CONCISE ARTICLE

99mTc-HMPAO labelled white blood cell scintigraphyin patients with osteoarticular infection: the value of lateimages for diagnostic accuracy and interobserverreproducibility

P. Fernandez & A. Monet & C. Matei & H. De Clermont &M. Guyot & R. Jeandot & H. Dutronc & C. Dumoulin &

M. Dupon & D. Ducassou

Received: 4 February 2008 /Accepted: 25 May 2008 /Published online: 27 June 2008# Springer-Verlag 2008

Abstract The objectives of this study were to evaluate thediagnostic value of 99mTc-HMPAO labelled white bloodcell scintigraphy (WBCS) in patients with suspectedosteomyelitis using late images and to study interobserverreproducibility. This study prospectively included 120patients, and after a follow-up of one year, only 70 patients(n = 49 with implants, n=21 without implants) wereselected. The final diagnosis of infection was based eitheron microbiological data (n = 54) or follow-up (n = 16). We

performed WBCS with 4 h and 24 h scans. Sensitivity,specificity, positive predictive value, negative predictivevalue, and accuracy were 77%, 72%, 83%, 64%, and 75%at 4 h, and 74%, 87%, 91%, 59%, and 79% at 24 h,respectively. The interobserver reproducibility shows a 63%prevalence of agreement between results (κ=0.5) at 4 h and80% (κ=0.74) at 24 h, respectively. WBCS with 24-h imagesimproves specificity and interobserver reproducibility inpatients with suspected osteoarticular sepsis.

Introduction

Osteomyelitis subsequent to an orthopaedic procedure repre-sents a real public health problem. Because of the increasingincidence of nosocomial infections, the prevalence of osteo-articular infections is still too high, associated with 1–2% ofhip prostheses and 2–4% of knee prostheses each year inFrance, in addition to the consequent social and medico-economic impacts. These rates may even be underestimates.The early recognition of bone or joint infection is of greatimportance because unrecognised osteomyelitis can be asevere disease resulting in significant morbidity ranging fromchronic pain and fistula to loss of function, amputation, oreven death. The three principal mechanisms of infectionaround implants are: direct per-operative inoculation (70–85%of total joint replacement infections), hematogenous infectionvia bacteremia, and contiguous extension from regionalsepsis, associated with numerous risk factors. Among variousphysiopathological events, the development of slime leads toexceptional resistance to therapy on infection [1].

Because clinical impressions, biological data, and plainX-rays are often inconclusive or nonspecific, clinicians rely

Eur J Clin Microbiol Infect Dis (2008) 27:1239–1244DOI 10.1007/s10096-008-0563-x

P. Fernandez (*) :A. Monet :C. Matei :M. GuyotService de Médecine Nucléaire, CHRU Bordeaux,Hôpital Pellegrin,Place Amélie Raba-Léon,33076 Bordeaux Cédex, Francee-mail: philippe.fernandez@u-bordeaux2.fr

P. Fernandez :H. De Clermont : R. Jeandot :M. Dupon :D. DucassouUniversité Victor Segalen Bordeaux2,Bordeaux 33076, France

P. FernandezInserm, U577,Bordeaux 33076, France

H. De Clermont :D. DucassouService de Médecine Nucléaire, CHRU Bordeaux,Hôpital Pellegrin,Pessac 33600, France

H. Dutronc :M. DuponDépartement de Maladies Infectieuses, CHRU Bordeaux,Hôpital Pellegrin,Bordeaux 33076, France

C. DumoulinService de Rhumatologie, CHRU Bordeaux, Hôpital Pellegrin,Bordeaux 33076, France

on imaging examinations to confirm their opinions [2]. Innumerous situations, computed tomography (CT) andmagnetic resonance imaging (MRI) are not helpful inpatients with prosthetic implants or osteosynthesis materialsbecause altered bone marrow signals remain abnormal forseveral months after surgery and metallic implants causemajor imaging artifacts excluding the detection of minorchanges [3]. Nuclear medicine can therefore play animportant role in the early stages of the infection diagnos-ing process. However, bacteriological results only providecertain confirmation of the diagnosis of sepsis, and indeedin the case of a negative result, the absence of sepsis after 1year of follow-up is required to confirm sterility [4].

