Q Fever Serology: Cutoff Determination for ... · Q FEVER SEROLOGY 191 TABLE 1. Results for various cutoff titers ofantibodies against phase II C. burnetii in the diagnosis ofactive

CLINICAL AND DIAGNOSTIC LABORATORY IMMUNOLOGY, Mar. 1994, p. 189-196 Vol. 1, No. 21071-412X/94/$04.00+0Copyright (C 1994, American Society for Microbiology

Q Fever Serology: Cutoff Determination forMicroimmunofluorescence

HERVE TISSOT DUPONT,"2 XAVIER THIRION,2 AND DIDIER RAOULTI*Unite des Rickettsies, CNRS EP J 0054,1 and Laboratoire de Sante Publique,2

Faculte de Medecine, Marseille, France

Received 21 October 1993/Returned for modification 29 November 1993/Accepted 13 December 1993

Q fever, a worldwide zoonosis caused by Coxiella burnetii, lacks clinical specificity and may present as acuteor chronic disease. Because of this polymorphism, serological confirmation is necessary to assess the diagnosis.Although microimmunofluorescence is our reference technique, the cutoff titers that are currently used to makea diagnosis of active or chronic Q fever were determined years ago with limited series of patients and sera. Wedetermined the titers of immunoglobulin G (IgG), IgM, and IgA against both phases (I and II) of Coxiellaburnetii. Rheumatoid factor was removed before testing IgM and IgA. We report here the various cutoff titersand the kinetics of antibody development from 2,218 first serum samples of patients, among whom 208 sufferedfrom acute Q fever and 53 had chronic Q fever. In active Q fever, we have defined a low cutoff (phase II IgGtiter .100) below which the diagnosis cannot be made and would need further confirmation and confirmed ahigh cutoff (phase II IgG titer .200 and phase II IgM titer .50) over which the diagnosis can be made. Forchronic Q fever diagnosis, phase I IgA titers are not contributive despite previous works claiming theirusefulness; a phase I IgG titer of .800 is highly predictive (98%) and sensitive (100%). We have also studiedthe possibility of rejecting or evoking the diagnosis of chronic Q fever by using phase II IgG and IgA titers. Thismethod is useful when phase I testing is not available, but the sensitivity remains low (57%).

Q fever is a worldwide zoonosis caused by the rickettsiaCoxiella burnetii, an obligate intracellular organism which livesin phagolysosomes of the host cell. The main characteristic ofQ fever is its clinical polymorphism. In acute cases, the mostcommon clinical syndromes are self-limited febrile illnesses ofunknown origin, granulomatous hepatitis, pneumonia, andmeningoencephalitis (23). Moreover, cases of febrile eruption,myocarditis, and pericarditis have been reported. In chroniccases, endocarditis is the main syndrome (3). Osteomyelitis,infections of vascular grafts, or aneurisms and infectionsduring pregnancy were also reported. Because of such variousclinical presentations, serologic confirmation is required forthe diagnosis of Q fever. Currently used serological methodsare indirect microimmunofluorescence, complement fixation,and enzyme immunoassay.Unique to C. burnetii is its antigenic phase variation (6, 22).

The virulent phase I is isolated from natural or laboratoryinfections of animals and humans, whereas the avirulent phaseII occurs during serial passages of C. burnetii in immunoin-competent hosts, such as embryonated hen eggs or cell culturesystems. Phase I to II variation seems to correlate with smoothand rough lipopolysaccharide changes (6, 24). Although onlyphase I organisms are virulent for humans, anti-phase IIantibodies are the first to appear. Specific high levels ofanti-phase I antibodies are normally found in chronic Q feverpatients, whereas specific anti-phase II antibodies predominateduring acute Q fever. As with many other infectious diseases,immunoglobulin M (IgM) antibodies are the first to appear.Usually they are detectable for a few weeks or, at most, for afew months (6, 13). IgG appears somewhat later but can persist

* Corresponding author. Mailing address: Unite des Rickettsies,CNRS EP J 0054, Faculte de Medecine, 27, boulevard Jean Moulin,13385 Marseille Cedex 5, France. Phone: (33) 91 38 55 17. Fax: (33) 9183 03 90.

for years, even for life (6). The presence of IgA to phase Iantigen is strongly correlated with endocarditis (15, 16, 25).Although microimmunofluorescence is our reference tech-nique, it becomes positive only 7 to 15 days after the onset ofclinical symptoms and a definite diagnosis can be confirmedonly by seroconversion or a fourfold increase in antibody titer.When serologic confirmation is required for diagnosis, theabsence of cross-reactions is important for specificity. For C.burnetii immunofluorescence, the only cross-reaction that hasbeen suggested was with Legionella pneumophila. However, wehave recently demonstrated the absence of such a cross-reaction (10). In order to shorten the delay before diagnosis,thus allowing earlier antibiotic treatment, we studied the valueof various cutoff titer combinations in the diagnosis of evolvingQ fever (acute or chronic) and in the diagnosis of chronic Qfever.

