Diagnosis of Mycobacterium ulcerans infection (Buruli

0 downloads 0 Views 106KB Size Report
1 Ghana Health Service, Ministry of Health, Tema, Ghana. 2 Noguchi Memorial Institute for Medical Research, University of Ghana, Legon, Ghana. 3 Section of ...... Journal of Leprosy and Other Mycobacterial Diseases 73,. 5–12. Darie H, Le ...
Tropical Medicine and International Health

doi:10.1111/j.1365-3156.2007.01990.x

volume 13 no 2 pp 191–198 february 2008

Diagnosis of Mycobacterium ulcerans infection (Buruli ulcer) at a treatment centre in Ghana: a retrospective analysis of laboratory results of clinically diagnosed cases Ernestina Mensah-Quainoo1, Dorothy Yeboah-Manu2, Caroline Asebi1, Francis Patafuor1, David Ofori-Adjei2, Thomas Junghanss3 and Gerd Pluschke4 1 Ghana Health Service, Ministry of Health, Tema, Ghana 2 Noguchi Memorial Institute for Medical Research, University of Ghana, Legon, Ghana 3 Section of Clinical Tropical Medicine, Department of Tropical Hygiene and Public Health, University Hospital Heidelberg, Heidelberg, Germany 4 Swiss Tropical Institute, Basel, Switzerland

Summary

Clinical diagnosis of Mycobacterium ulcerans infection is currently accepted as sufficient basis for treating the disease. Inadequate laboratory resources in the highly endemic areas of Africa often limit possibilities for in-country confirmation of clinical judgement. We analysed records of 99 Buruli ulcer (BU) patients diagnosed clinically and treated surgically at Amasaman Health Centre in Ghana, for whom post-treatment diagnostic laboratory tests were performed. Comparison of clinical diagnoses with test results obtained by an in-country laboratory on samples of excised tissue showed a high specificity of clinical judgement. Among lesions with three laboratory tests (microscopy for acid fast bacilli, culture and IS2404 polymerase chain reaction) done, 94% tested positive at least once and 83% twice. Thus correct clinical diagnosis of BU by well trained health workers is achievable, although the quality of clinical diagnosis should be monitored by intermittent testing in national reference laboratories. However, being retrospective, this study did not permit sensitivity and negative predictive value analysis. keywords Buruli ulcer, clinical diagnosis, Mycobacterium ulcerans

Introduction Currently, the diagnosis of Buruli ulcer (BU), caused by Mycobacterium ulcerans, is mainly clinical, and this is considered sufficient basis for treatment decision (The Uganda Buruli Group 1970; WHO 2000, 2001b). The brunt of the global burden of BU falls on West Africa, where cases occur in thousands (Kanga & Kacou 2001; Amofah et al. 2002; WHO 2004) in remote areas. These countries often lack the resources to meet the high cost of care (Asiedu & Etuaful 1998). In particular, laboratories are inadequate or absent at peripheral centres where majority of cases are treated. Pre-treatment diagnosis of non-ulcerated lesions is further limited. Fine needle biopsy to obtain specimens from closed lesions for polymerase chain reaction (PCR) analysis is currently being evaluated. Clinical judgement will most likely remain important for pre-treatment diagnosis in endemic developing countries until simple and inexpensive field-compatible tests become available. The accuracy of this method of diagnosis is therefore of interest.

ª 2008 Blackwell Publishing Ltd

Although clinical diagnosis is considered unchallenging (van der Werf et al. 1999), incidents of misdiagnosis have been reported. BU cases have been missed initially (Semret et al. 1999; Evans et al. 2003; Coloma et al. 2005) and, conversely, presumed BU lesions proved to be other conditions (Revill et al. 1973; Mwanatambwe et al. 2002; Guarner et al. 2003; Stienstra et al. 2003; Debacker et al. 2004; Bretzel et al. 2005; Siegmund et al. 2005). Even with careful assessment, nearly a third of the cases in a study (Etuaful et al. 2005) on the efficacy of antibiotic treatment turned out to be other conditions, indicating that in particular the clinical diagnosis of pre-ulcerative lesions is error-prone. WHO requires confirmation of suspected cases with positive results of at least two tests (WHO 2001a) from among microscopy with staining for acid fast bacilli (AFB), culture, histopathology and IS2404 PCR. Ga West District (GWD) in Ghana is highly endemic for BU. A survey in 1997 which found 340 active cases showed the disease clustered in rural communities in the lowland basin of the Densu River (Mensah-Quainoo, unpublished). A national case search in 1999 recorded 1113 cases in the district (467 being active), constituting 20% of the total 191

Tropical Medicine and International Health

volume 13 no 2 pp 191–198 february 2008

E. Mensah-Quainoo et al. Clinical and laboratory diagnosis of Buruli ulcer

national figure of 5619 (Amofah et al. 2002). In the endemic communities BU is recognized well, with several local names, such as Aboa gbonyo, Aboa fon (bad animal or terrible disease), Odonti fla (cotton wool wound), Odonti hela, Detifu dor (cotton wool disease) and Tsina asane (boil of cattle, carbuncle), which aptly describe its features. As reported for other areas (Johnson et al. 2004), many patients seek medical care only after traditional treatment failed. BU cases in GWD are treated at a rural health facility, Amasaman Health Centre (AHC). Surgical services were established at this centre to counter difficulties in referring cases for treatment elsewhere. Treatment decisions at this centre relied exclusively on clinical diagnosis. In this study, results of post-treatment tests on excised tissue of clinically diagnosed cases, performed in an in-country research laboratory, were retrospectively analysed.