Nuclear medicine provides a wide variety of radiopharma-ceuticals for infection imaging such 99mTc diphosphonates(DP), 67gallium [5], In111-oxine or Tc99m exametazimeautologous granulocytes, In-111 or Tc-99m labelled humannonspecific immunoglobulin, but also 99mTc labelledmonoclonal antigranulocyte antibodies and radiolabelledantibiotics [6, 7]. Nevertheless, it is well established thatthe gold standard in nuclear imaging of infection, except forvertebral sepsis, remains In-111 oxinate labelled autologouswhite blood cell (WBC), which requires in vitro granulocyteisolation [8, 9]. For a few years 99mTc HexaMethyl-PropyleneAmineOxime (HMPAO or exametazime) replacedIn-111 oxine labelling because of more favourable dosimetrywith fewer functional alterations [10]. Nevertheless, a reviewof the literature shows that the studies are heterogeneouswith regard to patient characteristics, the gold standard usedfor diagnosis, and the various nuclear medicine proceduresmake comparison difficult. Indeed, diagnostic performancesmay differ depending on whether infection occurred afterorthopaedic surgery, the presence of an implant, whether it isan acute or chronic process, the microbiological goldstandard, the existence of prolonged clinical follow-up, thecombination of different isotopic procedures and theirinterpretation. An evaluation of radiolabelled leukocytescintigraphy in each institute therefore appears to be animportant factor and may also provide answers to clinicalquestions. In our study, we evaluate 99mTc-HMPAO radio-labelled leukocytes to explore patients with suspectedosteomyelitis with or without implants, analysing 4- and24-h scans blindly to evaluate the value of late imagesand interobserver reproducibility between three trainedphysicians.

Patients and methods

Patients

Two hundred twenty patients were referred to the nuclearmedicine department over a period of 18 months. All had

suspected osteoarticular infection based on clinical signs orlaboratory blood tests. Only patients with a positive bonescan underwent a WBC scintigraphy, because a negativebone scan rules out osteomyelitis. One hundred twentyselected patients were followed prospectively. Among thosepatients, only patients with a microbiological resultobtained in satisfactory conditions or patients followed-upfor at least 1 year when biopsy and antibiotherapy weredisproved were selected for the study. With these criteria,only 70 patients qualified for evaluation: 45 men, 25women, mean age 59 years (range, 21–81).

WBC scintigraphy

Blood samples (60–120 ml) were collected on heparinizedtubes. Cell rich plasma was obtained after sedimentation at37°C. The granulocytes were labelled for 20 min with1110MBq of freshly (<30 min) prepared 99mTc-HMPAO(Ceretec, Amersham, France) after separation of the cell-rich plasma on a LymphoprepTM (Nycomed Pharma, Oslo,Norway) gradient. The labelled granulocytes were washed,resuspended, and reinjected intravenously within a timeframe of no more than 2 h after the initial blood sampling.The lower injected activity value was 185MBq.

Scintigraphic acquisition

First, a three-phase bone scan was performed after theadministration of 99mTc-MDP (Osteocis, Schering, Saclay,France) with activity between 700 and 960MBq. Theimages were acquired with a gamma camera (DSX orDST XLi, GE, Buc, France) equipped with a low energyparallel-hole collimator. Only patients with a positive bonescan underwent a WBC scintigraphy. The WBC scanimages were obtained 48 h later, 4 and 24 hours after theinjection of the 99mTc-HMPAO labelled polymorphonu-clear cells. The static images were acquired in the sameposition as for the bone scan with a 256×256 matrix and anacquisition time of 10 min. A liver-spleen image wassystematically obtained as a control.

Scintigraphic interpretation

The coupled exams were analysed blindly by three trainedphysicians and the results were expressed according to thefollowing algorithm (Table 1). Physicians interpreted theimages without any other imaging modality results orbiological findings. In cases of cold defect appearance,scans were scored as high probability, because dataobtained in our institution has shown that all patients witha cold defect in the hip with 99mTc-HMPAO radiolabelledleukocytes present a confirmed infection [11].

