Feasibility of the string test for tuberculosis diagnosis in children between 4 and 14 years old

The World Health Organization (WHO) estimated that one million of the 10 million cases of tuberculosis (TB) worldwide in 2017 occurred in children < 15 years of age [1]. This is likely an underestimate given that TB is notoriously difficult to diagnose in children. The WHO guidelines for diagnosing TB in children indicate evaluation of the child’s TB contact history and exposure, a test of infection (tuberculin skin test, interferon gamma release assay), chest radiograph, and bacteriological confirmation by culture [2]. As noted by Chiang, et al. [3] there are serious limitations to each step in the diagnostic process. Bacteriologic confirmation of pulmonary TB is the most certain method for determining disease status, but this test depends on sputum, which is difficult to collect in young children who are often unable to expectorate spontaneously [3]. The alternative sample type for bacteriologic confirmation of Mycobacterium tuberculosis in children who cannot produce sputum is gastric aspirate; however, the procedure to obtain this sample is invasive and it is not broadly available in resource-constrained settings due to a limited number of health providers trained in the procedure and/or a lack adequate space within primary health facilities [4‐9]. Furthermore, the paucibacillary nature of TB in children decreases the sensitivity of acid-fast smear microscopy, mycobacterial culture and nucleic acid amplification tests, all of which often are negative in children with clinically diagnosed TB disease [10, 11].

The enteric string test, typically used to diagnose intestinal parasites, consists of ingestion of a small, dissolvable gelatin capsule containing a string that absorbs stomach fluids [12]. Because the string could also absorb sputum that has been swallowed, it has been proposed as a technique for obtaining lower respiratory specimens for tuberculosis testing in children. Several studies in adults have found that string samples have a diagnostic yield that is roughly similar to induced sputum [13‐15]. The handful of studies in children have reported that > 80% could swallow the capsule and complete the test [4, 16, 17]; though this rate may vary by the child’s age, with the youngest children least able to swallow the capsule [13]. Among children who complete the test, diagnostic yield is comparable to sputum or gastric aspirate [13, 18]. In spite of its perceived simplicity and promise, the string test has not been widely adopted.

To respond to the need for a noninvasive clinical sample for routine TB diagnosis that is easily collected from children, we examined the feasibility and diagnostic effectiveness of the string test in children with suspected pulmonary TB in Lima, Peru.

We included 148 children (median age: 8.2 years, IQR: 5.5–10.0 years) in whom the string test was attempted. Table 1 provides descriptive statistics for the children and Table 2 shows an overview of study enrollment and procedures. Over half of children (54%) were under seven years old, and no child had known HIV-infection. Nearly all participants provided at least one sputum or gastric aspirate sample (n = 146, 99%) with 57% (n = 82) providing sputum (74 spontaneous and 8 induced samples, two children provided one of each), and 51% (n = 75) providing gastric aspirates. Thirty-two children were diagnosed with TB by a pediatric pulmonologist, of whom 12 (38%) had bacteriologic confirmation of TB by culture of sputum or gastric aspirate.

This study demonstrated that the string test, although generally tolerable and accepted by children and their caregivers, was not reliably feasible, especially among younger children. This is the group least likely to be able to produce a spontaneous sputum sample and therefore, in greatest need of an alternative sample type. There were no positive string cultures in this group. We did not observe increased bacteriological detection when using the string test as compared to conventional sample types. Among children who were able to produce a sputum sample (either spontaneously or through induction), we found that culture positivity for the string test was lower than sputum (6 of 9 children with positive sputum cultures, had positive string cultures). The single gastric aspirate sample in our study that was positive by culture was also positive by string.

Most children in the 4–10 year-old age group (78%) were not able to swallow the capsule. In contrast, older children were generally able to swallow the capsule and successfully complete the test. String test success rates were higher (> 80%) in other studies with a similar median age [4, 16, 17]; though these studies also found the string test to be least successful in young children. Had gastric aspirate and sputum induction not been available in our setting, the string test would have resulted in a sample for culturing in 13 (20%) children 4–10 and 6 (100%) children 11–14 who would not have otherwise had one because they could not produce a spontaneous sputum sample.

Our capsule sizes for children < 10 years were the same size or smaller than the pediatric Entero-Test (HDC Corporation, San Jose, CA), which was used in previous studies [4, 17] or adapted from Entero-test [18]. Smaller capsules and additional interventions to facilitate pill swallowing [25] could improve string test feasibility. One potential explanation for differential string test success across settings relates to pill taking familiarity. For example, one third of children that participated in the study conducted in Uganda by Nansumba, et al. [4] were living with HIV, and therefore may have been more accustomed to swallowing pills.

For the diagnosis of pediatric TB in primary health centers, the string test may have advantages over gastric aspiration, which requires more supplies, experienced, well-trained personnel and an adequate environment in which to perform the procedure. Both methods require overnight fasting. Notably, older children and guardians of children in both age groups were more likely to report willingness to repeat gastric aspiration than to repeat the string test. Study nurses reported that many guardians praised the brevity of gastric aspiration as compared to the long duration of the string test, which could be disruptive to daily activities. While gastric aspirate takes an average of 20 min and is completed in one attempt, the string test requires up to 4 h of intragastric downtime. Although one study in adults suggested that intragastric downtime could be reduced to one hour without a loss in diagnostic yield, it is unknown to what extent these findings apply to children, who are more likely to have paucibacillary disease [26]. Though longer, most children reported that the string test was relatively painless, as compared to the gastric aspirate.

We compared the diagnostic yield from a single string sample to the yield from up two sputum and/or gastric aspirate samples. An additional string sample may have improved detection relative to sputum; however, given that the string test would be most valuable in children unable to produce sputum, we do not perceive this as a major study limitation. A second limitation is the small sample number of children with culture-confirmed TB, which precluded us from formally testing the diagnostic yield of the string sample relative to conventional respiratory samples.

Our findings reinforce those from a small body of literature examining the string test as a diagnostic specimen for children with suspected TB. The string test is a well-tolerated and acceptable alternative sample type for diagnosing TB in children among those who are able to swallow the capsule. In contexts where gastric aspirate is not available, it may be an acceptable substitute; however, the sample and case detection yield may be small to moderate.