Reducing tuberculosis transmission: a consensus document from the World Health Organization Regional Office for Europe

The actors: patients, health staff and visitors

The healthcare facility occupants playing a role in TB transmission are the patients, health staff and visitors. Although patients are traditionally seen as index cases transmitting the infection and HCWs as the potential victims to be protected, visitors (and other patients) also need to be considered, both as potential index cases or as individuals who might be infected. Although HCWs are known to be at risk of infection [38], recent reports underline the possibility that they can both infect patients [39] and be infected by them [40]. Interestingly, the report by Richeldi et al. [40] followed disease transmission in a micro-epidemic originating from a highly infectious sputum smear-positive/culture-positive MDR-TB woman who gave birth in a maternity ward. Of the 92 contacts, 88 only had nosocomial contact, whereas the other four had had both nosocomial and household contact: 41 were neonates, 42 were mothers on the ward and nine were visiting relatives (four husbands and five grandparents).

Who is infectious?

There is clear evidence that the sources of transmission in hospitals are patients (or health staff) with undetected, untreated TB or patients with known TB, but unknown drug resistance (receiving ineffective therapy). Several studies have documented the prevalence of undetected TB [41, 42]. Gelmanova et al. [43] documented the impact of undiagnosed MDR-TB on the emergence of MDR-TB in Tomsk in Siberia in the Russian Federation (a large territory almost of the same size as Poland). The core problem is that most transmission control efforts focus on patients with known TB, most of who are on effective therapy and therefore no longer infectious, whereas most transmission occurs elsewhere [14, 15].

Three historical studies (Bedfordshire in the UK (1948–1952) [44], Rotterdam in the Netherlands (1967–1969) [45] and Saskatchewan in Canada (1966–1971) [46]) show that untreated sputum smear-positive/culture-positive individuals are much more infectious to household contacts compared with sputum smear-negative individuals (either sputum culture negative or positive) prior to beginning effective anti-TB treatment. Interestingly, there was not much difference in tuberculin skin test (TST) conversion among the contacts of culture-positive/sputum smear-negative and culture-negative/sputum smear-negative cases, underlining that sputum smear positivity is the better predictor of infectiousness. These studies show that 25–65% of household contacts exposed to sputum smear-positive/culture-positive cases undergo TST conversion compared with <10% of those exposed to sputum smear-negative index cases (either sputum culture negative or positive). A more recent Finnish study found that among 609 contacts of 134 index cases (sputum culture positive with positive or negative sputum smear findings), four developed TB within 4 years. All had been in contact with untreated sputum smear-positive index cases [47].

A large Brazilian study of 160 index cases and 934 household contacts reported the rates of TB infection and secondary TB cases among the household contacts of sputum smear-positive/culture-positive cases related to the intensity and location of the exposure before treatment started [48]. The sleeping proximity (same bed, room or house), biological (father or mother) or social (spouse) relationship, and the average number of days per month of contact or hours per day nursing the index case and number of meals shared per day were evaluated as significant variables. A proximity score was developed, consisting of clinical variables correlating with an increased risk of TB infection (presence of cough, pulmonary TB, sputum smear positivity, being the index case's caregiver or mother, sleep location and living in the same house). For intense exposure, TST and interferon-γ release assay (IGRA) positivity rates were similar (81% versus 77%, respectively, for contacts sleeping in the same bed; 69% versus 67%, respectively, for contacts sleeping in the same room); for milder exposure, the TST positivity rate was higher than the IGRA positivity rate. The rate of both TST and IGRA positivity increased with the proximity score. Of the 710 individuals who underwent both IGRA and TST, 69% of contacts sleeping in the same bed were positive for both IGRA and TST compared with 62% of contacts sleeping in the same room, 54% of those sleeping in the same house and 39% of those sleeping in another house [48, 49]. The rates of secondary TB cases and TB incidence among household contacts increased with the proximity score. The proximity score was also associated with microbiologically confirmed TB disease (sputum smear positive/culture positive). The findings suggest that TB transmission before treatment was due to sputum smear-positive/culture-positive cases, although the study design did not generate information on how many infections or TB cases were specifically due to sputum smear-positive (only) or culture-positive cases.