1240 Eur J Clin Microbiol Infect Dis (2008) 27:1239–1244

Diagnosis

The positive diagnosis of sepsis was established accordingto strict international bacteriological criteria as follows:

– Association of clinical signs (local inflammationprocess or pain or fistula) with two positive culturesfor the same microorganism obtained from a jointpuncture or surgical biopsy

or

– Association of clinical signs as described above, withone positive culture for a microorganism from a deeptissue specimen, with a fistula or evidence of infectionduring the surgery

In the absence of these criteria, sepsis was ruled out.When the bacteriological sample obtained was not accept-able, the patient was excluded from the study.

The diagnosis of sepsis was ruled out when:

– Biopsies were sterile in satisfactory conditions asrecommended in guidelines.

– Biopsies were not available, but 1-year follow-up withno sign of infection and antibiotherapy was consideredas the standard to confirm noninfection at the timeof WBC scintigraphy. In the case of sepsis during this1-year interval, the patient was excluded from the study.

Data analysis

This study was based on an observational cohort withprospective harvesting of the data. For data analysis ofdiagnostic parameters, results were gathered as follows: verylow/low probability results and high/very high probabilityresults for 4-h WBC scans, and, low probability results andintermediate/high probability results for 4-h/24-h WBCscans. Comparison between two proportions was performedwith either a chi-square test or a Fisher’s test depending ontheoretical patient group sizes. Comparison between twomeans was calculated with the Student’s t test and thecomparison between two medians with the Mann-Whitney’stest (α=0.05, p<0.05). The agreement among observers wasevaluated with the κ statistic and reported with the relative95% confidence interval (CI).

Results

Among 70 patients selected for the study following diagnosticcriteria, the diagnosis of infection was confirmed for 45patients and ruled out for 25 patients. Microbiological sampleswere obtained in 73% of patients via deep biopsy (42 infectedpatients and nine without infection), and in 4% during asurgical procedure (three infected patients). Sixteen patients(23%) were followed-up (noninfected patients). Amongpatients with ostearticular implants (n = 49), 73% had aninfection, versus 43% in patients with no implant (p<0.05).Among the infected patients, 19 had a recurrence and 26 hada de novo infection (23% acute infections, 77% chronicinfections). For half the patients, Staphylococcus aureus orcoagulase negative was diagnosed, and a bacillus Gramnegative in 25% of infected patients. Table 2 summarizesdiagnostic parameters for each of the physicians. Combiningresults we obtain the following results for sensitivity,

Table 1 Interpretation criteria based on 4-h image scans only or inassociation with 24-h image scans

Uptake Score Result

4 h WBCscan

Normal 1 Very lowprobability

Heterogenous uptake >background

2 Lowprobability

Moderate focal uptake 3a Highprobability

Intense focal uptake 3b Very highprobabilityCold defect 4

4 h/24 hWBC scan

Normalized uptake 1 LowprobabilityUptake decrease

compared to medullaNo modified uptake 2 Intermediate

probabilityFocal uptakeintensification

3 Highprobability

Cold defect 4

Table 2 Distribution of scintigraphic data according to the positive diagnosis

Physician 1 Physician 2 Physician 3

4 h 24 h 4 h 24 h 4 h 24 h

Sensitivity 73% 77% 82% 69% 77% 76%Specificity 76% 96% 64% 80% 76% 84%PPV 84% 97% 80% 86% 85% 89%NPV 61% 70% 66% 41% 66% 66%Accuracy 74% 84% 75% 73% 77% 79%

Eur J Clin Microbiol Infect Dis (2008) 27:1239–1244 1241

specificity, positive and negative predictive value, andaccuracy: 77%, 72%, 83%, 64%, and 75% for 4-h scans,and 74%, 87%, 91%, 59%, and 79% for 24-h scans,respectively. The statistical comparison between the valuesshowed a significant difference for specificity (p<0.05)between each interpretation, with an improvement ofspecificity with 24-h scans, especially when patients withimplants were considered. The prevalence of agreementamong three physicians was 63% with a κ index of 0.5 for4-h scans, and the prevalence increased to 80% with a κindex of 0.74 for 24-h scans. According to Landis’classification, the interobserver agreement was moderate for4-h scintigraphy and good for 24-h scintigraphy [12]. Thus,24-h scan results have significantly greater reproducibilitythan the 4-h scan results because we found more agreementwhen 24-h scans were considered.

