Infection status for contacts is not associated with severity of TB in the index case.

Dear Editor,

Contact investigation is an important tool for finding prevalent TB and preventing future incident cases.^1,2 Contact investigation is typically conducted after index cases present to health facilities with TB symptoms.³ However, active case finding (ACF), including testing of asymptomatic people in communities, has the potential to identify people with TB at an earlier stage of the disease. This early detection could prevent people from developing characteristics such as high cough frequency and sputum bacillary load associated with TB transmission.^4–6 Early detection could also shorten the time that contacts are exposed.

To explore the relationship between earlier detection of cases and risk of infection among contacts, we compared the TB risk in contacts whose index cases were identified through community-based ACF versus routine healthcare in the same district. From February to November 2019, we conducted community-based ACF among adults (age≥15 years) in a high TB-burden urban district of Kampala, Uganda (total population ~50,000 residents and area 2.2km²). We offered sputum Xpert MTB/RIF Ultra (“Ultra”) testing to all adults irrespective of symptoms or risk factors both door-to-door and at public venues. Simultaneously, we enrolled all adults diagnosed with pulmonary TB through routine care. All individuals diagnosed with TB named their household contacts and two closest non-household contacts, among whom we evaluated symptoms, HIV, and tuberculin skin test (TST) status. All participants provided written informed consent, with assent and parent/guardian consent for participants <17 years old.

Our primary outcome was the proportion of contacts with TB infection, defined by TST induration ≥10 mm at 48–96 hours. For two cases simultaneously identified through ACF in the same household, one was randomly designated as the index case, and the other as a TB-infected household contact without requiring TST. We estimated the proportion of contacts with TB infection overall and after stratification by demographic, clinical, and household characteristics. The mode of index case diagnosis (ACF versus routine care), level of Ultra positivity (trace versus greater than trace), and duration (≥2 weeks versus <2 weeks) of TB symptoms were our primary exposures of interest. We constructed multilevel Poisson regression models with index-level random effects. Given correlation of Ultra positivity level with mode of TB diagnosis and duration of symptoms (Pearson correlation coefficient 0.4 and 0.6, respectively), each was evaluated in a separate model. We selected contact-level covariates after considering biological plausibility, univariable association, and sparse data concerns. We used R v3.6.1(R Foundation, Vienna, Austria, 2019).

Of 331 named contacts from 62 index cases (of 140 cases diagnosed), 165 contacts (50%) completed contact investigation (81 adults and 84 children); 126 (38%) did not participate, and 40 (12%) had TST placed but no valid read. 100 contacts (61%, 62 adults and 38 children) had TB infection (99 TST-positive, one co-prevalent active TB). The prevalence of positive TST among adult contacts (62/81, 77%) was higher than in a randomly-selected control population of adults from the same community (35/71, 49%). Adult (versus child) contacts and household (vs. non-household) contacts had a higher prevalence of TB infection (adjusted prevalence ratio [aPR] for age 30–45 vs. <15 years: 2.11[95% confidence intervals=1.13–3.82]; aPR for household vs. non-household contacts=1.47[0.80–4.26]) (Table).

Table.

Characteristics associated with TB infection among contacts of cases in urban Uganda

	Descriptive characteristics		Model analyses
	TB infection (n = 100, 61%) n (%)	No TB infection (n = 65, 39%) n (%)	Contact-level models* PR (95% CI)	Index-level models† PR (95% CI)	Contact- and index-level, Model 1‡ aPR (95% CI)	Contact- and index-level, Model 2‡ aPR (95% CI)	Contact- and index-level, Model 3‡ aPR (95% CI)
Index-level Ultra positivity§
Trace	28 (55)	23 (45)	—	Reference	Reference
More than trace	60 (64)	34 (36)	—	1.16 (0.69–2.02)	1.26 (0.81–2.03)
Index-level TB symptoms
Cough/hemoptysis	74 (60)	49 (40)	—	0.92 (0.56–1.51)	—
Fever	39 (57)	29 (43)	—	0.88 (0.52–1.35)	—
Weight loss	62 (60)	41 (40)	—	1.03 (0.66–1.68)	—
Night sweats	22 (52)	20 (48)	—	0.85 (0.46–1.34)	—
Duration of any TB symptoms ≥2 weeks	81 (63)	48 (37)	—	1.38 (0.88–2.93)		1.37 (0.82–3.46)
Index-level enrollment method					—
Community active case finding	80 (60)	54 (40)	—	Reference	—		Reference
Healthcare facility	20 (65)	11 (35)	—	1.12 (0.60–1.85)			1.22 (0.67–1.95)
Household crowding, n, median [IQR]¶	3 [1.5–4]	3 [2–5]	—	0.96 (0.87–1.04)	—
Contact-level age, years
<15	38 (45)	46 (55)	Reference	—	Reference	Reference	Reference
15–30	25 (66)	13 (34)	1.40 (1.00–1.95)	—	1.56 (0.83–2.46)	1.54 (0.86–2.49)	1.61 (0.89–2.97)
30–45	24 (89)	3 (11)	1.94 (1.48–2.54)	—	2.11 (1.13–3.82)	2.06 (1.09–3.39)	2.10 (1.08–3.93)
≥45	13 (81)	3 (19)	1.77 (1.27–2.47)	—	2.06 (1.01–3.72)	1.95 (0.88–3.42)	2.03 (0.94–4.04)
Contact-level male sex	53 (58)	38 (42)	0.92 (0.72–1.17)	—	0.92 (0.60–1.41)	0.91 (0.58–1.42)	0.94 (0.60–1.49)
Contact-level HIV-positive	7 (54)	6 (46)	0.88 (0.52–1.48)	—	—
Contact-level TB symptoms
Cough	55 (57)	41 (43)	0.88 (0.69–1.12)	—	—
Fever	25 (66)	13 (34)	1.11 (0.85–1.46)	—	—
Weight loss	13 (76)	4 (24)	1.30 (0.97–1.75)	—	—
Night sweats	7 (58)	5 (42)	0.96 (0.59–1.57)	—	—
Cough ≥2 weeks	35 (57)	26 (43)	0.92 (0.71–1.19)	—	—
Household contact	88 (60)	59 (40)	0.90 (0.63–1.28)	—	1.47 (0.80–4.26)	1.43 (0.71–3.53)	1.42 (0.82–4.19)