MATERIALS AND METHODS

Serum collection and selection. Our laboratory, located inMarseille (southern France), is the French National ReferenceCenter for Rickettsial Diseases, which performs most of the Qfever serologic testing for the area of Marseille and receivessera from other parts of France, often for seroconfirmation.Each time a serologic result is consistent with Q fever, thephysicians are asked to provide further sera, blood, or anytissue for culture, epidemiological, and clinical data.

Since 1991, our results have been computerized in a database. The present study was conducted using 12,428 serumsamples tested for Q fever at the National Reference Centerbetween January 1991 and May 1993. From these, we selectedthe 2,218 patients who had paired sera and whose clinicalevolution of disease was known.

Serologic procedures. (i) Antigen preparation. Phase II C.burnetii Nine Mile (ATCC VR 615) was fgrown in confluentlayers of L929 mouse fibroblasts in 150-cm culture flasks with

189

on February 1, 2021 by guest

http://cvi.asm.org/

Dow

nloaded from

190 TISSOT DUPONT ET AL.

minimum essential medium supplemented with 2 mM L-glutamine and 4% fetal bovine serum. The infection wasmonitored by microscopy of Gimenez-stained cells scrapedfrom the bottoms of the flasks (11). When a heavy rickettsialinfection was seen, the supernatants of 15 flasks were individ-ually pelleted by centrifugation (5,000 x g, 15 min) andresuspended in 1 ml of phosphate-buffered saline, pH 7.3(PBS), with 0.1% formaldehyde. All further steps were con-ducted under cooling conditions. After pooling, the remainingcells were broken by sonication. Cellular debris were removedby two successive centrifugations (100 x g, 10 min each). The15-ml suspension was then centrifuged through 20 ml of PBSwith 25% sucrose (6,000 x g, 30 min, without break). Theresulting pellet was washed three times in PBS (6,000 x g, 10min), resuspended in the smallest possible volume of steriledistilled water, and adjusted to 2 mg/ml by UV spectroscopy.Antigen prepared in this manner was frozen at - 20°C forfurther immunofluorescence tests.

In order to reactivate phase I C. burnetii, mice were inocu-lated with phase II C. burnetii Nine Mile. Ten days afterinfection, each spleen was removed, ground in 7.5 ml ofminimum essential medium, and inoculated into three 75-cm2culture flasks containing L929 cell monolayers (2.5 ml perflask). The infected cells were then harvested and the bacteriawere purified by the same technique as described above. Allmanipulations of live C. burnetii were conducted in a biohazardsafety laboratory (P3).

(ii) Microimmunofluorescence. The two antigens, preparedas described above, were applied with a drawing pen at the twopoles of each well of 30-well microscope slides (DynatechLaboratories Ltd., Billingshurst, United Kingdom), air dried,and fixed in acetone for 10 min. Sera were serially diluted(twofold dilutions initially ranging from 1:25 to 1:1,600 andmore if needed) in PBS with 3% nonfat powdered milk. Theoverlaid antigens were incubated in a moist chamber for 30min at 37°C and then washed three times in PBS for 10 mineach time. After air drying, the complex was overlaid withfluorescein isothiocyanate-conjugated goat anti-human IgG(-y) (Fluoline G; Biomerieux, Marcy l'Etoile, France), anti-human IgM (Fluoline M; Biomerieux), or anti-human IgA(Fluoline A; Biomerieux), diluted at 1:300. Incubation, washes,and drying were performed again as described above. Theslides were mounted in buffered glycerol (Fluoprep; Bio-merieux) and read with a Zeiss fluorescence microscope at ax400 magnification. In order to remove IgG, rheumatoidfactor adsorbant (Behring, Mannheim, Germany) was usedprior to IgM and IgA determination.Case definition. (i) Acute infection. Patients were consid-

ered to have acute Q fever when they seroconverted orpresented a fourfold increase in specific antibodies titers witha suggestive clinical background, including fever of unknownorigin, flu-like syndrome, granulomatous hepatitis, pneumonia,or meningoencephalitis. In a few cases, the isolation of C.burnetii from blood or other tissue provided a definite micro-biological diagnosis.

(ii) Chronic infection. A diagnosis of chronic Q fever wasmade when a patient had a high antibody titer (.800) (15)associated with a chronic infectious syndrome (such as endo-carditis, valvular or vascular prosthesis infection, or boneinfection) without any other clinical or microbiological expla-nation. Isolation of C. burnetii from blood or other tissue wasalso diagnostic.

(iii) Non-Q fever. We included in the definition of non-Qfever the two following groups of patients: those with lateserologic evolution and those with negative serology. In theformer group, 0 fever serology was considered a sequela of

previous infection when the patient did not show clinicalsymptoms consistent with Q fever yet presented with antibod-ies against C. burnetii at low but constant titers. For the lattergroup, we considered the serology negative when patients hadno detectable antibodies in at least two consecutive serumsamples, irrespective of their clinical presentation.