Discussion

An aging population means that an increasing number ofjoint replacement procedures will be required in the future.Methods to prevent, diagnose, and treat infection must beperfected to reduce the social and medico-economic impactof total hip and total knee arthroplasty procedures. The rateof infection after primary or revision procedures hasremained stable for the past few years and treatment ofsuch complications is costly. After an orthopaedic proce-dure, infection at the operative site can be difficult todiagnose. Radiological imaging lacks the specificity neces-sary to confirm diagnosis [2, 13]. Literature on thediagnostic parameters of radiolabelled leukocyte scansdiffers significantly with regard to the patient populationselected, the gold standard chosen for the diagnosis ofinfection, and the length of follow-up [14–18]. Data showsimilar results for both radiolabelling methods. Despiteunfavourable dosimetry, some authors prefer 111In-oxinateleukocyte radiolabelling systematically and simultaneouslycoupled with medullar scintigraphy to increase specificityin patients with implants who may present modification ofmedullar distribution [15–17]. In this case, the absence of

congruence between images signals infection. A scintigra-phy with colloids does not appear to be necessary for99mTc-HMPAO radiolabelled leukocyte scintigraphy ifscans are performed at 24 h [18]. In our study, patientsunderwent a bone scan to be elected for WBC scan becauseosteoarticular infection can be ruled out when the bone scanis normal. Our results show the great value of performinglate images to differentiate between infection or medulla

A B

Fig. 1 Case of an infected lefthip prosthesis (bacteriologicaldiagnosis) showing an intenseuptake of radiolabelled leuko-cytes at 4 h (a), increasing at24 h with a decreasing medullarcontrolateral uptake (b)

R L

R L

R L

A

B

C

Fig. 2 Patient with suspicion of infection in the right knee prosthesis.Biopsy ruled out the infection diagnosis. 99mTc-HMPAO radiolabelledleukocyte scan show leukocyte accumulation in the right knee (a) witha decreasing uptake at 24 h (b), and the same pattern with colloidimaging (c), arguing for medullar uptake

1242 Eur J Clin Microbiol Infect Dis (2008) 27:1239–1244

(Fig. 1). Late images increase specificity significantly (87%versus 72%). These results are in agreement with somearticles in the literature. For example, Larrika et al. haveshown improvement in diagnostic parameters such asspecificity in hip and knee prosthesis infections (100%versus 90% and 82% versus 77% for hip and for kneeprostheses, respectively) and also sensitivity (83% versus50% and 100% versus 87% for hip and knee prostheses,respectively) when they compared early and late images[19, 20]. Dutton et al. carried out a study to show the abilityof early images of 111In-oxinate radiolabelled leukocytescintigraphy to replace colloid scintigraphy to imagemedulla distribution [21]. In our experience, when colloidscintigraphy is realized, images obtained with colloids didnot contribute to diagnosis but showed the same results asthose obtained with 24-h WBC scans (Fig. 2). To improvediagnostic parameters, Pelosi et al. have suggested adding aquantification method to determine uptake differencesbetween 4- and 24-h images considering a 10% up ratiopositive for infection [22]. Specificity is improved whencompared with values obtained with qualitative interpreta-tion (96% versus 70%). For interobserver reproducibility,only Pelosi et al. report a prevalence of agreement in theresults of 72.6% between three physicians for 99mTc-HMPAO radiolabelled leukocyte scintigraphy. Our studyexamined the differences of the prevalence of agreementbetween 4- and 24-h images among three physicians with63% agreement if 4-h images only were considered and80% agreement if both 4- and 24-h images were used forinterpretation.

