886 Gut, 1990,31, 886-889 Investigation of seemingly pathogen-negative

diarrhoea in patients infected with HIV1 Gut: first published as 10.1136/gut.31.8.886 on 1 August 1990. Downloaded from

G M Connolly, A Forbes, B G Gazzard

Abstract Patients Thirty three consecutive patients infected by Thirty three consecutive male patients (median human immunodeficiency virus type 1 (HIV1) age 28 years; range 18-54) with serum antibodies with persistent diarrhoea which remained un- to HIVI and with undiagnosed diarrhoea were diagnosed after microbiological examination studied. All had passed more than three loose of six stool samples and rectal histology were stools daily for at least one month, with no investigated for malabsorption. All had xylose microbiological diagnosis established from rectal and Schilling tests, distal duodenal biopsy, histology and at least six stool examinations comprehensive barium studies, micro- performed in a microbiology laboratory biological examination of six further stool accustomed to the particular pathogens associ- samples, and repeat rectal histology. A micro- ated with this patient population. The CDC biological or histological diagnosis ofinfection criteria for full AIDS3 were fulfilled by 16 of the was made in 12 patients (multiple organisms in 33 patients at the time of presentation (group IV three). Cryptosporidia were identified on five C1 in each case); the remaining 17 (by definition) occasions, cytomegalovirus on four, Giardia had HIV disease group IV A. lamblia on two, and herpes simplex, Campy- lobacter jejuni, Salmonella enteritidis, and Entamoeba histolytica once each. No Methods organism was found when weight loss was less Daily stool volume (mean of three days) and than 5 kg or stool volume less than 400 ml/day estimated weight loss from usual premorbid (n=9). Pathogens were identified in nine of 13 weight (checked against ideal 'weight for height' patients (69%) with weight loss greater than 10 in cases of premorbid obesity or doubt) were kg and stool volume more than 800 ml/day. recorded. Upper gastrointestinal endoscopy for

Barium studies were normal except for ileal distal duodenal biopsy, barium small bowel http://gut.bmj.com/ flocculation in two patients with cryptospori- follow through examination, and double contrast diosis. Evidence for malabsorption existed in barium enema were performed using standard 24 patients - impaired xylose absorption methods. (n= 19) and abnormal Schilling test (n=21). Of Serum concentrations of vitamin B- 12 and the patients with a severely abnormal Schilling serum and red cell folate concentrations were test, a pathogen was identified in 11 (79%) measured, and double labelled cobalt Schilling

(including all five with cryptosporidia, and two tests carried out (normal greater than 11% on September 23, 2021 by guest. Protected copyright. of the patients with only moderate diarrhoea excreted). Xylose absorption testing (5 g) was and weight loss). A simple scoring system performed with assay of a five hour urine collec- based on degree of weight loss and Schilling tion (normal greater than 1-4 g excreted). test result may help to identify the HIV positive Sixfurtherstool sampleswere examined micro- patient with seemingly pathogen-negative biologically with special attention to oppor- diarrhoea in whom further investigations are tunistic pathogens; accordingly the modified likely to show a specific cause. Ziehl-Nielsen stain and use of both Lowenstein- Jensen and Kirchner media were used routinely as previously described.45 Diarrhoea is a common symptom in AIDS Rectal and duodenal histology were stained patients and occurs in up to 50% at some time with haematoxylin and eosin, and with Ziehl- during their illness.' In a proportion of patients Nielsen stain. Inclusion bodies and an inflam- no pathogenic cause of their diarrhoea is found, matory response were required for a histological and it has been hypothesised that direct human diagnosis of cytomegalovirus infecton. immunodeficiency virus type 1 (HIV1) infection Receiver operating characteristics curves6 of the gut mucosa leading to malabsorption may were constructed for each ofthe precisely quanti- be responsible.2 The frequency of finding a fiable parameters measured. Sensitivity, specifi- Departments of pathogen obviously increases with the number of and values were defined and Gastroenterology, city, predictive St Stephen's and tests performed but the clinician needs advice to calculated conventionally.7 Westminster Hospitals, decide when further investigation is unlikely to A scoring system based on Bayes' theorem' London yield diagnostic information. We prospectively was developed to assist early identification of G M Connolly A Forbes studied a group of 33 patients in whom no those in whom an infective aetiology for diar- B G Gazzard pathogen was detected despite examination of rhoea would eventually be found. Correspondence to: six stool samples and rectal biopsy specimen. Dr B G Gazzard, Department Further histo- of Gastroenterology, microbiological analysis, Westminster Hospital, pathology, and tests of malabsorption were Results Horseferry Road, London, to see was to SW1P 2AP. performed if it possible predict Stool volume varied from 300 ml to more than Accepted for publication which patients would benefit from these 3000 ml per 24 hours. Use of arbitrary cut off 5 October 1989 tests. points at 400 and 800 ml led to division of Investigation ofseeminglypathogen-negative diarrhoea inpatients infected with HIVI 887