Materials and methods Patients All patients in this study were reviewed by one clinician who made the final diagnostic decision. Cases and lesions were classified according to the standard WHO definitions (WHO 2001b). Patients clinically diagnosed as having BU and surgically treated at AHC during the 3-year period beginning of July 2001 to end of July 2004, who had single lesions on first presentation, and at least two diagnostic tests done, and whose complete clinical records were available, were included in the study. Only one patient received anti-mycobacterial therapy prior to surgical excision. We reviewed patients’ clinical notes, Bu 01 Forms (WHO 2000) and laboratory records. Laboratory diagnostic methods Specimens were taken from lesions intra-, not post-operatively, after routine thorough surgical site cleaning with chlorhexidine, hypochlorite and boiled water. This reduces the risk of contamination of culture. Specimens were taken from the observable centre of pre-ulcerative lesions, and from parts of ulcers with necrotic subcutaneous tissues still in situ. This increases the likelihood that tissue with the highest load of M. ulcerans bacilli is sampled. In all cases, care was taken to prevent contamination of the specimen during and after collection. Samples were placed in modified Dubos transport medium as described by Yeboah-Manu et al. (2004), and transported within one week to the Noguchi Memorial Institute for Medical Research (NMIMR) in Ghana, where culture, Ziehl-Neelsen (ZN) staining and IS2404 PCR were 192

performed. For each patient, all tests done in each instance were performed on the same sample. Cultivation was done as described by Yeboah-Manu et al. (2004). Specifically, about 1 cm3 (one half) of tissue sample of the specimens was cut into smaller pieces, homogenized and suspended in Dulbecco’s phosphate buffered saline (PBS). The suspension was decontaminated by the oxalic acid method. After decontamination, the homogenate was concentrated by centrifugation, and 100 ll aliquots of the 1 ml suspensions was used to inoculate in duplicate LJ tubes. These were incubated at 32 ºC and all suspected M. ulcerans isolates were confirmed by IS2404 PCR. Two drops of the remaining suspension were used to prepare two slides per sample for ZN staining. Stained smears were read (Yeboah-Manu et al. 2004). The remaining half of the tissue specimen was used for IS2404 PCR. DNA was extracted by heating the sample for 1 h at 95 C in 500 ll of an extraction mixture (50 mM Tris–HCl, 25 mM EDTA, and 5% monosodium glutamate). After cooling, 100 ll of a 50 mg ⁄ ml lysozyme solution was added and incubated for 2 h at 37 C. Seventy micro litre of proteinase K-10x buffer [100 mM Tris–HCl, 50 mM EDTA, 5% sodium dodecyl sulphate (pH 7.8)] and 10 ll of a 20 mg ⁄ ml proteinase K solution were then added and incubated at 45 C overnight. The bacterial cell wall was fully disrupted by adding 200 ll of 0.1 mm-diameter zirconia beads (BioSpec Products) to each sample and vortexing at full speed for 4 min. Beads and undigested tissue fragments were removed by brief centrifugation, and the supernatants were transferred to fresh tubes for phenol-chloroform (Fluka) extraction. The DNA contained in the upper phase was precipitated with ethanol and resuspended in 100 ll of water. The IS2404 sequence was amplified in a 50 ll reaction volume using the Qiagen Taq DNA polymerase kit. The primer sets used for the amplification were as described by Ross et al. (1997). The reaction mixture contained a 1.0 lM concentration of each primer, 200 lM concentration of each deoxynucleoside triphosphate, 1.5 mM MgCl2, 1x PCR buffer, 1 U of Taq DNA polymerase (Qiagen, Hilden, Germany), and approximately 50 ng of DNA. Thermocycling parameters were 95 C for 3 min; 35 cycles of 1 min at 94 C, 1 min at 60 C, 1 min at 72 C and 7 min at 72 C. Ten microlitres of amplified DNA was subjected to electrophoresis in a 2% agarose gel and detected by ethidium bromide staining and UV transillumination. Negative and positive controls were routinely included. The amplicons were sized by comparing to a 1-kb ladder. The results of the laboratory tests of treated patients were compared with their clinical diagnosis retrospectively. The levels of agreement between different tests overall, as