Statistical methods. (i) Serum selection. For all the patientswhose diagnosis was known (acute Q fever, chronic Q fever, ornon-Q fever), we considered only the titer of the first serumsamples. Serum selections and counts were then performed byusing Epi-Info 5.01 software (1).

(ii) Quality indicators. Four usual indicators have beencalculated for various cutoff titers of antibodies against variousantigens: sensitivity [true positive/(true positive + false nega-tive)], specificity [true negative/(false positive + true nega-tive)], predictive value of a positive result (PPV) [true positive/(true positive + false positive)], and predictive value of anegative result (NPV) [true negative/(false negative + truenegative)] (5).

(iii) Kinetics. In order to study the kinetics of antibodydevelopment in acute Q fever, we selected the 58 patients witha confirmed diagnosis of acute Q fever and for whom thephysicians had provided the date of onset of the clinicalsymptoms. We considered all available serologic results forthese patients. This interval was defined as the time (days orweeks) between the onset of clinical signs and the date of bloodsampling for serology. For groups at each interval, using anonlinear scale (due to the decreasing number of serumsamples with the increase of delay), we calculated the meantiter for each antigen.

Additionally, the kinetics were studied in a manner, accord-ing to the method of Worswick and Marmion (25), in which foreach group of intervals, the percentage of patients whoreached a cutoff titer (after determination of this cutoff), i.e.,the kinetics of the sensitivity, was calculated.

(iv) Factorial analysis. In order to study the relationsbetween the three different clinical presentations and theserologic response, we have carried out a multiple correspon-dance analysis (MCA) (2, 14, 19) which gives a simultaneousmultidimensional representation of the patients and the sero-logic results.The variables that we considered were the clinical presenta-

tion and the titers of each class of antibodies (IgG, IgM, andIgA) against both phases of C. burnetii. The modalities wereacute, chronic, or inactive for the clinical presentation and 0,25, 50, 100, 200, 400, 800, and : 1,600 for the antibody titers.The first step consisted of changing the initial data into asymmetric partitioned matrix of crosstabulations among thecategorical variables (Burt table) (4), i.e., each modalityfor each variable was coded as 1 when present or 0 whenabsent, leading to a matrix of 51 modalities by 1,079 serumsamples.The MCA is used to find a low-dimensional graphical

representation of the association between rows and columns ofthe table. Each row and column is represented by a point in aEuclidian space determined from cell frequencies. This MCAwas carried out by using SAS software (20). We have com-pleted the MCA by calculating indices to assess the quality ofthe representations, including absolute and relative contribu-tions to the axes, which permit, respectively, the identity of theheaviest modalities in building the axes and the measure of thequality of the representation of each modality in the system ofaxes which was considered. This was carried out by using SASand ANADEV, software especially developed at the PublicHealth Laboratory of Faculty of Medicine of Marseille forusing factor analysis in clinical research.

CLIN. DIAGN. LAB. IMMUNOL.

on February 1, 2021 by guest

http://cvi.asm.org/

Dow

nloaded from

Q FEVER SEROLOGY 191

TABLE 1. Results for various cutoff titers of antibodies againstphase II C. burnetii in the diagnosis of active

(acute or chronic) Q fever

Phase II cutoff Sensitivity Specificity PPV NPVtiter(s) (%) (%) (%) (%)

IgG 2 100 86.0 92.9 59.0 96.6IgG 2 200 81.9 96.1 70.8 95.8IgM 2 50 67.2 98.8 88.3 95.7IgG 2 100 and 61.5 99.7 96.2 94.7IgM 2 50

IgG 2 200 and 58.4 100.0 100.0 94.0IgM 2 50

Groups of patients and sera. (i) Predictive values. The PPVand NPV, respectively, were calculated among groups ofpositive and negative sera, and they depended on the preva-lence of the disease. In such conditions, we used 2,218 firstserum samples of patients, among whom 208 suffered fromacute Q fever and 53 had chronic Q fever.

(ii) Sensitivity and specificity. Sensitivity and specificity areintrinsic properties of tests in relationship to detecting disease.They were calculated among groups of diseased and nondis-eased people, respectively, and did not depend on the preva-lence of the disease. Under these conditions we have enlargedthe sample size of the diseased groups, using all available caseswhich were in our files. Thus, the total number of patients was2,594, with 486 acute Q fever patients, 148 chronic patients,and 1,960 patients without active Q fever.

RESULTS

Quality indicators. (i) Diagnosis of active Q fever. Theindicators of active Q fever are summarized in Table 1. For thediagnosis of active (i.e., acute or chronic) Q fever, we haveconsidered anti-phase II C. burnetii IgG and IgM. The sensi-tivity, specificity, PPV, and NPV have been calculated forvarious cutoff titers of these antibodies, independently and incombination. The highest sensitivity and NPV (86.0 and96.6%, respectively) were obtained with an IgG titer of .100,decreased when the titer increased or when IgM was consid-ered, and reached 58.4 and 94.0%, respectively, for the com-bination of an IgG titer of .200 and an IgM titer of -50.However, although specificity and the PPV were as low as 92.9and 59.0% for an IgG titer of .100, respectively, they reached100% for the combination of an IgG titer of .200 and an IgMtiter of .50.