Scattergram to show relation 0 Sensitivity, specificity, and positive predictive value (PPV) between xylose excretion, 0 -24 for each ofthe degrees ofabnormality ofthe parameters Schilling test result, and evaluated, with negative predictive value (NPV)for a normal 00 result (percentages in each case). final evidencefor or against Gut: first published as 10.1136/gut.31.8.886 on 1 August 1990. Downloaded from infective aetiology for 20 diarrhoea. Xylose results are 0 Sensitivity Specificity PPV NPV excretion in g/5/hour urine .16 collection and Schilling O" a* Weight loss: cm >5 kg 100 43 50 100 results are presented as the co >10 kg 75 90 64 - percentage ofradioisotope x 00 0 Abnormal xylose absorption 83 57 53 86 excretion; in both cases the 1-0' a n/ <50% normal lower limit 42 81 56 horizontal line marks the S. Abnormal Schilling test 92 52 52 92 lower end ofthe normall co <50% normal lower limit 92 86 79 1. 0 reference range. a 00 * 00 .0 oJ v-.. sr were identified. Weight loss of less than 5 kg Without With Without With therefore had a negative predictive value of 100% infection infection infection infection for an eventual microbiological diagnosis, whereas weight loss of greater than 10 kg had a patients between three approximately equal positive predictive value of 64% and a sensitivity sized groups: less than 400 ml, n=9; 400-800 ml, of 75% (Table). n=11; more than 800 ml, n=13. Similarly Abnormality of xylose absorption (whether or arbitrary cut off points for weight loss led to not severe), however, was less discriminatory - distribution of patients between three groups: an abnormal result had only a 53% positive less than 5 kg loss, n=9; 5-10 kg, n=10; more predictive value for a final microbiological diag- than 10 kg, n= 14. The groupings form natural nosis and a normal result only an 86% negative pairs and in all but one patient the stool volume predictive value. Severe abnormality of xylose grouping was 'equivalent' to that for weight loss. absorption had a specificity of 80% associated Serum B- 12 and folate and red cell folate with a sensitivity of only 42% (Table). The concentrations were normal in all patients. receiver operating characteristics curve con- Xylose excretion, however, was impaired in 19 structed from the xylose absorption data patients and in nine of these excretion was less indicated no better discrimination from any than 50% of the lower limit of normal (severely other given level of cut off. abnormal) (Figure). Unsurprisingly, the Schil- A microbiological diagnosis was made in only ling tests showed no evidence of intrinsic factor one patient with a normal Schilling test but of 14 deficiency, and both labels were detected at patients with a severely abnormal result, patho- similar radioconcentration: the mean of the two gens were found in 11 all five (including patients http://gut.bmj.com/ values for each patient was therefore used. In 21 with cryptosporidial infecton, and two patients patients abnormal values were detected and in 14 with only moderate weight loss and stool of these, mean values of less than 50% of the volume). A normal Schilling test therefore had a lower limit of normal were recorded (severely 92% negative predictive value for a final micro- abnormal). biological diagnosis. A severely abnormal Schil- Infection was finally shown in 12 patients ling test, however, had a sensitivity of92% and a (more than one organism in three): crypto- positive predictive value of 79% for subsequent sporidium, n=5; Giardia lamblia, n=2; Enta- microbiological diagnosis (Table). The receiver on September 23, 2021 by guest. Protected copyright. moeba histolytica, n=1; Campylobacter jejuni, operating characteristics curve for Schilling test n=1; Salmonella enteritidis, n=1; cytomegalo- results indicated that a cut offpoint at 50% ofthe virus, n=4; and herpes simplex virus, n=1. lower limit of normal (by chance) corresponded Diagnosis of cytomegalovirus was from biopsy to the best combination ofsensitivity and specifi- specimen alone. Diagnosis of herpes simplex city achievable. virus was from biopsy specimen and from swabs A barium follow through study was abnormal taken from rectal vesicles. Cryptosporidium was in only two patients who had cryptosporidial detected in both biopsy specimen and stool infection, appreciable weight loss, and severely microbiology in two patients. The remaining abnormal Schilling test results. All endoscopies diagnoses were from stool microbiology alone. were macroscopically normal. All histological diagnoses were obtained from rectal biopsy specimens except for one patient where rectoscopy was normal but typical cyto- TREATMENT AND OUTCOME megalovirus ulcers were seen in the transverse Patients were treated on the basis of the results colon at colonoscopy and a biopsy confirmed obtained, using specific antibiotic drugs for the cytomegalovirus infection. bacterial infections (ampicillin for S enteritidis, erythromycin for C jejuni), metronidazole for giardiasis and amoebiasis, and high dose CORRELATION OF MALABSORPTION AND acyclovir for herpes simplex virus. In all patients INFECTION relevant treatment eradicated the organism for An estimated weight loss of more than 10 kg was the duration of microbiological follow up (more strongly associated with eventual identification than two months) and stool volumes were of gastrointestinal pathogens: nine of 14 patients reduced. The five patients with cryptosporidial with this extent of weight loss had organisms diarrhoea were treated with a variety of anti- detected, accounting for nine of the 12 patients microbial agents without success, but starting in whom an organism was eventually found. them on zidovudine and opioid antidiarrhoeal When weight loss was less than 5 kg or daily stool drugs was associated with control of symptoms volume was under 400 ml, or both, no pathogens and considerably reduced stool volumes in two 888 Connolly, Forbes, Gazzard