ª 2008 Blackwell Publishing Ltd

Tropical Medicine and International Health

volume 13 no 2 pp 191–198 february 2008

E. Mensah-Quainoo et al. Clinical and laboratory diagnosis of Buruli ulcer

well as with respect to lesion stage, were quantified using Kappa statistic, where k = 0 is equivalent to the level of agreement expected by chance and k = 1 indicates perfect agreement. NMIMR ethical review board gave permission for performing the study. Results Patient population and clinical features During the study period, 131 BU patients were clinically diagnosed and treated at AHC. Administrative difficulties prevented the testing of 20. Of the remaining 111 patients, one with incomplete notes and 11 who had multiple concurrent lesions were excluded. Ninety-two of the 99 remaining patients had one sample each tested; the other seven had additional samples from same-site and differentsite recurrent lesions tested. Altogether, we tested 109 samples of 99 eligible patients. Forty-two males aged 3–70 years and 57 females aged 2– 75 were included in the study. Median age was 11 years and 62.6% were 14 years old and younger (Figure 1) and 91.7% of lesions were on the limbs (Table 1). There was a statistically significant difference between upper limbs and lower limbs in the distribution of lesions between the left and right sides (v2 = 6.3, P = 0.01). The left leg and right arm were more often affected than the corresponding

opposite limbs. Forty-eight percent (52 ⁄ 108) of lesions were pre-ulcerative, 33.3% (36 ⁄ 108) in earlier ulcer stages and 18.5% (20 ⁄ 108) were chronic ulcers. Laboratory test results in patients with recurrences As the history, clinical features and course of illness left no doubt that the primary and recurrent lesions in the seven patients with recurrences were caused by M. ulcerans infection, the 17 sets of laboratory results of these patients were first examined separately (Table 2). AFB and culture were performed on all 17 samples. The culture for one sample was contaminated and for two samples was not PCR-tested. Thus, 17 AFB and 15 PCR tests were done, and 16 cultures were evaluable for the seven patients, totalling 48 evaluable tests. Thirty-eight (79.2%) of these on the 17 samples were positive. Each test method yielded at least one negative result. Interestingly, the original lesions in patients #R2 and #R3 were positive for all three tests, whereas their second lesions returned only one positive test each. Patient #R2 then had a third lesion which was again all positive. Patient #R1 had a positive culture of the primary abdominal lesion but a negative result on the sample from the same-site recurrent lesion. Case #R4 demonstrates similarly inconsistent test results on three samples from the same-site lesions. Of the seven patients, only case #R6 had no negative result.

18

Number of patients

16

Clinical diagnosis vs. laboratory results

Male Female

14 12 10 8 6 4 2 0

0–4 5–9 10– 15– 20– 25– 30– 35– 40– 45– 50– 55– > 60 14 19 24 29 34 39 44 49 54 59

Age-group Figure 1 Age-sex distribution of BU patients enrolled. Table 1 Anatomical site distribution of BU lesions (108 lesions in 99 patients) Location of lesion

Male

Female

Left arm Right arm Left leg Right leg Trunk

6 11 15 10 4

10 17 19 11 5

16 28 34 21 9

14.8 25.9 31.5 19.4 8.3

Total

46

62

108

100.0

ª 2008 Blackwell Publishing Ltd

Total

%

Next we compared the clinical diagnoses at first presentation for all 99 patients with their test results. For the seven patients who had recurrences, only the test results of their primary lesions, which posed a true challenge for clinical decision, were included in this assessment. For the 99 lesions, AFB and culture were performed in all cases, whereas PCR was done for 85. The cultures for two of the 85 lesions were contaminated. This resulted in 83 patients with all three tests done and evaluable: 14 patients with only AFB and culture done and evaluable, and two patients with AFB and PCR evaluable. The 92 single-lesion results, together with the primary-lesion results of the seven patients with recurrences are summarised in Table 3. Of the evaluable tests, 78.8% (78 ⁄ 99) were positive for AFB, 79.4% (77 ⁄ 97) for culture and 72.9% (62 ⁄ 85) for PCR. In 36.4% (36 ⁄ 99) of patients, negative results with one test were associated with positive results for one or both of the two other tests, an indication of false negative results. Among the 83 patients for whom all three tests were carried out, 94.0% (78 ⁄ 83) tested positive at least once, 83.1% (69 ⁄ 83) at least twice, and 53.0% (44 ⁄ 83) tested positive in all three tests (Table 3). Of the 14 cases with 193

Tropical Medicine and International Health

volume 13 no 2 pp 191–198 february 2008

E. Mensah-Quainoo et al. Clinical and laboratory diagnosis of Buruli ulcer

Table 2 Laboratory test results in seven patients with multiple recurrent lesions Case ID*

Date

Classification and (location)

