SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Summary

Background

Diagnosis of Whipple's disease, a rare systemic infection affecting predominantly the small bowel, is based on the identification of the bacterium Tropheryma whipplei.

Aims

To make explicit diagnostic uncertainties in Whipple's disease through a decision analysis, considering two different clinical scenarios at presentation.

Methods

Using appropriate software, a decision tree estimated the consequences after testing different strategies for diagnosis of Whipple's disease. Probabilities and outcomes to determine the optimum expected value were based on MEDLINE search.

Results

In patients with clinically-predominant intestinal involvement, diagnostic strategies considering intestinal biopsy for histology (including appropriate staining) and the polymerase chain reaction testing for bacterial DNA were similarly effective. In case of failure of one procedure, the best sequential choice was a polymerase chain reaction analysis after a negative histology. Of the five strategies tested for cases with predominant focal neurological involvement, the stereotaxis cerebral biopsy evidenced the highest expected value. However, using quality-adjusted life-years considering the morbidity of methods, intestinal biopsy for PCR determination was the best choice.

Conclusions

In patients with Whipple's disease having predominant digestive involvement, intestinal biopsies for histology should be indicated first and, if negative, a bacterial polymerase chain reaction determination should be the next option. Although the molecular polymerase chain reaction assessment of cerebral biopsies has the highest diagnostic yield in neurological Whipple's disease, its associated morbidity means that analyses of intestinal samples are more appropriate.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Whipple's disease (WD) is a rare multisystem infectious disease with protean clinical manifestations. Although a bacterial cause was suspected for many years, a bacterium, classified as an Actinomycete and named Tropheryma whipplei, has only been recently identified.1–8 The disease is characterized by an insidious onset followed by highly variable clinical manifestations. Weight loss, chronic diarrhoea, arthralgias and low-grade fever are the characteristic features found in most patients.5, 8 Although clinical gastrointestinal (GI) involvement is very common, atypical clinical forms are increasingly recognized.5, 7–9 In this context, central nervous system (CNS) involvement occurs in up to 43% of cases,5, 10–12 and may be the sole clinical manifestation of the disease.13, 14 As CNS involvement has a poor prognosis and carries the highest risk for relapse, treatment should be rapidly instituted.15–17 Antibiotic treatment is mandatory, which results in a rapid clinical improvement and remission in most patients.18–22 This treatment has been shown to reduce the number of relapses and is also effective in prevention and/or treatment of the neurological aspects of the disease.16, 19, 22

The diagnosis of WD can be difficult and requires the histological assessment of diseased tissue showing the presence of foamy appearing vacuolated macrophages that infiltrate the lamina propria. These macrophages are periodic acid–Schiff (PAS) positive and Ziehl–Nielsen (ZN) negative.9, 10, 23 Although PAS-positive staining is also observed with other predominantly intracellular infectious agents such as Rhodococcus equi, Histoplasma capsulatum, Bacillus cereus and others, 10, 24 only Mycobacterium avium-intracellulare in patients with acquired immunodeficiency syndrome presents a practical problem solved by ZN staining where, different from WD bacillus, M. avium-intracellulare is acid-fast.10, 21 Confirmatory tests such as electron microscopy and/or the bacterial DNA molecular identification using polymerase chain reaction (PCR) have special utility in these cases.23, 25, 26 Although histological examination of the small bowel is a common diagnostic test, no abnormality is found in about 30% of patients.5, 7, 9, 23

Polymerase chain reaction of the 16S ribosomal RNA of T. whipplei has now become an important test. It is useful in duodenal biopsies or biopsies of any other tissues involved, including the CNS.25–28 However, some laboratories have reported a number of potential false-positive results from control individuals without clinical evidence of WD, such as in saliva (35% in one series), in duodenal biopsy (4.8%) and gastric juice (11.4%), suggesting that the organism may be a normal commensal agent.25, 28–31