(ii) Diagnosis of chronic Q fever. The indicators of chronicQ fever are summarized in Table 2. For the diagnosis ofchronic Q fever, we have considered the IgG and IgA phase Iand phase II C. burnetii antibody titers.With phase I antigen, the highest sensitivity and NPV were

obtained with an IgG titer of -800 (100% for both), an IgAtiter of .25 (96.6 and 100%, respectively), and a combinationof these two cutoff titers (96.6 and 100%, respectively). Spec-ificity reached 99.8% for IgG titers of .1,600, 99.9% for IgAtiters of .400, and 100% for IgG titers of .800 and IgA titersof -25. However, the PPV was 100% for IgG titers of .1,600,97.4% for IgA titers of .400, 98.1% for IgG titers of .800 andIgA titers of >-25, and 100% for IgG titers of -1,600 and IgAtiters of .400. However, it is remarkable that with IgG titers of.1,600 alone, although the sensitivity was only 80.4%, thespecificity was 99.8%, the PPV was 100%, and the NPV was99.6%.With phase II antigen, the sensitivity and NPV were com-

TABLE 2. Results for various cutoff titers of antibodies againstphase I and phase II C. burnetii in the diagnosis of chronic Q fever

Cutoff titer(s) Phase Sensitivity Specificity PPV NPV(%) (% (% (%

IgG 2 800 I 100.0 99.6 98.1 100.0II 100.0 93.4 53.5 100.0

IgG 2 1,600 I 80.4 99.8 100.0 99.6II 88.5 96.4 71.4 99.9

IgA 2 25 I 96.6 96.2 48.6 100.0II 96.6 87.0 23.8 100.0

IgA 2 50 I 94.6 98.0 63.9 100.0II 93.2 91.9 34.0 100.0

IgA 2100 I 85.8 98.9 77.8 99.8II 83.1 95.8 52.1 99.9

IgA 2 200 I 71.6 99.4 83.3 99.6II 72.3 97.9 66.2 99.8

IgA 2 400 I 56.8 99.9 97.4 99.3II 59.5 99.0 80.0 99.4

IgG 2 800 and I 96.6 100.0 98.1 100.0IgA 2 25 II 96.6 95.3 57.6 100.0

IgG 2 800 and I 85.8 100.0 98.0 99.8IgA 2 100 II 83.1 97.8 74.6 99.9

IgG 2 800 and I 56.8 100.0 97.4 99.3IgA 2 400 II 59.5 99.3 87.0 99.4

IgG 2 1,600 and I 52.7 100.0 100.0 99.3IgA 2 400 II 56.8 99.5 95.0 99.3

parable. Although the specificity was lower than with phase Iantigens, it reached 96.4% for IgG titers of .1,600, 99% forIgA titers of .400, and 99.5% for IgG titers of .1,600 and IgAtiters of .400. The PPV was 71.4% for IgG titers of .1,600,80% for IgA titers of .400, and 95% for IgG titers of .1,600and IgA titers of .400.

Kinetics. The results of the kinetics study of acute Q fever,in terms of mean titers, have been separated in two figures.

Figure 1 shows the kinetics for the first 8 weeks after theonset of clinical signs. Phase II IgG and IgM reached thedetection level (titer = 25) at week 2 and increased concur-rently. Phase II IgM reached the diagnostic level (titer = 50)between weeks 2 and 3, and the titer remained below 200.Phase II IgG titers were over 200 after week 4 and remainedaround 800. Phase II IgA became detectable between weeks 2and 3, and the titer remained below 50. Phase II IgM showedthe same profile, and the titer remained around 25. Phase IIgG and IgA did not reach detectable levels.

Late in the evolution of acute Q fever (Fig. 2), phase II IgGtiters remained between 800 and 1,600 until week 48. Phase IIIgM titers decreased quickly between weeks 10 and 15 andremained around 50 until week 48. Phase I IgG titers increasedbetween weeks 8 and 25 and remained around 100 until week48. Phase II IgA titers were still around 50 at week 48, whereasphase I IgM and IgA titers were around the detection level.

Figure 3 represents the kinetics of the percentage of patientswho have reached the low cutoff (i.e., IgG phase II titer .100)and the percentage of those who have reached the high cutoff(i.e., IgG phase II titer .200 and IgM phase II titer .50),during 45 days after the onset of the disease. Seventy percentof the patients reach the low cutoff during the third week, and100% reach it after 1.5 months. For the high cutoff, themaximum (75% of patients) was reached during week 6 andremained stable.

Factorial analysis. The total number of axes, calculated bythe MCA, was 44 (51 modalities - 7 variables). We reporthere the results of the projection following the two mostrepresentative axes. We first studied the representation qualityindices.