patients. The other three patients failed to initial investigation. There was, however, a good respond and died within one month with con- and long lasting (more than two months) remis-

tinued watery diarrhoea. Three of the patients sion from diarrhoea after antibiotic treatment. Gut: first published as 10.1136/gut.31.8.886 on 1 August 1990. Downloaded from with cytomegaloviris colitis were treated (the Stool electron microscopy in the routine evalua- fourth was terminally ill by the time of diag- tion of these patients may be thought valuable, nosis). One responded to continuous ganciclovir, although in a large series of stool samples from one to a single three week course of phosphono- patients with AIDS and diarrhoea (unpublished formate but the last showed no response to observations), no viral particles were observed or antiviral treatment. other evidence provided for infection missed by other investigations. Electron microscopy of small bowel mucosa, however, is of considerable SCORING SYSTEM value in the diagnosis of microsporidial infec- Stool volume and estimated weight loss provided tion, which may be associated with malabsorp- almost entirely concordant information; they are tion and altered small bowel histology but with- clearly not independent variables as far as their out evidence of the organism in conventional diagnostic potential is concerned. Since stool duodenal pinch biopsy specimens. In one recent volume, despite being a valuable marker of study of patients similar to those considered disease activity and response to treatment, is here, more than a third had Microsporidia more difficult to measure (particularly in out- present at jejunal electron microscopy.'0 patients) and proves relatively expendable diag- Good evidence for malabsorption was found nostically, we used estimated weight loss as the frequently in this group ofpatients, but although preferred variable, grouping patients between there has been speculation that HIV itself causes the three arbitrary bands of severity described malabsorption,2 the present data argue for above. Of the other variables considered, Schil- malabsorption being linked to secondary micro- ling test result, with receiver operating charac- bial infection in many patients. The normal teristics curve-determined cut offat 5 5% or less, serum vitamin B- 12 values is unsurprising showed most promise. Application of Bayes' given the relatively short total illness period in theorem led to the allocation of the following the patients concerned, but the site of the scores: weight loss less than 5 kg: +0-19; abnormality in the two positive barium follow 5-10 kg: +0 04; more than 10 kg: -0-25; through examinations and the normal folate Schilling result: more than 5-5%: +0-17; 5-5% studies suggest that the malabsorption is pre- or less: - 0-48. Fortunately, almost as effective dominantly of ileal origin. Recent reports of is a simplified version where weight loss is scored possible relevance here indicate that there may

with 2 points for more than 10 kg loss, 1 point for be two forms of cryptosporidial infection, one http://gut.bmj.com/ a loss of 5-10 kg, and 0 points for weight loss of where organisms are detected in stool only with less than 5 kg, and where points are allocated for difficulty but small bowel involvement and the Shilling result as follows: more than 5 5%: 0 B- 12 malabsorption are prominent, and the point; 5-5% or less: 2 points. Retrospective other where oocysts are readily identified in stool application to the present patients (predictably) and the Schilling test is usually normal." shows a 5% frequency of pathogen recognition As we have shown in other contexts,'2'"