The cold defect pattern in certain hip infections hasnever been described, whereas there have been numerousreports of cold defect in the axial skeleton. In the lattercase, bone marrow replacement is one of the generallyaccepted reasons for such a result. In our institution, wehave shown that all patients with a cold defect in the hipwith 99mTc-HMPAO radiolabelled leukocytes presented aconfirmed infection [11]. Tc-99m ciprofloxacin would seemto be a good agent highlighting a hot spot in vertebralosteomyelitis [23]. This tracer could be used in thisparticular pattern.

Since PET can detect increased glucose metabolism invivo, various studies have shown increased fluorine-18fluorodeoxyglucose (F-18 FDG) uptake in activated in-flammatory cells, abdominal abscesses, and also in boneinfection diseases with promising results [24]. Given thelow sensitivity and the specificity of radiolabelled granulo-cytes in chronic bone infections, some authors have studiednew tracers. 99mTc-PEG (polyethyleneglycol) liposomesand 99mTc-HYNC (hydrazinonicotinamide)-IgG have beensuccessfully used to image experimental chronic ostemye-litis. Despite their ease of preparation and a lower radiationdose, they cannot differentiate inflammation or infection.

Preliminary results have been obtained in 35 patients withradiolabelled liposomes and have shown a lack of speci-ficity because their uptake is related to the increase invascularization [25].

Another encouraging research avenue is the use ofchemotactic peptides with some interesting results withvarious Tc-99m labelled peptides, which bind to bacteria,allowing significant accumulation at sites of infection [26].

In summary, our results highlight the value of 24-h imagesto improve the specificity of 99mTc-HMPAO radiolabelledleukocytes in osteomyelitis diagnosis with or without thepresence of implants. Moreover, late images also improveinterobserver reproducibility when early and late images arecompared.

References

1. Bernard L et al (2004) Value of preoperative investigations indiagnosing prosthetic joint infection: retrospective cohort studyand literature review. Scand J Infect Dis 36(6-7):410–416doi:10.1080/00365540410015240

2. Munk PL et al (1994) Imaging after arthroplasty. Can AssocRadiol J 45(1):6–15

3. Deely DM, Schweitzer ME (1997) MR imaging of bone marrowdisorders. Radiol Clin North Am 35(1):193–212

4. Atkins BL et al (1998) Prospective evaluation of criteria formicrobiological diagnosis of prosthetic-joint infection at revisionarthroplasty. The OSIRIS collaborative study group. J ClinMicrobiol 36(10):2932–2939

5. Turpin S, Lambert R (2001) Role of scintigraphy in musculoskel-etal and spinal infections. Radiol Clin North Am 39(2):169–189doi:10.1016/S0033-8389(05)70271-2

6. Becker W et al (1994) Detection of soft-tissue infections andosteomyelitis using a technetium-99m-labeled anti-granulocytemonoclonal antibody fragment. J Nucl Med 35(9):1436–1443

7. Rubello D et al (2004) Role of anti-granulocyte Fab’ fragmentantibody scintigraphy (LeukoScan) in evaluating bone infection:acquisition protocol, interpretation criteria and clinical results.Nucl Med Commun 25(1):39–47 doi:10.1097/00006231-200401000-00006

8. Seabold JE et al (1997) Procedure guideline for indium-111-leukocyte scintigraphy for suspected infection/inflammation. JNucl Med 38(6):997–1001

9. Palestro CJ, Love C, Miller TT (2006) Infection and musculo-skeletal conditions: imaging of musculoskeletal infections. BestPract Res Clin Rheumatol 20(6):1197–1218 doi:10.1016/j.berh.2006.08.009

10. Ak I et al (2002) Labeling of mixed leukocytes with (99m)Tc-HMPAO causes severe chromosomal aberrations in lymphocytes.J Nucl Med 43(2):203–206