All these diagnostic uncertainties make WD an interesting clinical dilemma for decision analysis techniques.32–34 These represent a method for synthesizing medical facts (probabilities) and human values helping to determine the best course of action. Decision analysis can be used to provide guidelines for managing patients with similar clinical features based on Bayesian probabilities and those values assigned to different outcomes.34 Thus, it can offer a rational means for allowing the medical community to move from finding evidence to implementing it.32, 33 Our goal in this study was to evaluate diagnostic uncertainties in WD using decision analysis techniques in two very different and challenging clinical situations represented by: (i) a patient with predominant intestinal involvement; and (ii) a patient predominantly with a neurological presentation.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We constructed a decision analysis model to estimate the clinical consequences of diagnostic procedures currently used for the diagnosis of WD. The various decision points and their consequences, including their associated probabilities and utilities, were mapped to form a tree as a visual representation of the decisions available. Probabilities and outcomes to estimate the optimum expected values were assembled (using Data 32 TreeAge Software Inc., Williamstown, MA, USA) to focus on different strategies.

Data sources

In order to estimate probabilities, a review of the literature involving a computerized search and a review of reference lists from retrieved papers was undertaken. MEDLINE searches through July 2005 were conducted using the terms ‘Whipple's disease’, ‘neurological Whipple's disease’, combined with ‘diagnosis’ (sensitivity and specificity), ‘treatment’ and ‘reviews’. While we identified 1256 different papers, all relevant publications were retrieved and pertinent data extracted. Probability data were obtained from papers reporting series of cases or from narrative reviews. Expert opinions were also taken into account when no data was found or when discrepancies resulted in the literature.

Strategies

Different strategies were considered for the two clinical cases: (i) a 55-year-old man with chronic diarrhoea, weight loss, fever and arthralgias, on the one hand; and (ii) a 55-year-old man with neurological symptoms (oculomasticatory myorhythmia, myoclonus and epilepsy) and fever but no digestive signs, on the other. For the first case, three alternatives were used: (i) watchful waiting, (ii) endoscopic intestinal biopsies for histology including PAS and ZN staining; and (iii) PCR testing for bacterial DNA in biopsy samples. In the neurological case, five strategies were introduced: (i) watchful waiting, (ii) endoscopic intestinal biopsy for histology (including PAS and ZN staining), (iii) PCR determination of either intestinal biopsy samples, (iv) cerebrospinal fluid (CSF) or (v) stereotactic CNS biopsies. When a diagnosis of WD was reached, specific antimicrobial therapy was considered. If treatment was successful, the remission rate was estimated. Remission was defined as a resolution of the symptoms, with a return to normal life styles and equal life expectancy as a person of the same age without WD.

Outcome measures

The outcome of interest was life expectancy. It was estimated as 20 years for a healthy 55-year-old man, according to life expectancy in Argentina.35 As there are no validated relevant outcome measures for WD, a hypothetical average survival rate was estimated as 7 years from the time of first visit, according to the clinical descriptions in reported cases. This was the outcome assigned when tests were negative and, consequently, no therapy administered. In neurological WD, it was possible to tailor diagnostic approaches according to patients’ preferences assuming that they might perceive endoscopy as being ‘better’ or more comfortable than brain biopsy or lumbar puncture, which are more invasive. This allowed us to propose alternatives base upon quality-adjusted life-years (QALYs).33, 34 QALYs reflect length of life adjusted for the relative desirability of a health state, where a value of 1 was assigned to represent perfect health and 0 to death.34 QALYs were calculated using the following formula: assigned preference × life expectancy in perfect health.34

Sensitivity analysis

To determine the robustness and the impact of varying inputs on the model structure of the decision trees, a sensitivity analysis using plausible ranges based on the lowest and highest numbers reported in different studies was performed on all probability and outcome values.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Patient with clinically predominant intestinal involvement