In a fingerprinting study in California in the USA, 17% of transmission occurred among contacts of sputum smear-negative index cases before their treatment started [50]. The relative transmission rate from sputum smear-negative versus sputum smear-positive cases was 0.22 (or around one-fifth). The study by Acuna-Villaorduna et al. [51] showed that cough aerosol sampling was more predictive of transmission than sputum smear positive alone, presumably because that methodology incorporates the ability to generate aerosol.

Once on effective therapy, sputum smear/culture status no longer predicts transmission. Gunnels et al. [52] reported no difference in infection among household contacts of 86 culture-positive cases (52 were sputum smear negative/culture positive) and 69 culture-negative cases undergoing anti-TB chemotherapy (with 284 and 216 contacts, respectively). The human-to-guinea pig studies of Riley [53, 54] demonstrated more precisely the rapid effect of treatment on transmission. Smear-positive patients, immediately treated on admission, were 98% less infectious than smear-positive patients with delayed treatment. More recently, Dharmadhikari et al. [55] reported similar responses to standard treatment for MDR-TB.

In summary, transmission is multifactorial, relating to the concentration of viable, infectious bacilli in airborne droplet nuclei, itself the result of source strength, ventilation, proximity, drug susceptibility and, most importantly, the administration of effective treatment. Until effective treatment is provided, sputum smear positivity can be used to estimate infectiousness. We describe later that this is no longer true when the patient starts treatment, as long as the regimen is effective (i.e. there is no undetected resistance when a regimen for susceptible cases is administered) [14, 20].

Exposure time needed to generate infection

Various factors can influence the effectiveness of TB transmission, including factors related to the index case and the contact (discussed later) [6, 8]. Other important factors include the proximity, frequency and duration of contact. The Rotterdam study demonstrated the importance of proximity in TB transmission: 35% of close contacts of a sputum smear-positive index case were infected versus 10% of casual contacts. For sputum smear-negative cases, the proportions were lower (10% and <10%, respectively) [45]. The core question remains: what are the minimum frequency and duration of contact for infection to occur?

Historical studies in animal models enabled the study of airborne M. tuberculosis in droplet nuclei [53, 54, 56]. In the studies by Riley [53, 54], a TB ward with single rooms housing TB untreated patients was connected via a controlled, calibrated closed ventilation circuit to a large exposure chamber containing (exposed) guinea pigs. The animals were monitored by TST and their organs were examined if signs of disease developed. This method allowed the number of infectious doses in the air leading to effective transmission to be quantified. On average, one infectious dose was present in 340 m³ of air. The authors also estimated that the time needed for a nurse to breathe in this volume of air and become infected in the pre-antibiotic era (i.e. have TST conversion) was 12–18 months, on average. The infectious dose estimate for naturally exposed guinea pigs worked reasonably well for naturally exposed people.

A cut-off time of 8 h for exposure to an index case is used in several countries to initiate contact tracing. This approach has no scientific basis, as infection clearly can result from multiple brief exposures (the seconds needed to inhale an infectious particle) [57–60], and the (modest) evidence derives from transmission in airplanes and other closed settings [10, 11, 34, 57, 61, 62].

In airplanes, two major episodes of transmission occurred in a study of 2600 passengers on 191 flights (in nine different aircraft models). The first episode involved the transmission of M. tuberculosis from a crew member to colleagues in a flight lasting >12 h [58, 63]. The second episode involved a few passengers seated close to the index case in a flight lasting >8 h [63, 64]. However, the specific conditions in airplanes (i.e. a small air volume, overcrowded, with pre-defined air circulation and limited filtered air exchange while in flight) cannot be extrapolated to other settings.