11. Galperine T et al (2004) Cold bone defect on granulocytes labelledwith technetium-99m-HMPAO scintigraphy: significance andusefulness for diagnosis and follow-up of osteoarticular infec-tions. Scand J Infect Dis 36(3):209–212 doi:10.1080/00365540310018851

12. Landis JR, Koch GG (1977) An application of hierarchical kappa-type statistics in the assessment of majority agreement amongmultiple observers. Biometrics 33(2):363–374 doi:10.2307/2529786

Eur J Clin Microbiol Infect Dis (2008) 27:1239–1244 1243

13. Dutronc H, Bocquentin F, Dupon M (2004) Radiographicdiagnosis in bone and joint infection management. Med MalInfect 34(6):257–263 doi:10.1016/j.medmal.2004.04.006

14. Devillers A, Moisan A, Jean S, Arvieux C, Bourguet P (1995)Technetium-99m hexamethylpropylene amine oxime leucocytescintigraphy for the diagnosis of bone and joint infections: aretrospective study in 116 patients. Eur J Nucl Med 22(4):302–307 doi:10.1007/BF00941845

15. Palestro CJ et al (1990) Total-hip arthroplasty: periprostheticindium-111-labeled leukocyte activity and complementary tech-netium-99m-sulfur colloid imaging in suspected infection. J NuclMed 31(12):1950–1955

16. Palestro CJ, Love C, Tronco GG, Tomas MB, Rini JN (2006)Combined labeled leukocyte and technetium 99m sulfur colloidbone marrow imaging for diagnosing musculoskeletal infection.Radiographics 26(3):859–870 doi:10.1148/rg.263055139

17. Seabold JE et al (1991) Postoperative bone marrow alterations:potential pitfalls in the diagnosis of osteomyelitis with In-111-labeled leukocyte scintigraphy. Radiology 180(3):741–747

18. Segura AB et al (2004) What is the role of bone scintigraphy inthe diagnosis of infected joint prostheses? Nucl Med Commun 25(5):527–532 doi:10.1097/00006231-200405000-00016

19. Larikka MJ et al (2001) Extended combined 99mTc-white bloodcell and bone imaging improves the diagnostic accuracy in thedetection of hip replacement infections. Eur J Nucl Med 28(3):288–293 doi:10.1007/s002590000463

20. Larikka MJ et al (2001) Improved method for detecting kneereplacement infections based on extended combined 99mTc-whiteblood cell/bone imaging. Nucl Med Commun 22(10):1145–1150doi:10.1097/00006231-200110000-00015

21. Dutton JA, Bird NJ, Skehan SJ, Peters AM (2004) Evaluation of a3-hour indium-111 leukocyte image as a surrogate for atechnetium-99m nanocolloid marrow scan in the diagnosis oforthopedic infection. Clin Nucl Med 29(8):469–474 doi:10.1097/01.rlu.0000132880.77924.b1

22. Pelosi E et al (2004) 99mTc-HMPAO-leukocyte scintigraphy inpatients with symptomatic total hip or knee arthroplasty: improveddiagnostic accuracy by means of semiquantitative evaluation. JNucl Med 45(3):438–444

23. Amaral H, Morales B, Pruzzo R, Britton KE (1999) Cold-hotmismatch between Tc-99m HMPAO-labeled leukocytes and Tc-99m ciprofloxacin in axial skeleton infections: a report of threecases. Clin Nucl Med 24(11):855–858 doi:10.1097/00003072-199911000-00007

24. Kalicke T et al (2000) Fluorine-18 fluorodeoxyglucose PET ininfectious bone diseases: results of histologically confirmed cases.Eur J Nucl Med 27(5):524–528 doi:10.1007/s002590050538

25. Dams ET et al (2000) 99mTc-PEG liposomes for the scintigraphicdetection of infection and inflammation: clinical evaluation. JNucl Med 41(4):622–630

26. WellingMM et al (1999) Imaging of bacterial infections with 99mTc-labeled human neutrophil peptide-1. J Nucl Med 40(12):2073–2080

1244 Eur J Clin Microbiol Infect Dis (2008) 27:1239–1244