^*Poisson regression and robust standard errors.

^†Multilevel Poisson regression and bootstrap standard errors.

^‡Adjusted for the covariates indicated (contact-level age, sex, and household contact).

^§Index cases with negative Ultra result or no Ultra result were excluded.

^¶Measured as the number of household members divided by the number of rooms.

PR = prevalence ratio; CI = confidence interval; aPR = adjusted PR; Ultra = Xpert® MTB/RIF Ultra ; IQR = interquartile range.

Controlling for contacts’ age, sex, and household (versus non-household) exposure, the prevalence of TB infection among contacts was not associated with the mode of index case diagnosis (healthcare facility evaluation vs. ACF, aPR=1.22[0.67–1.95]), level of Ultra positivity (more-than-trace-positive vs. trace-positive, aPR=1.26[0.81–2.03]), or duration of TB symptoms (aPR=1.37[0.82–3.46]). These findings were robust to changes in the covariates (e.g. each TB symptom), the handling of co-prevalent cases and the allowed time window for TST reading (48–72 hours).

In this study of TB patients and their known contacts in a high-TB-burden community, we found that the majority of adult and household contacts were TST positive, but contacts’ infection status was not significantly associated with index bacillary burden, duration of TB symptoms, or location/mode of diagnosis (symptom-driven healthcare versus community-based ACF). Although our sample size cannot exclude positive associations of small magnitude, these findings indicate that close contacts are at high risk of TB infection even before index cases develop classic features associated with being infectious, or begin to seek care for TB-related illness.

These findings might be explained in at least two ways. First, the prevalence of TB infection in contacts of individuals with minimally symptomatic TB and trace-positive Ultra results may indicate transmission of M. tuberculosis during early stages of disease. Although individuals with lower sputum bacillary loads are likely less infectious on a per-person-day basis,^7,8 they can be potential sources of community transmission.⁸ Long durations of paucibacillary disease⁹ and exhalation of M. tuberculosis before it is detectable in sputum¹⁰ could facilitate transmission from Ultra trace-positive cases. Second, TB infection in contacts may reflect exposures and/or risk factors shared between index cases and contacts rather than direct transmission. Most TB transmission in high-burden settings occurs outside of households,¹¹ and household members often share risk factors, such as social environment, or genetic predisposition.¹² Our finding that contacts have a high prevalence of TB infection regardless of risk factors for index infectivity aligns with previous research showing that although household contacts are at high risk of TB infection and disease,^11,13 a relatively low proportion of mycobacterial strains isolated from index patients and contacts are genetically matched in high TB-burden settings.^12,14,15

This study has important limitations. Only 50% of referred contacts (corresponding to 44% of enrolled index cases) were included in the analysis. Low enrolment reflects the logistical challenges of recruiting contacts and reading TSTs (especially for non-household contacts and contacts who worked long hours far from home). Hence, our enrolled population is representative of contacts who could be reached through routine contact investigation, but not of all contacts in the community. If contacts without TB infection were less likely to participate or have their TSTs read, the overall prevalence of TB infection among contacts could have been overestimated. Our results were also underpowered to identify small effects (prevalence ratios <1.6) that could nevertheless be epidemiologically relevant. Furthermore, because TST cannot distinguish recent from remote infection, our reported prevalence ratios may underestimate the magnitude of association between index-case characteristics and recent infection. Thus, we may underestimate the relative risk of progression to active disease in the most highly exposed contacts.

In summary, contacts of people with TB are at high risk of infection even when index cases have very low sputum bacillary burden, or have yet to seek care. ACF interventions should therefore consider including contact investigation even when index cases are thought to have been identified at an early stage of disease.