Lesion ⁄ sample

AFB

Culture

PCR

#R1

8 May 2002 12 November 2002 14 January 2003

Plaque Plaque Papule

Positive Positive Positive

Positive Negative Contaminated

Not done Not done Positive

#R2

15 October 2002 8 November 2002

Primary lesion abdomen Recurrent lesion same-site (abdomen) Recurrent lesion different – site (left thigh) Primary lesion (anterior left knee) Recurrent lesion different –site (posterior right thigh) Recurrent lesion different –site (postero-lateral aspect left thigh) Primary lesion right leg Recurrent lesion different –site (left foot) Primary lesion (right knee and leg) Recurrent lesion same-site The same recurrent lesion Primary lesion (left thigh) Recurrent lesion same-site Primary lesion (right arm) Recurrent lesion same-site Primary lesion (right arm) Recurrent lesion same-site

Plaque Nodule

Positive Negative

Positive Positive

Positive Negative

Nodule

Positive

Positive

Positive

Nodule Nodule

Positive Negative

Positive Negative

Positive Positive

Plaque Ulcer Sample 1 Sample 2 Plaque Plaque Plaque Plaque Plaque Plaque

Positive Positive Positive Negative Positive Positive Positive Positive Positive

Positive Negative Positive Positive Positive Positive Positive Positive Positive

Positive Negative Negative Positive Positive Positive Positive Negative Positive

15 January 2003 #R3

13 November 2003 11 February 2004

#R4

12 April 2002 14 March 2003 13 November 2003 23 March 2002 17 October 2002 18 November 2002 8 April 2003 24 April 2002 5 December 2002

#R5 #R6 #R7

*Sex and age: #R1: female, 4 years; #R2: male, 3 years; #R3: male, 10 years; #R4: female, 10 years; #R5: female, 8 years; #R6: male, 23 years; #R7: female, 26 years. Table 3 Laboratory test results of 99 clinically diagnosed primary BU lesions

Laboratory tests AFB + Culture* + PCR  + All three test evaluable Any one test positive Any two tests positive All three tests positive

Number of lesions

% of lesions

78 ⁄ 99 77 ⁄ 97 62 ⁄ 85

78.8 79.4 72.9

78 ⁄ 83 69 ⁄ 83 43 ⁄ 83

94.0 83.1 51.8

*2 ⁄ 99 cultures were contaminated.  PCR was not done in the case of 14 ⁄ 99 samples.

only AFB and culture performed, 10 tested positive in both, two had positive culture and two tested negative in both. The two patients with contaminated cultures tested positive for both AFB and PCR. Nine cases tested positive only once: two for AFB, two for culture, and five for PCR. The two patients with AFB-only positive both had very extensive ulcers, with one needing repeat excisions for secondary lesions that developed beyond the excised margins subsequently during admission. One of the culture-only-positive patients had an ulcer and the other a nodule. Of the five PCR-onlypositive cases, one was a large plaque in the popliteal fossa 194

and four were ulcers; one of which – a 7-year-old lesion – was devoid of active disease. We assessed the impact of lesion stage on test results, and found a difference in test outcomes between early and late stage lesions for culture, but not for AFB and PCR (Table 4). Culture positive rate was statistically significantly higher for pre-ulcerative lesions than for ulcers (v2 = 4.04, P = 0.04). Two-tests-positive rate for preulcerative lesions was significantly higher than for ulcerated lesions (v2 = 4.50, P = 0.03), whereas there was no difference for one- and three-tests-positive rates. Agreement between the AFB ⁄ culture test-pair was statistically significantly higher than what can be expected by chance [Kappa coefficient = 0.47 (SE 0.10) and P = 0.000001]. The coefficient was equally high for pre-ulcerative [k = 0.42 (SE = 0.13), P = 0.00008] and ulcerated lesions [k = 0.50 (SE = 0.13), P = 0.0001]. In contrast, agreement for the test-pairs AFB ⁄ PCR and culture ⁄ PCR for both pre-ulcerative and ulcerated lesion groups were not statistically significant. Features of clinically diagnosed Buruli ulcer patients with only negative test results Seven (7.1%) of the 99 patients tested negative in all tests: five had all three tests done and two (#N2 and #N3) had only AFB and culture done. All seven patients were living

ª 2008 Blackwell Publishing Ltd

Tropical Medicine and International Health

volume 13 no 2 pp 191–198 february 2008

E. Mensah-Quainoo et al. Clinical and laboratory diagnosis of Buruli ulcer

Table 4 Effect of BU lesion stage on laboratory test results (109 tests in 99 clinically diagnosed patients)

Laboratory tests AFB + Culture* + PCR  + AFB+ and culture+ AFB+ and cultureAFB- and culture+ AFB- and cultureAgreementà AFB ⁄ culture§ AFB+ and PCR+ AFB+ and PCRAFB- and PCR+ AFB- and PCRAgreementà AFB ⁄ PCR Culture+ and PCR+ Culture+ and PCRCulture- and PCR+ Culture- and PCRAgreementà culture ⁄ PCR All three test evaluable Any one test positive Any two tests positive All three tests positive

