From a total of 88 patients suspected of having gastrointestinal TB on presentation at Queen Elizabeth Hospital and the Women’s and Children’s Hospitals, Kota Kinabalu, Sabah, between March 2015 and June 2017, consent for inclusion in the study was obtained from 77 patients. Of these, eight patients were excluded from the analysis due to incomplete follow-up and final diagnosis. Among the 69 patients included in the study: 52 (75%) had a final diagnosis of gastrointestinal TB (25 confirmed, 13 probable, 14 possible cases), and 17 patients were assessed to have non-TB diagnoses, the majority of which (n=12) were malignancy.
Among the 52 cases of confirmed, probable and possible gastrointestinal TB, 42 (81%) had infection of the intestine, 14 (27%) had infection of an abdominal lymph node, 13 (25%) had infection of the peritoneum, 8 (15%) had infection of the liver, and 1 (2%) had biliary infection (figure 1). A total of 45 (87%) gastrointestinal TB cases were recorded as having involvement of multiple sites or organs within the gastrointestinal system.
The mean age of patients with a final diagnosis of gastrointestinal TB (confirmed, probable, possible) was 42.7 years, significantly lower when compared with the mean age of 61.5 years for patients with non-TB diagnosis (t (46 d.f.) = 5.07, p < 0.01) (figure 2). There was no significant difference in sex, with 51% of gastrointestinal TB cases being male, compared with 60% of non-TB cases.
Microbiological and histological results
A total of 110 biopsy samples were collected from patients during initial work-up and ongoing clinical follow-up. The largest proportion of samples were collected from the large bowel (34%), followed by the ileum (14%) and peritoneum (10%). Samples were also collected from the stomach, small bowel, liver and omentum. Peritoneal biopsy samples were collected predominantly via laparoscopy (n=7) or laparotomy (n=2).
Twenty-four of 42 TB cases tested by GeneXpert® were positive for M.tuberculosis. Of 39 TB cases with mycobacterial culture results, 12 had M.tuberculosis detected by culture. GeneXpert® positivity by sample type is detailed in supplementary table 4. Rifampicin resistance was detected by GeneXpert® in three cases, of which one was found to be sensitive to rifampicin and another was confirmed rifampicin resistant by phenotypic drug-susceptibility testing. The third was treated as rifampicin resistant based on the GeneXpert® result. Two cases were found to be isoniazid mono-resistant by phenotypic drug-susceptibility testing.
Sensitivity and specificity of diagnostic tests for M.tuberculosis are presented in table 1. GeneXpert®, culture, PCR and smear microscopy for Acid Fast Bacilli all achieved 100% specificity against the gold standard of final case classification of confirmed TB. GeneXpert® achieved high sensitivity (95.7%), compared with very low sensitivity of culture (35.0%), PCR (50.0%) and microscopy (31.0%). Histology achieved moderate sensitivity (68.0%) and specificity (77.1%) when using the lower threshold for evidence of TB (‘suggestive’ or ‘highly suggestive’ of TB).
Features distinguishing TB from non-TB and development of a clinical score
Among confirmed and probable TB cases, the most common symptoms reported at presentation were weight loss (91%), fever (54%) and abdominal pain (52%). For non-TB cases, the most common presenting symptom was abdominal pain (71%), followed by weight loss (64%) and vomiting (36%) (figure 3). Confirmed and probable TB cases were significantly more likely to present with weight loss compared with non-TB cases (χ2=4.93, p = 0.03). The proportion of probable and confirmed TB cases with a recorded temperature of 38 degrees Celsius or higher on presentation was higher (24%) than non-TB cases (14%), however the difference was not statistically significant. Confirmed and probable TB cases presented with longer median duration of fever (4 weeks vs. 1 week) and abdominal pain (8 vs. 4 weeks) but shorter duration of weight loss (10 vs. 14 weeks), vomiting (3 vs. 18 weeks), cough (8 vs. 53 weeks) and fatigue (8 vs. 12 weeks), compared with non-TB cases. However, none of the differences in duration of symptoms between TB and non-TB cases were statistically significant.