According to our assessment of the available literature, a sensitivity of 0.78 was assigned to the intestinal histology with PAS and ZN staining, and 0.84 for the bacterial DNA identification on intestinal samples by PCR. Based on published data and expert consensus, we estimated a 0.80 probability for remission when diagnosis is obtained and treatment with antibiotics is implemented (Tables 1 and 2). We estimated 20- and 7-year life expectancy outcomes for responders and non-responders to antibiotic therapy respectively (see Materials and methods). Thus, the analysis yielded identical values for both histology and PCR methods (17.0 survival years) (Figure 1). This finding suggests that either staining or PCR are essentially equivalent in terms of outcome. However, if the first diagnostic test was negative, a stepwise approach was utilized to determine the best alternative to adopt. It appeared that the most useful sequence of testing was for histological analysis followed by molecular analysis (15.7 vs. 15.1 survival years). Therefore, in a classical patient with symptoms suggesting small intestinal WD it would be appropriate to start with an upper endoscopy and duodenal biopsies for histological assessment. If these results were negative then this should be followed by PCR for detecting bacterial DNA.

Table 1.  Sensitivity rates for the different diagnostic methods in WD
 Intestinal WDNeurological WDCSFCNS biopsy
Intestinal histologyIntestinal PCRIntestinal histologyIntestinal PCR
  1. WD, Whipple's disease; PCR, polymerase chain reaction; CSF, cerebrospinal fluid; CNS, central nervous system.

Baseline probability (%)788471808087
Range (used in the sensitivity analysis)70–8580–9866–7579–9075–8575–90
References used5, 8, 10, 21, 2810, 21, 25, 26, 2810, 21, 28, 3810, 11, 28, 3827, 28, 3810, 11, 28, 38
Table 2.  Rate of response to antibiotic treatment as it was reported in the literature
 Response in predominant intestinal involvementResponse in neurological Whipple's disease
Baseline probability (%)8070
Range (used in the sensitivity analysis)65–9060–80
References used5, 8, 10, 16, 17, 20, 2510, 21, 22
image

Figure 1. Decision tree for diagnosis of Whipple's disease in a patient with gastrointestinal-predominant symptoms.

Download figure to PowerPoint

Patient with a predominant CNS involvement

In patients with mainly neurological symptoms at the time of diagnosis of WD, all available data suggest that the sensitivity for the diagnostic methods is different. Thus, a sensitivity of 0.71, 0.80, 0.80 and 0.87 was calculated for intestinal histology with PAS and ZN staining, PCR determination in intestinal samples, PCR in CSF and in cerebral biopsy respectively (Table 1). Furthermore, a remission rate of 0.70 was considered for this atypical clinical picture, a value which is lower than that observed for cases with predominant intestinal involvement (Table 2). This lower value is related to, at least, three observations: (i) access to the CNS is restricted for certain antibiotics,16 (ii) the documented persistence of minor symptoms despite a positive therapeutic response;23 and (iii) the irreversible nature of some neurological symptoms.10 As expected, ‘watchful waiting’ had a negative impact on outcome. The best expected value was for the PCR analysis of CNS biopsies yielding 14.9 survival years (Figure 2). Moreover, PCR analysis of either CSF or intestinal samples had a similar but lower expected value (14.2 survival years). Finally, intestinal histology with PAS and ZN staining had the lowest value (13.4).

image

Figure 2. Decision tree for diagnosis of Whipple's disease in a patient with neurological symptoms at presentation.

Download figure to PowerPoint

Sensitivity analysis

The robustness and the impact of varying inputs on the model structure of the decision trees were challenged in both clinical scenarios. In the model designed for the intestinal involvement, sensitivity probabilities were tested from 0.70 to 0.85 for histology and from 0.80 to 0.98 for PCR, and choices were the same as in the basal analysis. Similar results were obtained when the response to therapy was elevated to 0.90 and when the non-responder to therapy branch was tested changing expected survival years from 2 to 15 years. These findings suggest that the structure of this tree is robust and firm.