VOL. 1, 1994

on February 1, 2021 by guest

http://cvi.asm.org/

Dow

nloaded from

192 TISSOT DUPONT ET AL. CLIN. DIAGN. LAB. IMMUNOL.

, IgG II

1go,IgM II

- IgA II

.... IgM I

,, IgA I-IgG I

1 2 3 4 5 6 7 8Weeks

FIG. 1. Kinetics of acute 0 fever serology during the first 8 weeks after the onset of clinical signs. I and II refer to phase.

(i) Absolute contributions. The following modalities contrib-ute mostly to the first factorial axis: chronic (12.9%), phase IIgG titer of -1,600 (12.5%), phase II IgG titer of >1,600(10.1%), phase II IgA titer of -1,600 (7.8%), phase I IgAtiter of .1,600 (7.6%), and phase II IgM titer of .1,600(5.5%).The following modalities contribute mostly to the second

factorial axis: acute (7.8%), inactive (6.4%), phase II IgM titerof .1,600 (7.9%), phase I IgA titer of .1,600 (5.6%), phase IIgM titer of -1,600 (5.5%), phase II IgG titer of 25 (5.5%),and phase II IgA titer of .1,600 (5.3%).

(ii) Relative contributions. All the studied modalitiesshowed relative contributions to the studied axes high enoughto allow analysis of the graphic closeness of these modalities.

(iii) Scattergram description. The scattergram (Fig. 4)showed a remarkable inverted-V shape, with a regular distri-bution of the modalities on the first axis, opposing low andelevated antibody titers. The second axis opposed the modal-ities phase II IgG titer of 25 and phase I IgA titer of 25 at a lowlevel of axis one, whereas at a high level, it individualized thespecial positions of elevated IgM and IgA titers in both phases.

Inactive patients were associated with negative phase I,

- IgG II

IgG I..IgA II@@-|@u@@u-||-u""Me#s"""""__..........1...IAII

,,,,, ....... ...... I11.0% gM IIIgA I..

....

__------- _---- IgM I

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48

Weeks

FIG. 2. Kinetics of acute Q fever serology during 48 weeks after the onset of clinical signs. I and II refer to phase.

on February 1, 2021 by guest

http://cvi.asm.org/

Dow

nloaded from

Q FEVER SEROLOGY 193

14

o0 5 10 15 20 25 30 35 40 45

Days

FIG. 3. Kinetics of the sensitivity and NPV of phase II IgG titers of .100 and kinetics of the sensitivity of phase II IgG titers of .200 and phaseII IgM titers of .50 in the diagnosis of acute Q fever, during 45 days after the onset of clinical signs.

negative phase II IgA, phase II IgG titers of up to 100, andphase 11 IgM titers of up to 50.Acute Q fever was associated with phase II IgG titers

between 200 and 800, phase 11 IgM titers of up to 400, phaseII IgA titers of up to 50, phase I IgG titers of up to 400, andphase I IgM and phase I IgA titers of up to 25.

Chronic Q fever was clearly associated with phase I andphase I IgG titers of -1,600, phase I IgA and IgM titers of.400, and phase II IgA titers of .800. The highest titers ofIgM and IgA in both phases were especially individualized,corresponding to late sera of chronic Q fever, mostly receivedfor confirmation.

DISCUSSION

Because of the low clinical and biological specificity of Qfever, laboratory diagnosis is usually based on serological tests,of which microimmunofluorescence is our reference techniqueand is also currently used by other laboratories in the UnitedStates (6, 15), Canada (15), Australia (9, 13, 25), Switzerland(6), and Spain (21). This test is economical and requires smallamounts of antigen (18). The technique permits one to distin-guish among the antibody classes as well as among their phaseI and phase II specificities.Antigen preparation. C. burnetii antigen is commercially

available in France (phase II; Nine Mile and Henzerlingstrains; Biomerieux). As the National Reference Center, weoften receive sera that have been tested in other laboratories inFrance, for confirmation and phase study. Discrepancies arevery often observed, mostly in sera found positive with thecommercial antigen and negative in our hands. This is probablybecause our highly purified preparations yield antigen free ofcellular debris, whereas the commercial antigen is not purified.

Diagnosis of active Q fever. When physicians entrust abiologist with patients' sera, they wait for a reliable answer in

terms of PPV and NPV. Therefore, Table I leads us to definetwo different cutoff titers.The low cutoff should give a high NPV, so that below this

titer we can reject the diagnosis with a high probability. Wemay now consider a phase II IgG titer of 100 as this low cutoff.Under these conditions, the NPV is 96.6% and the specificityis 92.9%.To the contrary, the high cutoff must emphasize the PPV in

order to assess the diagnosis with a high probability. In ourhands, the combination of a phase II IgG titer of -200 andphase II IgM titer of 250 gives a PPV of 100%. Under theseconditions, in a patient with clinical signs and in a compatibleepidemiological context, we can state positively the diagnosisof active Q fever, without need of paired sera. Between thesetwo cutoff titers exists a grey zone which requires paired sera toconfirm the diagnosis.These findings confirm by using a large sample size the high

cutoff which had already been established. This is in agreementwith the results of Peacock et al. (15), Hunt et al. (13), andField et al. (9), who have shown the presence of phase II IgGand IgM in acute cases. However, Edlinger (7) did not detectIgM in acute cases.