Number preulcerative lesions

% of preulcerative lesions

Number ulcerated lesions

% of ulcerated lesions

Total number of lesions

40 ⁄ 52 43 ⁄ 50 32 ⁄ 42 36 ⁄ 50 2 ⁄ 50 7 ⁄ 50 5 ⁄ 50 41 ⁄ 50 26 ⁄ 42 7 ⁄ 42 6 ⁄ 42 3 ⁄ 42 29 ⁄ 42 27 ⁄ 40 9 ⁄ 40 3 ⁄ 40 1 ⁄ 40 28 ⁄ 40

76.9 86.0– 76.2 72.0 4.0 14.0 10.0 82.0§2 61.9 16.7 14.3 7.1 69.0 67.5 22.5 7.5 2.5 70.0

45 ⁄ 57 39 ⁄ 56 36 ⁄ 52 36 ⁄ 56 7 ⁄ 56 3 ⁄ 56 8 ⁄ 56 44 ⁄ 56 31 ⁄ 52 9 ⁄ 52 5 ⁄ 52 5 ⁄ 52 36 ⁄ 52 26 ⁄ 51 8 ⁄ 51 9 ⁄ 51 6 ⁄ 51 32 ⁄ 51

78.9 69.6 69.2 64.3 12.5 5.4 14.3 78.6§3 59.6 17.3 9.6 9.6 69.2 51.0 15.7 17.6 11.8 62.7

85 ⁄ 109 82 ⁄ 106 68 ⁄ 94 72 ⁄ 106 9 ⁄ 106 10 ⁄ 106 13 ⁄ 106 85 ⁄ 106 52 ⁄ 94 16 ⁄ 94 11 ⁄ 94 8 ⁄ 94 60 ⁄ 94 53 ⁄ 91 17 ⁄ 91 12 ⁄ 91 7 ⁄ 91 60 ⁄ 91

39 ⁄ 40 35 ⁄ 40 23 ⁄ 40

97.5 87.5** 57.5

47 ⁄ 51 39 ⁄ 51 25 ⁄ 51

92.2 76.5 49.0

86 ⁄ 91 74 ⁄ 91 48 ⁄ 91

% 78.0 77.4 72.3 67.9 8.5 9.4 12.3 80.2§1 55.3 17.0 11.7 8.5 63.8 52.8 18.7 13.2 7.7 65.9 94.5 81.3 52.7

*3 ⁄ 109 cultures were contaminated.  PCR was not done in the case of 15 ⁄ 109 samples. àAgreement: both tests positive, or both negative. §Kappa coefficient: 1k = 0.47 (SE 0.10), P = 0.000001. 2k = 0.42 (SE = 0.13), P = 0.00008. 3k = 0.50 (SE = 0.13), P = 0.0001. –Culture positive rate pre-ulcerative lesions compared to ulcers: v2 = 4.04, P = 0.04. **Two-tests-positive rate for pre-ulcerative lesions compared to ulcers: v2 = 4.50, P = 0.03.

in BU endemic communities and their lesion features, as detailed in Table 5, were clinically consistent with BU, as with the rest of the cases. For instance, the nodule of patient #N1 and the ulcer of #N4 were beside typical sunken scars (Muelder 1992) – an indication of same-site previous BU lesions. Another clinician also independently judged case #N3 as BU and started him on anti-mycobacterial drug treatment. Discussion The clinical diagnosis of BU is considered easy in endemic settings (van der Werf & van der Graaf 1990; Muelder 1992; van der Werf et al. 1999; WHO 2000). Of cases clinically diagnosed in routine practice in the GWD, 94% could be reconfirmed by at least one of three tests performed in an in-country laboratory. To our knowledge, there are no previous reports of a large series assessing clinical diagnostic performance. Over-diagnosis in GWD could include known differentials of BU (WHO 2001a) common in the district such as: abscesses, cellulitis, lipoma,

ª 2008 Blackwell Publishing Ltd

yaws, insect bites, haematoma and traumatic wounds. Other known differential diagnoses such as tropical phagedenic ulcer, venous ulcers, deep fungal infection, leprosy, scrofuloderma, leishmaniasis, lymphoma and malignant melanoma may also occur. Our findings of most lesions on limbs are similar to previous ones (The Uganda Buruli Group 1971; van der Werf et al. 1989; Aguiar & Stenou 1997; Amofah et al. 2002; Noeske et al. 2004), with the left leg and right arm being more affected than the corresponding opposite limbs (van der Werf et al. 1989). Their suggested explanation might hold true for this cohort, as the cultural setting is the same. In contrast, Hospers et al. (2005) found equal distribution on the limbs, but arms more affected with correction for body surface area. Plaque forms appear over-represented for Ghana cases; the reason for this is unclear. But perhaps the definition of clinical forms of the disease needs further refinement to enable comparison of cases between studies. In our study the sensitivity of PCR at 72.3% was comparable to a finding of 74.8% on excised tissue samples by Stienstra et al. (2003), but relatively low, 195