In the MDR-TB micro-epidemic in a maternity ward in Modena in Italy, Richeldi et al. [40] reported an average exposure time of 6.05 h for the 88 main contacts and of >8 h for 15 contacts. Of the 88 contacts, 17 became IGRA (ELISPOT) positive. This study was the first to demonstrate the importance of casual nosocomial contacts and a correlation between exposure and TST and IGRA conversion. The sputum smear-positive/culture-positive index case (highly infectious, undiagnosed) remained in the maternity ward (with six beds per room) for 4 days post-natally, alongside other babies and mothers. The mothers slept in the same room and the babies slept separately in a nursery, with no specific infection control measures.

The risk of children becoming TST positive following exposure to sputum smear-positive adults at day care or after-school care facilities was considered low in a Norwegian study: <3% of children underwent TST conversion after a median contact period of 6.9 h [65]. However, a recent Italian study demonstrated that children can become infected within a rather shorter exposure time. A highly contagious paediatrician transmitted M. tuberculosis (resulting in several TB cases) in children undergoing vaccination against non-TB diseases, with an exposure time of only 15–20 min [39]. This study confirmed previous findings of rapid transmission in children exposed for <8 h while calling for improved workplace safety measures [66].

In a recent household contact study of sputum smear-positive/culture-positive cases performed in Brazil in a setting with a moderate TB incidence, the intensity of exposure and sleep location in relation to the TB index case were directly related to the TST/IGRA response and incidence of secondary TB [48]. The proximity score (a modified version of the Mandalakas score) provided a standardised way to capture the cumulative risk factors of proximity from 10 different variables [48]. These included cough features, existence of pulmonary TB, microbiological status, sleep location and living in the same house.

In summary, the available evidence does not enable the establishment of a cut-off time allowing infection to occur: it is not possible to define a lower time limit to exclude the possibility of M. tuberculosis transmission as several factors contribute, including source strength, dilution/ventilation, virulence of the infectious agent and host susceptibility. It is useful to underline that TB infection is a probabilistic and not a deterministic event.

The best approach available to establish infectiousness is the concentric circle or “stone-in-the-pond” approach to contact investigation [61, 67].

Features that influence transmission

Various factors can influence transmission; these can relate to the index case, the recipient, the bacterial strain or the environment [6, 8].

All factors favouring transmission deserve to be adequately evaluated by the clinician, including cough, clinical signs and symptoms, microbiological status (sputum smear, sputum culture, resistance pattern determined by both DST and rapid Xpert MTB/RIF testing), and radiological features.

As the FAST approach maintains, the most dangerous TB cases are the unknown/undiagnosed cases. Therefore, HCWs must keep a high level of clinical suspicion when dealing with patients reporting signs and symptoms compatible with TB (cough lasting for >2 weeks, general malaise, night sweats, fever, haemoptysis and weight loss) [35, 68]. The patient should undergo bacteriological examinations (rapid diagnostic tests, sputum smear, sputum culture and DST to confirm TB and exclude MDR-TB) and chest radiography [35, 69–71].

In the case of positive sputum smear findings and Xpert MTB/RIF confirmation of TB, the patient should be rapidly started on an effective TB treatment. If Xpert MTB/RIF testing confirms resistance, the patient should be treated as for MDR-TB. Rapid diagnostic tests for second-line drugs are useful to rapidly diagnose MDR-TB and exclude XDR-TB to avoid empirical treatment with insufficient active drugs [35, 68].

The contacts of a TB case should undergo epidemiological evaluation according to the concentric circle/“stone-in-the-pond” approach [67] and those infected (as evaluated by TST and/or IGRA, in line with national policies) should be treated for latent TB infection, as recommended by the WHO, after active TB is ruled out [35, 68, 72–74].

What is the effect of treatment on infectiousness?

Rapid diagnosis and effective treatment are the core factors to reduce the infectiousness of TB patients [13, 15, 17]. Clearly, adequate therapy should be given, in terms of drugs/regimens based on drug susceptibility, administration route, dose, duration, adherence and drug absorption [35, 75, 76].