Kinetics of antibodies in acute Q fever. In acute Q fever, wehave shown that phase 11 IgM increases quickly until week 4and decreases after week 8. Titers remain around 50 at leastuntil week 48, which was our longest follow-up. For phase IIgM, we obtained the same profile with mean titers below 25.These findings agree with these of Dupuis et al. (6) in terms ofintervals, but with lower titers (mean = 640 at week 4 for phase11 IgM and 160 for phase I 1gM in the Dupuis study). As forphase II IgG, we have shown a rapid increase, strictly parallelto that of phase 11 IgM, until week 6, yielding a constant titeraround 800 at least until week 48. Phase I IgG becomesdetectable at week 10, and the titer reaches a constant levelaround 100 at week 24. In the study of Dupuis et al. (6), phaseII IgG showed a comparable profile, whereas phase I IgG did

VOL. 1, 1 994

on February 1, 2021 by guest

http://cvi.asm.org/

Dow

nloaded from

194 TISSOT DUPONT ET AL.

9U800 Axis 1

i0 |CHRONIC*

50o fo * 1600

T~IVE25f

*Phase IgG

*Phase I IgM

OPhase I IgA |1600

* Phase 11 IgG 21600*Phase 11 IgM

Axis 2 DPhase 11 IgA * 1600

FIG. 4. MCA scattergram showing the relative positions of various titers of IgG, IgM, and IgA to phase I and II C. burnetii and the various Qfever clinical presentations (inactive, acute, or chronic).

not increase. However, in the study of Guigno et al. (12), IgMwas more precocious than IgG and both were still present athigher levels after 6 months than in the study of Dupuis et al.These discrepancies could be explained by the use, in the workof Guigno et al., of commercial antigen (Biomerieux), which isnot purified.

Figure 3 is of great importance in terms of medical decisionmaking, as it represents the kinetics of the sensitivity for thetwo cutoffs that we have determined previously for the diag-nosis of acute Q fever. Considering the high cutoff (phase IIIgG titer .200 and phase I IgM titer 250), although its PPVis 100% at any time, the sensitivity, i.e., the percentage of casesit allows one to diagnose, varies considerably with time andreaches 10% during the second week after the onset of clinicalillness, 50% during the third week, and 70% after 1 month. Itis interesting to note that this last sensitivity figure, calculatedon a smaller number of cases, is higher than the global one(58.4%). Considering that the set of sera used for the kineticsstudy is representative of all our sera, without technicaldifference in microimmunofluorescence, the previously calcu-lated specificity (which is an intrinsic property of the test)(92.2%) and the prevalence (261/2, 218 = 0.12) were the same.Under these conditions, we applied Bayes' theorem to estimatethe NPV: [specificity x (1 - prevalence)]/{[(1 - sensitivity)x prevalence] + [specificity x (1 - prevalence)]}. The top ofFig. 3 shows the kinetics of the calculated NPV, which in-creases from 88% (which is the minimum due to a prevalenceof 12%) to 96.6%. Thus, given a serum sample with a phase IIIgG titer of < 100, we may reject the diagnosis of active Q fever

with a probability varying from 88% (i.e., 1 - prevalence) theday after the onset of clinical signs to 99.6% 1.5 months afterthe onset.

Diagnosis of chronic Q fever. Since the study of Peacock etal. (15) it has been accepted that the diagnosis of chronic Qfever, and mainly endocarditis, is based on high titers of phaseI IgG and IgA. However, our study differs from that of Peacocket al. because we did not segregate granulomatous hepatitisfrom acute cases since hepatic granulomas are usually presentin acute cases and a liver biopsy sample is not often availableto confirm granulomatous hepatitis (18). These findings havebeen confirmed by Worswick et al. (25) and by Edlinger (7). Ina recent study, Soriano et al. (21) have determined that aserum sample with a phase I IgA titer of .640 (or phase I IgAtiter of .320 and complement fixation titer of -128) ispredictive of Q fever endocarditis.The results presented here suggest improved serodiagnostic

strategies for Q fever. When we consider the cutoff phase I IgGtiter to be .800, the PPV for chronic Q fever is 98.1% with asensitivity of 100% and the NPV is 100% with a specificity of99.6%. When we add a phase I IgA titer of .25, the PPV andNPV do not change, whereas the sensitivity decreases to 96.6%and the specificity rises to 100%. Moreover, a phase I IgG titerof -1,600 gives a PPV of 100%. Considering the importance ofdiagnosing such a disease, it seems important to emphasize thesensitivity so that, in our hands, we may assess a diagnosis ofchronic Q fever when the phase I IgG titer is .800, with a PPVof 98.1%. This is also confirmed by the MCA (Fig. 4) in whicha phase I IgG titer of .1,600 is closely associated with chronic

