromised individuals.
    Cancer Res 1981; 41: 4222-25.

28. Behan PO, Behan WM, Bell EJ.
    The postviral fatigue syndrome: ananalysis of the findings in 50 cases.
    J Infect Dis 1985; 10: 211-22.

29. Aoki T, Usada Y, MiyakoshiH,TamuraK,Herberrnan RB.
    Lownatutal killer syndrome: clinical and immunologic features.
    Nat Immune Cell Growth Regu11987; 6: 11&-28.

30. Borysiewicz LK, Haworth SJ, Cohen J, Mundin J, Rickinson A, Sissons JGP.
    Epstein Barr virus specific immune defects in patients with persistent symptoms following
    infectious mononucleosis.
    Q J Med New Set 1986; 58: 111-21.

31. Straus SE, Dale JK, Peter ]B, Dinarello CA.
    Circulating Iymphokine levels in the chronic fatigue syndrome.
    J Infect [)is 1989; 160: 1085-86.

32. PrietO J, Subira L, Castilla A, Serrano M.
    Naloxone reversible monocyte dysfunction in patients with chronic fatigue syndrome.
    Scand J Immuno11989; 30: 13--20.

33. Gratama JW, Naipal A, OostervL'Cr M, et al.
    Effects of herpcs virus carrier  status on peripheral T lymphocyte subsets.
    Blood 1987; 70: 51&-23.

                                                    END