Tropical Medicine and International Health

volume 13 no 2 pp 191–198 february 2008

E. Mensah-Quainoo et al. Clinical and laboratory diagnosis of Buruli ulcer

Table 5 Features of clinically diagnosed BU patients with all laboratory tests negative Case ID

Age

Sex

Clinical form

Key features of patient ⁄ lesion

#N1

22

Male

Nodule

#N2

15

Male

Nodule

#N3

19

Male

Plaque

#N4

9

Female

Ulcer

#N5

65

Female

Ulcer

#N6

2

Female

Ulcer

#N7

9

Male

Ulcer

Patient lives in endemic community Firm lesion attached to skin, but not to underlying tissues Lesion near characteristic previous BU scars Patient lives in endemic community Firm lesion attached to skin, but not to underlying tissues Patient lives in endemic community Firm lesion attached to skin, not attached to underlying tissues initially, but later attached Discoloration of skin overlying lesion (purplish) On referral to the Korle Bu Teaching Hospital, placed on anti-TB drug treatment Patient lives in endemic community Lesion at first a small discharging sinus, later enlarged into a typical BU Necrotic slough in situ Lesion near characteristic previous BU scars Patient lives in endemic community Large ulcer with undermined edges Necrotic sloughing base Indurated surrounding tissues and darkened surrounding skin Patient lives in endemic community Extensive ulcer with undermined edges Necrotic slough in situ Indurated surrounding tissues and darkened surrounding skin Patient lives in endemic community Large ‘old-looking’ ulcer with signs of active disease on edges Indurated surrounding tissues and darkened surrounding skin

compared to 85% found by Guimaraes-Peres et al. (1999), and to a recent publication of 98% by Phillips et al. (2005) who examined punch biopsy specimens. The finding of 78.7% culture sensitivity was comparable to our previous finding of 75.6% (Yeboah-Manu et al. 2004) performed on excised tissue, and high compared to other studies: 54.5% reported by Portaels et al. (1997), 39.2% by Stienstra et al. (2003) and 49% by Phillips et al. (2005). The AFB sensitivity of 78.1% was high compared to 63% reported by Guarner et al. (2003) (in histopathology sections), 42% by Phillips et al. (2005), and nine of 25 cases by Darie et al. (1993); but lower than the 100% (n = 8) reported by Goutzamanis and Gilbert (1995). All three tests gave negative results in a set of samples from 7 patients who had recurrences; this may explain the outcomes for the seven patients with all-negative results. Mycobacterium ulcerans bacilli are distributed unevenly in tissues (Josse et al. 1995; Rondini et al. 2003, 2006; Phillips et al. 2005; Siegmund et al. 2005). Thus, inadequacy in size or site of specimen may result in false negative test outcome. The seven all-test-negative patients fulfilled the standard criteria for clinical diagnosis the same as those lesions confirmed as BU. But whether they were true 196

negatives, that is a clinical diagnostic error or false negatives due to inadequate test sensitivity, remains unclear. On the hand, the likelihood of false positive results must be considered, principally among lesions with only one test positive. A number of pointers suggest a valid BU diagnosis among the nine single-test-positive cases: the two with culture-only positive are unquestionably definite BU cases; the two cases with AFB-only positive had typical BU clinical scenarios and ⁄ or disease course. Three of the PCRonly-positive cases had additional features that lend support to a diagnosis of BU: one had another positive PCR result from independent testing in an external laboratory; one had recurrence and the third, a chronic upper limb lesion, had been diagnosed with newly active BU several years previously. Lastly, ulcers, which are clinically easier to diagnose, were over- rather than under-represented among the one-test-positive cases. For culture, we found a statistically significantly higher test positive rate for pre-ulcerative than ulcerated lesions (P = 0.04). Higher test yield in pre-ulcerative lesions may reflect the presence of the total load of M. ulcerans bacilli, whereas in ulcers the bulk of bacilli may be lost along with sloughing necrotic tissues (Portaels et al. 1997).