On the contrary, in the case of a patient with neurological WD at diagnosis, the risks and discomforts of diagnostic methods impacted on the results, rendering a weak and unstable tree. The diagnostic approach linked to patients’ preference was conditioned to test morbidities based on their eventual complications. In this context, while the reported morbidity for upper endoscopy and biopsy is about 0.03%,36 that for CNS biopsy is 9%.37 Considering the impact on quality of life, patients’ preferences were estimated in 0.98, 0.90 and 0.85 for endoscopy, lumbar puncture and CNS biopsy, respectively, resulting in 19.5, 18 and 17 QALYs for outcome instead of 20 years of life expectancy in perfect health. This approach was accompanied by a change in the expected values determining that the upper endoscopy with duodenal biopsies for PCR analysis is the method of choice (14.0 survival years), followed by the histological assessment of the intestinal biopsy (13.2), the PCR assessment of CSF (13.2) and, finally, the PCR in CNS biopsy samples (13.1) (Figure 3).

image

Figure 3. Decision tree for diagnosis of Whipple's disease with neurological involvement using quality-adjusted life-years.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Whipple's disease is a very interesting disorder where diagnostic tests can be misused.9, 10, 28 The clinical suspicion of WD has to be confirmed by procedures applied on specimens obtained either from the small bowel or from affected tissues such as CNS or CSF.12, 13, 15 As stated before, all diagnostic tools have limited sensitivity and specificity, and in the case of those devoted to the identification of T. whipplei, these may be challenging. Furthermore, as its clinical presentation may be variable with a potentially lethal course if untreated, a carefully developed diagnostic algorithm is essential.20, 21, 28 In the present study, we used a decision model to estimate the clinical benefits of different diagnostic strategies in hypothetical patients with two different clinical presentations. In a decision analysis, the optimal operating point depends upon the pretest probability of disease, the expected net benefit of correctly diagnosing the disease (true positives) and the expected net harms associated with false-positive results.32–34 We came across these difficulties while creating the first scenario, which was based on a middle-age patient with WD presenting with symptoms suggestive of predominant GI involvement. Because the prevalence of WD is extremely low, the pretest probability for a given patient is impossible to estimate and the quantitative comparative assessment of the different diagnostic methods is limited. Furthermore, as there is no gold standard for the diagnosis of WD, sensitivity and specificity are difficult to assess in both, histological studies and PCR determinations. The information used to create the decision model was based on published series of cases or reviews, combined with expert criteria. As these estimates are subject to bias, the use of a sensitivity analysis based on a wide range of possible values was mandatory. Although different probabilities and outcomes were introduced in the sensitivity analyses, equivalent results were obtained in every run of the tree. Based on these results, we suggest that the diagnostic work-up for patients presenting with GI symptoms suggestive of WD should follow a sequential approach beginning with mucosal histology (including PAS and ZN staining), and if negative followed by PCR amplification of bacterial DNA in the biopsy.

The diagnosis of CNS WD remains difficult and it is often delayed in the absence of the systemic manifestations of the disease.11, 12, 14, 15, 38 The clinical picture is often non-specific and similar to that of vasculitis or other subacute or chronic encephalopathies, or tumours.38 The diagnosis can be obtained by looking for PAS-positive cells and/or bacterial DNA in the CSF or in the CNS. Furthermore, the identification of T. whipplei by either or both the molecular analysis or the conventional histology and staining in intestinal samples can diagnose WD even in the absence of overt GI clinical involvement.28, 38 Thus, the second scenario proposed in our decision analysis was included to determine the best diagnostic approach for those patients with neurological WD, but without overt GI symptoms. According to the data reported in the literature, the sensitivity for molecular and histological procedures on intestinal samples of patients without overt intestinal involvement is lower than that assessed in patients with clinical GI involvement.15, 21, 28, 38 Thus, based on probabilities and outcomes we observed that the procedure with the highest expected value was the CNS biopsy followed by the examination of CSF. However, the associated morbidity of methods impacting on quality of life might be expected to alter both physicians’ and patients’ choices. For instance, upper GI endoscopy with biopsy has less morbidity than either CSF extraction or stereotactic cerebral biopsy.36, 37 Therefore, when QALYs were introduced into the decision tree, intestinal sampling for PCR or histology was demonstrated to be a better choice for diagnosis.