for patients with scores of 0-2 (n= 19), whereas barium radiology and endoscopy are only rarely on September 23, 2021 by guest. Protected copyright. 90% of patients with a score of 4 (n= 10) had at helpful in patients with AIDS and will not often least one pathogen identified. be indicated, but rectal biopsy is a useful early investigation9 and remains the only route to diagnosis in some patients. Discussion Based on the results presented here, the Microbiological examination of numerous stool diagnostic yield of further microbiological samples is evidently the most useful diagnostic samples in patients who have lost less than 5 kg in procedure, and this applies equally to the present weight and have a normal Schilling test is difficult group as to the generality of HIV extremely low. A putative scoring system is infected patients with diarrhoea.9 The present suggested but this clearly needs prospective results, however, help to indicate the patient in evaluation in future patient groups. whom it is reasonable to conclude that a specific diagnosis is to however We thank our patients and our many colleagues without whose unlikely emerge many assistance this work would not have been possible, but are samples are studied. The value of a specific particularly indebted to Drs Shanson, Harcourt-Webster, and diagnosis is not always obvious when no estab- Gleeson. lished treatment exists (for example in crypto- 1 Antony MA, Brandt LJ, Klein RS, Bernstein LH. Infectious sporidiosis), but valuable prognostic informa- diarrhea in patients with AIDS. Dig Dis Sci 1988; 33: tion may be obtained (several of the present 1141-6. 2 Nelson JA, Wiley CA, Reynolds-Kohler C, Reese CE, patients were moved from HIV group IV A to Margaretten W, Levy JA. Human immunodeficiency virus group IV Cl or IV C2) and, increasingly, effec- detected in bowel epithelium from patients with gastro- intestinal symptoms. Lancet 1988; i: 259-62. tive new treatments (for example for cytomega- 3 Centers for Disease Control. Classification system for human lovirus) are becoming available. T-lymphotropic virus type III/lymphadenopathy-associated virus infections. MMWR 1986; 35: 334-9. It is acknowledged that the infecting 4 Garcia LS, Bruckner DA, Brewer TC, Shimizu RY. Tech- organisms identified may not necessarily have niques for the recovery and identification ofcryptosporidium oocysts from stool specimens. J Clin Microbiol 1983; 18: been relevant to the chronic diarrhoea - particu- 185-90. larly the bacterial and amoebic infections - as 5 Mitchison DA, Allen BW, Manickavasagar D. Selective Kirchner medium in the culture of specimens other than these may have arisen acutely and been super- sputum for mycobacteria.JT Clin Pathol 1983; 36: 1357-61. imposed on (still) undiagnosed chronic diarrhoea 6 Freedman ALS. Evaluating and comparing imaging tech- nique: a review and classification of study designs. Br J thus explaining why they were not detected on Radiol 1987; 60: 1071-81. Investigation ofseemingly pathogen-negative diarrhoea inpatients infected with HIVI 889

7 Griner PF, Mayewski RJ, Mushlin Al, Greenland P. Selec- in AIDS. Vth International Conference on AIDS, Montreal, tion and interpretation of diagnostic tests and procedures: 1989: 209 (Abstract). principles and applications. Ann Intern Med 1981; 94: 11 Heller TD, Tierney AR, Kotler D. Variable localization of 553-600. intestinal cryptosporidiosis in AIDS. Vth International

8 Knill-Jones RP. Diagnostic systems as an aid to clinical Conference on AIDS, Montreal, 1989: 358 (Abstract). Gut: first published as 10.1136/gut.31.8.886 on 1 August 1990. Downloaded from decision making. BrMedJ 1987; 295: 1392-6. 12 Connolly GM, Forbes A, Gleeson JA, Gazzard BG. Investiga- 9 Connolly GM, Shanson D, Hawkins DA, Harcourt-Webster tion of upper gastrointestinal symptoms in patients with JN, Gazzard BG. Non-cryptosporidial diarrhoea in human AIDS. AIDS 1989; 3: 453-6. immunodeficiency virus (HIV) infected patients. Gut 1989; 13 Connolly GM, Forbes A, Gazzard BG. Value of barium enema 30: 195-200. and colonoscopy in HIV positive patients with diarrhoea. 10 Orenstein J, Steinberg W, Chiang J, Smith P, Rotterdam H, [Abstract] Gut 1989; 30: A735. Kotler D. Intestinal microsporidiosis as a cause ofdiarrhoea http://gut.bmj.com/ on September 23, 2021 by guest. Protected copyright.