ª 2008 Blackwell Publishing Ltd

Tropical Medicine and International Health

volume 13 no 2 pp 191–198 february 2008

E. Mensah-Quainoo et al. Clinical and laboratory diagnosis of Buruli ulcer

The high proportion of pre-ulcerative lesions (48%) in our cohort thus probably contributed to the high culture and AFB yields obtained. Also, for high AFB and culture yield, careful sample collection and handling, as detailed, are of crucial importance. For culture, method of decontamination and culture conditions (Yeboah-Manu et al. 2004) are also important. Laboratory diagnosis to confirm BU cases can either support pre-treatment decision making or assess the accuracy of clinical diagnosis after treatment. Both functions are increasingly important where chemotherapy is introduced (Etuaful et al. 2005; Chauty et al. 2007) and its efficacy needs to be assessed. For infectious diseases, culture of the causative organism is the strongest proof of disease. However, in the case of M. ulcerans culture takes several weeks and is thus impractical to aid pre-treatment diagnosis. Besides, it requires a specialised laboratory and is therefore inaccessible at peripheral centres. Assay formats suitable for peripheral health facilities, such as an agglutination or dipstick test, would be preferable over all currently available tests. Such assays are available for many infectious diseases and test formats could be easily adopted once a test principle suitable for BU has been identified. We encourage further research to develop highly predictive but simple diagnostic tests to aid fast pre-treatment confirmation of M. ulcerans infection at peripheral treatment centres in highly endemic countries. We advocate, based on our findings, accepting one positive test result in conjunction with clinical diagnosis done by well-trained clinicians as confirmation of BU cases. With this approach, the trade-off fraction of non-BU cases that would be counted as BU due to false single-testpositive results would probably be minimal, whereas more true BU cases would be identified. Given these considerations, we recommend training of peripheral level health workers to optimise clinical diagnosis of BU. For this, the predictive values of objective clinical features of BU must be further investigated. To sustain and monitor the quality of clinical diagnosis at the periphery, intermittent quality-control testing should be performed, preferably by in-country reference laboratories. Our findings provide an indication of clinical diagnostic reliability. However, due to the retrospective nature of the study, data that would have permitted sensitivity and negative predictive value analysis was unavailable. Nevertheless, the results of the analysis provide an indication of the degree of specificity achievable with clinical judgment. A prospective study design will be required to test the overall accuracy of clinical judgment more formally. In conclusion, our findings partly validate the

ª 2008 Blackwell Publishing Ltd

perception that clinical diagnosis is reliable and give an indication of the degree of accuracy achievable in routine practice in an endemic country of Africa. Acknowledgements We thank the patients and the Government of Ghana. We are indebted to the hardworking team at Amasaman Health Centre, Ghana, and members of the District Health Management Team. In addition, many individuals deserve our gratitude for various contributions to this work, in particular Drs Hector Addo, Moses Adibo, Charles SagoeMoses, Fabian Mork, and Mr. Abdulai Bukari; and Profs Tom Smith, Fred Binka and Samuel Ofosu Amaah. We are grateful to the Stanley-Thomas Johnson Foundation for partially funding this work. References Amofah G, Bonsu F, Tetteh C et al. (2002) Buruli ulcer in Ghana: results of a national case search. Emerging Infectious Diseases 8, 167–170. Asiedu K & Etuaful S (1998) Socioeconomic implications of Buruli ulcer in Ghana: a three-year review. American Journal of Tropical Medicine and Hygiene 59, 1015–1022. Bretzel G, Siegmund V, Racz P et al. (2005) Post-surgical assessment of excised tissue from patients with Buruli ulcer disease: progression of infection in macroscopically healthy tissue. Tropical Medicine and International Health 10, 1199– 1206. Chauty A, Ardant MF, Adeye A et al. (2007) Promising clinical efficacy of the combination streptomycin–rifampin for the treatment of Buruli ulcer (Mycobacterium ulcerans disease). Antimicrobial Agents and Chemotherapy 11, 4029–4035. Coloma JN, Navarrete-Franco G, Iribe P & Lopez-Cepeda LD (2005) Ulcerative cutaneous mycobacteriosis due to Mycobacterium ulcerans: report of two Mexican cases. International Journal of Leprosy and Other Mycobacterial Diseases 73, 5–12. Darie H, Le Guyadec T & Touze JE (1993) Epidemiological and clinical aspects of Buruli ulcer in Ivory Coast. 124 recent cases. Bulletin of the Exotic Pathology Society 86, 272–276. Debacker M, Aguiar J, Steunou C et al. (2004) Mycobacterium ulcerans disease (Buruli ulcer) in rural hospital, Southern Benin, 1997–2001. Emerging Infectious Diseases 10, 1391– 1398. Etuaful S, Carbonnelle B, Grosset J et al. (2005) Efficacy of the combination rifampin-streptomycin in preventing growth of Mycobacterium ulcerans in early lesions of Buruli ulcer in humans. Antimicrobial Agents and Chemotherapy 49, 3182– 3186. Evans MR, Phillips R, Etuaful SN et al. (2003) An outreach education and treatment project in Ghana for the early stage of Mycobacterium ulcerans disease. Transactions of the Royal Society of Tropical Medicine and Hygiene 97, 159–160.