In conclusion, we attempted to develop strategies that could be employed for the diagnosis of WD in two different clinical situations using decision analysis models. Our data allowed us to conclude that for an intestinal presentation of WD, intestinal biopsy and histology was a reasonable first test. If it was negative, then it should be followed by PCR testing of the biopsy. We believe that this mirrors conventional practice in most institutions. However, in patients with neurological presentations of disease there is a discrepancy. While tissue sampling of the CNS has the highest diagnostic yield, it also has the highest associated morbidity. Using decision analysis, we have demonstrated that although intestinal biopsy has a lower diagnostic yield it is the preferred strategy in these patients.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The authors are indebted to Professor J. Meddings from the University of Calgary, Canada, for his aid and helpful review of the manuscript.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    Wilson KH, Blitchington R, Frothingham R, et al. Phylogeny of the Whipple's-disease-associated bacterium. Lancet 1991; 338: 4745.
  • 2
    Relman DA, Schmidt TM, MacDermott RP, Falkow S. Identification of the uncultured bacillus of Whipple's disease. N Engl J Med 1992; 327: 293301.
  • 3
    Maiwald M, Schuhmacher F, Ditton H, von Herbay A. Environmental occurrence of the Whipple's disease bacterium (Tropheryma whippelii). Appl Environ Microbiol 1998; 64: 7602.
  • 4
    Dutly F, Altwegg M. Whipple's disease and ‘Tropheryma whippelii. Clin Microbiol Rev 2001; 14: 56183.
  • 5
    Ratnaike RN. Whipple's disease. Postgrad Med J 2000; 76: 7606.
  • 6
    Von Herbay A, Otto HF, Stolte M, et al. Epidemiology of Whipple's disease in Germany: analysis of 110 patients diagnosed 1965–1995. Scand J Gastroenterol 1997; 32: 527.
  • 7
    Bai JC, Mota AH, Mauriño E, et al. Class I, Class II HLA antigens in a homogeneous Argentinian population with Whipple's disease: lack of association with HLA-B 27. Am J Gastroenterol 1991; 86: 9924.
  • 8
    Durand DV, Lecomte C, Cathebras P, Rousset A, Godeau P. Whipple disease. Clinical review of 52 cases. Medicine (Baltimore) 1997; 76: 17084.
  • 9
    Dobbins WO. The diagnosis of Whipple's disease. N Engl J Med 1995; 332: 3902.
  • 10
    Bai J, Mazure R, Vazquez H, et al. Whipple's disease. Clin Gastroenterol Hepatol 2004; 2: 84960.
  • 11
    Louis ED, Lynch T, Kaufmann P, Fahn S, Odel J. Diagnostic guidelines in central nervous system Whipple's disease. Ann Neurol 1996; 40: 5618.
  • 12
    Lynch T, Odel J, Fredericks DN, Luis ED, Forman S. Polymerase chain reaction-based detection of T. whippelii in central nervous system in Whipple's disease. Ann Neurol 1997; 42: 1204.
  • 13
    Verhagen WI, Huygen PL, Dalman JE, Schuurmens MM. Whipple's disease and the central nervous system. A case report and a review of the literature. Clin Neurol Neurosurg 1996; 98: 299304.
  • 14
    Schnider P, Trattnig S, Kollegger H, Auff E. MR of cerebral Whipple disease. Am J Neuroradiol 1995; 16: 13289.
  • 15
    De Coene B, Gilliard C, Indekeu P, Duprez T, Trigaux JP. Whipple's disease confined to the central nervous system. Neuroradiology 1996; 38: 3257.
  • 16
    Keinath RD, Merrell DE, Vlietstra R, Dobbins WO III. Antibiotic treatment and relapse in Whipple's disease: long term follow-up of 88 patients. Gastroenterology 1985; 88: 186773.
  • 17
    Feurle GE, Marth T. An evaluation of antimicrobial treatment for Whipple's disease: tetracycline versus trimethoprim-sulfamethoxazole. Dig Dis Sci 1994; 39: 16428.