CLIN. DIAGN. LAB. IMMUNOL.

on February 1, 2021 by guest

http://cvi.asm.org/

Dow

nloaded from

0 FEVER SEROLOGY 195

Q fever, whereas a phase I IgG titer of 800 is in the grey zonebetween acute and chronic Q fever. For phase I IgA, titers of25 and 50, which were considered diagnostic of chronic Qfever, are in fact more closely associated with acute Q fever.The phase I IgA titers most closely associated with chronic Qfever are 400 and 800, which give a good PPV but a lowsensitivity. Thus, we conclude that phase I IgA titers are notuseful for diagnosis of chronic Q fever; however, IgA titersremain essential in follow-up of the disease (17).

In a previous study (18) that we carried out with a limitednumber of serum samples, we found that the specificity ofphase I IgA titers of -25 was 100%. In fact, the present studyshows that it is 96.2%. In the same study, the followingequation (phase I IgG titer - phase II IgG titer and phase IIgA titer - phase II IgA titer) gave a sensitivity of 80%, aspecificity of 79%, a PPV of 94.1%, and an NPV of 92.6%.These figures may now be modified to 60, 98, 53.2, and 98.7%,respectively.

Moreover, as in vitro propagation of phase I C. burnetii canbe hazardous for laboratory workers, we have performed thesame study with anti-phase II antibodies. A phase lI IgG titerof -1,600 gives a high NPV (99.9%) and specificity (96.4%),whereas the PPV is only 71.4% and the sensitivity is 88.5%.The highest PPV that we may obtain with phase II is with anIgG titer of :-1,600 and IgA titer of -400 (95%), but thesensitivity is as low as 56.8%. Thus, a phase 1-nonequippedlaboratory can reject the diagnosis of chronic Q fever with aprobability of 100% and assess this diagnosis with a probabilityof 95%. The sera showing titers which fall in the grey zoneshould be sent to a reference center in order to study anti-phase I antibodies.

Conclusions. Because of the results of this study, we are ablemore accurately to diagnose active and chronic Q fever.

In our hands, the diagnosis of active Q fever can now berejected in a serum sample with a phase II IgG titer of <100,with a probability of 1 - prevalence (88% in our series) theday after the onset of clinical signs, increasing to almost 100%1.5 months after the onset of the disease. At any time, a serumsample with a phase II IgG titer of -200 and phase 11 IgM titerof -50 is 100% predictive of active Q fever. However, we areable to detect only 10% of these cases during the second weekafter the onset of clinical illness, 50% during the third week,and 70% after 1 month. A serum sample with titers betweenthese two cutoff titers will need further confirmation.

In our hands, a serum sample with a phase I IgG titer of-800 is highly predictive of chronic Q fever (98%) and is 100%sensitive. Moreover, a serum sample with a phase I IgG titer of-1,600 is 100% diagnostic of chronic Q fever. Phase I IgAtiters, which were considered as necessary to state positively adiagnosis of chronic Q fever, are no longer useful for thisdiagnosis but remain of prime importance in the serologicalfollow-up. Whenever phase I testing is not available, thediagnosis of chronic Q fever may be rejected with a probabilityof 99.9% in a serum sample with a phase II IgG titer of < 1,600,whereas in a serum sample with a phase II IgG titer of - 1,600and phase II IgA titer of -400 (PPV = 95%) the diagnosis ishighly likely; however, the sensitivity is only 57%. Diagnosis ofa serum sample with titers between these two cutoffs will needto be confirmed by phase I testing.Although serology is the major way to confirm the diagnosis

of such a polymorphic disease, the only way to increase thenumber of detected cases of Q fever is continue physicians'education in the routine use of serologic testing for patientswith any clinical signs suggestive of Q fever or for patients atrisk for chronic Q fever (8).

ACKNOWLEDGMENTS

We thank Maryse Mutillod and Christine Rei for performing allserologic tests, J. Stephen Dumler for reviewing the manuscript, andall the physicians who kindly provided their clinical data.

REFERENCES

1. Anonymous. 1990. Epi Info: a word processing, database andstatistics system for epidemiology on microcomputers. Centers forDisease Control, Atlanta.

2. Benzekri, J. P. 1973. L'analyse des donnees: T.2, l'analyse descorrespondances. Dunod, Paris.

3. Brouqui, P., H. Tissot Dupont, M. Drancourt, Y. Berlan, J.Etienne, C. Leport, F. Goldstein, P. Massip, M. Micoud, A.Bertrand, and D. Raoult. 1993. Chronic Q fever: 92 cases fromFrance including 27 cases without endocarditis. Arch. Int. Med.153:642-648.

4. Burt, C. 1950. The factorial analysis of qualitative data. Br. J.Psychol. 3:166-185.

5. Dierksheide, W. C. 1987. Medical decisions: interpreting clinicaltests. ASM News 53:677-680.

6. Dupuis, G., 0. Peter, M. Peacock, W. Burgdorfer, and E. Haller.1985. Immunoglobulin responses in acute Q fever. J. Clin. Micro-biol. 22:484-487.