Two patients in our study cohort had HIV infection, both with confirmed gastrointestinal TB. There were no statistically significant differences in the proportion of patients who had: a history of Bacillus Calmette–Guérin (BCG) vaccination; past TB diagnosis; family history of TB; positive tuberculin skin test; HIV infection; diabetes; or immunocompromising condition, between confirmed TB cases and non-TB cases or confirmed and probable TB cases and non-TB cases. No significant risk factors emerged when limiting the analysis to either confirmed cases of TB or non-TB cases with malignancy.
There were no statistically significant differences in blood chemistry, blood count or liver function tests between confirmed and probable TB cases and non-TB cases (supplementary material, table 3). Non-TB cases had lower mean albumin levels (21.54 vs. 26.17 g/L) and higher mean bilirubin levels (24.0 vs. 13.35 mmol/L) from peripheral blood samples on initial diagnostic work-up—however these differences were not significant. No statistically significant differences emerged when limiting the analysis to confirmed TB cases or non-TB malignancy. Blood cultures were positive in four of four non-TB cases who had blood cultures collected (isolates were: Elizabethkingia meningoseptica, Escherichia coli, and two with Klebsiella pneumoniae) and one probable TB case (Acinetobacter baumannii).
A composite clinical score was developed to determine whether a scoring system could reliably distinguish TB from non-TB diagnosis (Table 2). The best performing algorithm for diagnosis of gastrointestinal TB included: age less than 44 years; weight loss; cough; fever; no vomiting; blood albumin level great than 26 g/L; platelet count greater than 340 x109/L; and history of immunocompromising illness/treatment. However, this algorithm achieved an area under the receiver operator characteristics curve of just 0.5873. The highest correct classification of cases (70.1%) occurred at a cut-point of eight points from a maximum of 11 points, this corresponded to a specificity of 96.2%, but sensitivity of only 16.0% (positive predictive value = 60.0%; negative predictive value = 3.8%). There was no significant difference in performance of the composite clinical score when assessed against an alternative gold standard, such as microbiologically confirmed M.tuberculosis.
Details of the prescribed anti-tuberculous treatment regimen was documented for 38 patients. Standard first-line treatment of isoniazid, rifampicin, ethambutol and pyrazinamide was initiated in 32 (84%) TB cases with documentation of treatment, with proposed treatment durations at initiation of, six-months, nine-months and 12-months for 9 (28%), 16 (50%) and 7 (22%) cases respectively. Most patients (91%) received pyridoxine (vitamin B6) during treatment. Six patients were given an alternative regimen: one due to adverse reaction to ethambutol, one due to baseline liver impairment, two due to detection of rifampicin resistance and two for an undocumented reason. Of the two cases with rifampicin resistance: one was treated with a regimen consisting of kanamycin, ethionamide, cycloserine and levofloxacin; the other with a regimen consisting of isoniazid, ethambutol, pyrazinamide, kanamycin, levofloxacin and ethionamide. Mean treatment duration at completion for cases receiving first-line regimens was 8.8 months (median 9.0 months, maximum 12.4 months).
A total of 12 adverse events were reported among eight patients undergoing anti-tuberculous treatment. The majority (67%) of adverse events occurred during the first two months of treatment. Hepatotoxicity and rash were the most frequent adverse events reported, accounting for 25% of adverse events each. Most adverse events were mild or moderate (75%), three required hospitalization but none were life-threatening or resulting in death. Changes in treatment regimen were required in six (16%) of the 38 patients for whom treatment details were available. Five adverse events resulted in permanent cessation of a drug, two resulted in a drug being stopped and then reintroduced and one resulted in a change in route of administration of a drug. No new agents were added to any regimens because of an adverse event.
TB cases were followed-up for a median of 7.1 months (maximum 16.9 months). Final outcome was recorded for 92% of gastrointestinal TB cases. Of these, 79% successfully completed treatment or were cured; no case of treatment failure was recorded. A total of six (13%) TB cases died. TB cases who died had a mean age of 54 years and all deaths occurred in the first five months following hospitalization. Four cases (8%) defaulted from treatment. Three (18%) non-TB cases died (two with malignancy and one with Helicobacter pylori gastritis).