Observational studies performed in the pre-MDR era demonstrated that anti-TB chemotherapy can rapidly render patients noninfectious [52–54, 77–81]. Based on this evidence, the 2-week rule after starting treatment was established (as the time needed for adequately treated patients to be considered noninfectious) [77].

Older guidelines tend to focus mainly on the 2-week rule, and still rely on sputum smear and culture conversion results, irrespective of whether effective treatment has been initiated. For example, infection control guidelines of the American Thoracic Society, US Centers for Disease Control and Prevention, and Infectious Diseases Society of America state that patients with pulmonary TB can be considered noninfectious when: 1) they have received adequate chemotherapy for 2–3 weeks, 2) they show clinical improvement and 3) there is a negligible chance of MDR-TB [9].

This rule was achieved through consensus among experts, who appeared not to consider the important results of Riley's human-to-guinea pig studies (although these were available at the time) [53, 54, 80].

In the first study, Riley et al. [80] observed three to four guinea pig infections per month following exposure to sputum smear-positive, previously treated, drug-resistant TB patients by means of a specially constructed TB ward. When newly diagnosed, previously untreated cases were admitted to the ward, the infections in guinea pigs stopped; they re-started when drug-resistant patients were admitted to the ward. These results were interpreted to indicate that the action of starting adequate treatment for susceptible cases on the day of admission to the ward was highly effective in stopping transmission.

In the second study, Riley et al. [80] exposed guinea pigs to drug-susceptible and drug-resistant TB patients (MDR-TB did not exist at the time) who had been untreated or treated since diagnosis and admission. The rather weak regimen used at the time (streptomycin, isoniazid and para-aminosalicylic acid) reduced infectiousness of drug-susceptible M. tuberculosis by 98% with an almost immediate effect since treatment was started at the time of admission. However, given the small number of drug-resistant cases, no firm conclusions could be made at that time, but have been made since based on further human-to-guinea pig transmission studies [55].

Fennelly et al. [56] cultured bacilli from a sample of cough aerosols using a novel method. They found that aerosol cultures from four adequately treated MDR-TB patients declined exponentially, much faster than results from sputum smear microscopy and smear culture. The numbers of patients were small, but the result supports the core role of treatment in reducing infectiousness and the unreliability of sputum smear/culture assays to monitor infectiousness after treatment starts.

Escombe et al. [82] studied guinea pigs exposed to exhaust air from a ward admitting TB/HIV co-infected patients. They showed that transmission occurred from nine patients with unsuspected MDR-TB, inadequately treated because underlying drug resistance was not suspected. A total of 292 guinea pigs were exposed for >505 days to exhaust air from 97 TB/HIV co-infected patients with pulmonary TB: 66 were sputum culture positive and 35 were sputum smear positive. Of 125 infected guinea pigs, 122 (98%) were infected by nine MDR-TB patients who were inadequately treated or whose treatment was delayed. Only three drug-susceptible TB patients infected one guinea pig each: two had delayed treatment and one interrupted treatment due to adverse events. Most TB patients on effective anti-TB treatment did not infect guinea pigs; only patients with undetected MDR-TB infected guinea pigs. This study supports the belief that effective treatment prevents transmission regardless of sputum smear and culture status.