197

Tropical Medicine and International Health

volume 13 no 2 pp 191–198 february 2008

E. Mensah-Quainoo et al. Clinical and laboratory diagnosis of Buruli ulcer

Goutzamanis JJ & Gilbert GL (1995) Mycobacterium ulcerans infection in Australian children: report of eight cases and review. Clinical Infectious Diseases 21, 1186–1192. Guarner J, Bartlett J, Whitney EA et al. (2003) Histopathologic features of Mycobacterium ulcerans infection. Emerging Infectious Diseases 9, 651–656. Guimaraes-Peres A, Portaels F, de Rijk P et al. (1999) Comparison of two PCRs for detection of Mycobacterium ulcerans. Journal of Clinical Microbiology 37, 206–208. Johnson RC, Makoutode M, Hougnihin R et al. (2004) Traditional treatment for Buruli ulcer in Benin. Medecine Tropicale: Revue du Corps de Sante Colonial 64, 145–150. Josse R, Guedenon A, Darie H et al. (1995) Mycobacterium ulcerans cutaneous infections: Buruli ulcers. Medecine Tropicale: Revue du Corps de Sante Colonial 55, 363–373. Kanga JM & Kacou ED (2001) Epidemiological aspects of Buruli ulcer in Cote d’Ivoire: results of a national survey. Bulletin of the Exotic Pathology Society 94, 46–51. Muelder K (1992) Wounds that will not heal: the Buruli ulcer. International Journal of Dermatology 31, 25–26. Mwanatambwe M, Yajima M, Etuaful S et al. (2002) Malignant melanoma of the buttock presenting as a Buruli ulcer. Journal of Dermatology 29, 357–361. Phillips R, Horsfield C, Kuijper S et al. (2005) Sensitivity of PCR targeting the IS2404 insertion sequence of Mycobacterium ulcerans in an assay using punch biopsy specimens for diagnosis of Buruli ulcer. Journal of Clinical Microbiology 43, 3650–3656. Portaels F, Agular J, Fissette K et al. (1997) Direct detection and identification of Mycobacterium ulcerans in clinical specimens by PCR and oligonucleotide-specific capture plate hybridization. Journal of Clinical Microbiology 35, 1097–1100. Revill WD, Morrow RH, Pike MC & Ateng J (1973) A controlled trial of the treatment of Mycobacterium ulcerans infection with clofazimine. Lancet 2, 873–877. Rondini S, Mensah-Quainoo E, Troll H, Bodmer T & Pluschke G (2003) Development and application of real-time PCR assay for quantification of Mycobacterium ulcerans DNA. Journal of Clinical Microbiology 41, 4231–4237. Rondini S, Horsfield C, Mensah-Quainoo E et al. (2006) Contiguous spread of Mycobacterium ulcerans in Buruli ulcer lesions

analysed by histopathology and real-time PCR quantification of mycobacterial DNA. Journal of Pathology 208, 119–128. Semret M, Koromihis G, MacLean JD, Libman M & Ward BJ (1999) Mycobacterium ulcerans infection (Buruli ulcer): first reported case in a traveler. American Journal of Tropical Medicine and Hygiene 61, 689–693. Siegmund V, Adjei O, Racz P et al. (2005) Dry-reagent-based PCR as a novel tool for laboratory confirmation of clinically diagnosed Mycobacterium ulcerans-associated disease in areas in the tropics where M. ulcerans is endemic. Journal of Clinical Microbiology 43, 271–276. Stienstra Y, van der Werf TS, Guarner J et al. (2003) Analysis of an IS2404-based nested PCR for diagnosis of Buruli ulcer disease in regions of Ghana where the disease is endemic. Journal of Clinical Microbiology 41, 794–797. The Uganda Buruli Group (1970) Clinical features and treatment of pre-ulcerative Buruli lesions (Mycobacterium ulcerans infection). Report II of the Uganda Buruli Group. British Medical Journal 2, 390–393. van der Werf TS & van der Graaf WT (1990) Buruli ulcer in West Africa. Lancet 336, 1440. van der Werf TS, van der Graaf WT, Tappero JW & Asiedu K (1999) Mycobacterium ulcerans infection. Lancet 354, 1013– 1018. WHO (2000) Buruli ulcer. Mycobacterium ulcerans infection, 1st edn. (WHO ⁄ CDS ⁄ CPE ⁄ GBUI ⁄ 2000) WHO, Geneva. WHO (2001a) Buruli ulcer: diagnosis of Mycobacterium ulcerans disease: In: Buruli Ulcer: A Manual for Health Care Providers, 4th edn (WHO ⁄ CDS ⁄ CPE ⁄ GBUI ⁄ 2001) WHO, Geneva, pp. 4– 92. WHO (2001b) Buruli ulcer; management of Mycobacterium ulcerans disease; a manual for health care providers, 3rd edn. (WHO ⁄ CDS ⁄ CPE ⁄ GBUI ⁄ 2001) WHO, Geneva. WHO (2004) Provisional guidance on the role of specific antibiotics in the management of Mycobacterium ulcerans disease (Buruli ulcer), 10th edn. (WHO ⁄ CDS ⁄ CPE ⁄ GBUI ⁄ 2004) WHO, Geneva. Yeboah-Manu D, Bodmer T, Mensah-Quainoo E et al. (2004) Evaluation of decontamination methods and growth media for primary isolation of Mycobacterium ulcerans from surgical specimens. Journal of Clinical Microbiology 42, 5875–5876.

Corresponding Author Gerd Pluschke, Swiss Tropical Institute, Socinstrasse 57, 4002 Basel, Switzerland. Tel.: +41 61 2848235; Fax: +41 61 2848101; E-mail: [email protected]

198

ª 2008 Blackwell Publishing Ltd