7. Edlinger, E. 1985. Immunofluorescence serology: a tool for prog-nosis of Q fever. Diagn. Microbiol. Infect. Dis. 3:343-351.

8. Fergusson, R. J., T. R. D. Shaw, A. H. Kitchin, M. B. Matthews,J. M. Inglis, and J. F. Peutherer. 1985. Subclinical chronic Q fever.Q. J. Med. 222:669-676.

9. Field, P. R., J. G. Hunt, and A. M. Murphy. 1983. Detection andpersistence of specific IgM antibody to Coxiella burnetii by en-zyme-linked immunosorbent assay: a comparison with immunoflu-orescence and complement fixation tests. J. Infect. Dis. 148:477-487.

10. Finidori, J. P., D. Raoult, N. Bornstein, and J. Fleurette. 1992.Study of cross-reaction between Coxiella burnetii and Legionellapteurnophila using indirect immunofluorescence assay and immu-noblotting. Acta Virol. 36:459-465.

11. Gimenez, D. F. 1964. Staining rickettsiae in yolk-sack cultures.Stain Technol. 39:135-140.

12. Guigno, D., B. Coupland, E. G. Smith, I. D. Farrell, U. Dessel-berger, and E. 0. Caul. 1992. Primary humoral antibody responseto Coxiella burnetii, the causative agent of Q fever. J. Clin.Microbiol. 30:1958-1967.

13. Hunt, J. G., P. R. Field, and A. M. Murphy. 1983. Immunoglobulinresponses to Coxiella burtnetii (Q fever): single-serum diagnosis ofacute infection, using an immunofluorescence technique. Infect.Immun. 39:977-981.

14. Lebart, L., A. Morineau, and K. M. Warwick. 1984. Multivariatedescriptive statistical analysis: correspondance analysis and relatedtechniques for large matrices. Wiley, New York.

15. Peacock, M. G., R. N. Philip, J. C. Williams, and R. S. Faulkner.1983. Serological evaluation of Q fever in humans: enhancedphase I titers of immunoglobulins G and A are diagnostic for Qfever endocarditis. Infect. Immun. 41:1089-1098.

16. Raoult, D., J. Etienne, P. Massip, S. Iaocono, M. A. Prince, P.Beaurain, S. Benichou, J. C. Auvergnat, P. Mathieu, P. Bachet,and A. Serradimigni. 1987. Q fever endocarditis in the south ofFrance. J. Infect. Dis. 155:570-573.

17. Raoult, D., P. Y. Levy, J. R. Harle, J. Etienne, P. Massip, F.Goldstein, M. Micoud, J. Beytout, H. Gallais, G. Remy, and J. P.Capron. 1990. Chronic Q fever: diagnosis and follow-up. Ann.N.Y. Acad. Sci. 590:51-60.

18. Raoult, D., J. Urvolgyi, J. Etienne, M. Roturier, J. Puel, and H.Chaudet. 1988. Diagnosis of endocarditis in acute Q fever byimmunofluorescence serology. Acta Virol. 32:70-74.

19. Saporta, G. 1990. Probabilites, analyse des donnees et statistiques.Editions Technip, Paris.

20. SAS Institute, Inc. 1988. SAS® technical report P-179, additionalSAS/STAT" procedures, release 6.03. SAS Institute, Inc., Cary.N.C.

21. Soriano, F., M. T. Camacho, C. Ponte, and P. Gomez. 1993.

VOL. 1, 1994

on February 1, 2021 by guest

http://cvi.asm.org/

Dow

nloaded from

196 TISSOT DUPONT ET AL.

Serological differentiation between acute (late control) and endo-carditis Q fever. J. Clin. Pathol. 46:411-414.

22. Stocker, M. G. P., and P. Fiset. 1956. Phase variation of the NineMile and other strains of Rickettsia burnetii. Can. J. Microbiol.2:310-321.

23. Tissot Dupont, H., D. Raoult, P. Brouqui, F. Janbon, D. Peyra-mond, P. J. Weiller, C. Chicheportiche, M. Nezri, and R. Poirier.1992. Epidemiologic features and clinical presentation of acute Q

CLIN. DIAGN. LAB. IMMUNOL.

fever in hospitalized patients: 323 French cases. Am. J. Med.93:427-434.

24. Williams, J. C., M. R. Johnston, M. Peacock, L. A. Thomas, S.Stewart, and J. L. Portis. 1984. Monoclonal antibodies distinguishphase variants of Coxiella burnetii. Infect. Immun. 43:421-428.

25. Worswick, D., and B. P. Marmion. 1985. Antibody responses inacute and chronic Q fever and in subjects vaccinated against Qfever. J. Med. Microbiol. 19:281-296.

on February 1, 2021 by guest

http://cvi.asm.org/

Dow

nloaded from