  • 18
    Petrides PE, Muller-Hocker J, Fredricks DN, Relman DA. PCR analysis of T. whippelii DNA in a case of Whipple's disease: effects of antibiotics and correlation with histology. Am J Gastroenterol 1998; 93: 157982.
    Direct Link:
  • 19
    Bai JC, Sen L, Diez R, et al. Impaired monocyte function in patients successfully treated for Whipple's disease. Acta Gastroenterol Latinoam 1996; 26: 859.
  • 20
    Bai JC, Crosetti EE, Mauriño EC, Martinez CA, Sambuelli A, Boerr LA. Short-term antibiotic treatment in Whipple's disease. J Clin Gastroenterol 1991; 13: 3037.
  • 21
    Franca MC, Castro R, Balthazar M, et al. Whipple's disease with neurological manifestation: case report. Arq Neuropsiquiatr 2004; 62: 3426.
  • 22
    Cooper GS, Blades EW, Remler BF, Salata RA, Bemment KW, Jacobs GH. Central nervous system Whipple's disease: relapse during therapy with trimethoprim-sulfamethoxazole and remission with cefixime. Gastroenterology 1994; 106: 7826.
  • 23
    Marth T, Raoult D. Whipple's disease. Lancet 2003; 361: 239246.
  • 24
    Strom RL, Gruninger RP. AIDS with Mycobacterium avium-intracellulare lesions resembling those in Whipple's disease. N Engl J Med 1983; 309: 13234.
  • 25
    Ramzan N, Loftus E, Burgart L, et al. Diagnosis and monitoring of Whipple's disease by polymerase chain reaction. Ann Intern Med 1997; 126: 5207.
  • 26
    Muller C, Peterman D, Stain C, et al. Whipple's disease: comparison of histology with diagnosis based on polymerase chain reaction in four consecutive cases. Gut 1997; 40: 4257.
  • 27
    von Herbay A, Ditton HJ, Scuhmacher F, Maiwald M. Whipple's disease: staging and monitoring by cytology and polymerase chain reaction analysis of cerebrospinal fluid. Gastroenterology 1997; 113: 43441.
  • 28
    Müller SA, Vogt P, Altwegg M, Seebach JD. Deadly diagnostic carrousel or difficult interpretation of new diagnostic tools for Whipple's disease: case report and review of the literature. Infection 2005; 33: 3942.
  • 29
    Dutly F, Hinrikson HP, Seidel T, Morgenegg S, Altwegg M, Bauerfeind P. Tropheryma whippelii DNA in saliva of patients with Whipple's disease. Infection 2000; 28: 21922.
  • 30
    Ehrbar HU, Bauernfeind P, Dutly F, Koelz HR, Altwegg M. PCR-positive test for Tropheryma whippelii in patients without Whipple's disease. Lancet 1999; 353: 2214.
  • 31
    Street S, Donoghue HD, Neild GH. Tropheryma whippelii in saliva of healthy people. Lancet 1999; 354: 11789.
  • 32
    Pauker S, Kassirer J. The threshold approach to clinical decision making. N Engl J Med 1980; 302: 110917.
  • 33
    Naglie G, Krahn MD, Naimark D, Redelmeier DA, Detsky AS. Primer on medical decision analysis: part 3 – estimating probabilities and utilities. Med Decis Making 1997; 17: 13641.
  • 34
    Hunink M, Glasziou P. Valuing outcomes. In: HuninkM, GlasziouP, ed. Decision Making in Health and Medicine. Cambridge: Cambridge University Press, 2001: 88127.
  • 35
    Organización Paramericana de la Salud, Argentina. Informe Anual 2004. 2004, http://www.ops.org.ar/Files/indicadores%202005.pdf .
  • 36
    Feldman M, Friedman L, Sleisenger M. Sleisenger and Fordtran. Gastrointestinal Disease 2002, 7th edn. Philadelphia: WB Saunders Company, 2004: 5434.
  • 37
    McGirt MJ, Woodwith G, Coon A, et al. Independent predictors of morbidity after image-guided stereotactic brain biopsy: a risk assessment of 270 cases. J Neurosurg 2005; 102: 897901.
  • 38
    Peters G, Du Plessis DG, Humphrey P. Cerebral Whipple's disease with a stroke-like presentation and cerebrovascular pathology. J Neurol Neurosurg Psychiatry 2002; 73: 3369.