In a recent study, guinea pigs were exposed to exhaust air from rooms hosting presumed MDR-TB cases [55]. In different human-to-guinea pig transmission experiments, the proportion of guinea pigs infected ranged from 1% to 77%. The reasons for the marked inter-experimental variation in transmission rates were unclear: patients had been selected based on similar characteristics associated with transmission (cough, sputum smear positivity and lung cavities) and all were receiving the standard South African MDR-TB treatment regimen. However, the explanation became evident when the fingerprint of several guinea pig isolates matched human isolates of XDR-TB, but not of MDR-TB. In one experiment, a cohort of 27 patients with similar transmission characteristics had infected just one out of 90 guinea pigs subjected to 3 months of continuous exposure. There were no XDR-TB patients in that cohort. As in the aforementioned studies by Riley, in most patients treatment started at the same time that the patient entered the experimental ward (not weeks earlier), again suggesting that effective treatment stops transmission almost immediately, including for MDR-TB. However, the same may not be true for XDR-TB, as effective drugs for MDR-TB such as fluoroquinolones and injectables are ineffective against XDR-TB. An unpublished study showed that supplementing a failed South African MDR regimen with bedaquiline and linezolid (according to the South African protocol) did not reduce transmission to guinea pigs over a 13-day period, but those drugs with pretomanid as part of the NIX regimen (i.e. pretomanid, bedaquiline and linezolid) were effective in rapidly stopping transmission (E. Nardell, Harvard Medical School, Boston, MA, USA; personal communication, 2018) [1, 2].

Dharmadhikari et al. [55] and Farmer and Raviglione [83] concluded that the standard MDR-TB treatment used in South Africa rapidly and effectively suppresses MDR-TB disease transmission, regardless of the sputum smear and culture status, similar to the results of Riley et al. [81] and of Escombe et al. [82] for drug-susceptible TB. These studies show the importance of rapidly detecting XDR-TB using rapid genetic tests and the timely initiation of appropriate treatment.

Recent evidence from the Russian Federation demonstrated that rapid diagnosis of MDR-TB by Xpert MTB/RIF resulted in a lower incidence of MDR-TB in the population (presumably due to decreased transmission and re-infection in hospitals), and that prolonged infectiousness in a patient under treatment can be due to inadequate chemotherapy, drug resistance and lack of treatment adherence [84–86]. We have been unable to find case reports of transmission from patients under effective treatment [15, 17, 87].

Fitzwater et al. [88] studied 95 drug-susceptible and -resistant TB cases in Peru treated with standardised short-course chemotherapy under strict directly observed treatment (DOT). The median time to convert sputum culture to a negative result was 38.5 days for the whole cohort. It was not affected by the sputum smear status at baseline. The median time to sputum culture conversion among drug-susceptible TB cases was 37 days; of these, 10% were still culture positive at day 60, which predicts the existence of MDR-TB. The results of this study should be interpreted as confirmation that conversion of sputum smear and culture might take time (particularly for MDR-TB cases), but that these outcomes cannot be used to predict transmission once effective treatment has started [15, 17]. Of course, on ineffective treatment, failure to convert sputum and culture (accompanied by deterioration of radiological and clinical indicators) are important clues, so clinical judgement is always required.

Some of the contents of a systematic review of available experimental and epidemiological evidence on the contagiousness of treated patients will be critically discussed later where relevant [89].

Experimental and epidemiological evidence

Although several studies used sputum smear and sputum culture conversion as indicators of effective treatment [90–92], we underline once more that they cannot be used to evaluate infectiousness [15, 17].

Aerosolisation and drug concentration may play a key role in limiting the spread of disease from patients on effective treatment, as suggested by Loudon and colleagues [21, 22]. In a trial to investigate the mechanism determining the rapid fall in infectiousness of patients after starting treatment, Loudon et al. [22] theorised that evaporation of the droplet nuclei could increase the drug concentration around bacilli. This mechanism might inactivate the bacilli or hamper their capacity to successfully infect human hosts. Although the drug concentration must increase as droplets evaporate, this phenomenon was never confirmed experimentally to reduce the transmission rate.

Only a few studies have evaluated the possibility that index cases under treatment might infect their contacts and only five studies have reported epidemiological evidence [52, 78, 79, 81, 93]. The most well known is the Madras study in India by Kamat et al. [79], which to the best of our knowledge is the only randomised study. The other four studies, performed in the USA, although consistent, could have inadvertently allocated less severe cases to the ambulatory arm (although there is no evidence that biased allocation occurred) [52, 78, 81, 93]. The main limitation of all five studies is that 30–65% of household contacts were already infected at the time treatment started and they did not consider the possibility of re-infection. The proportion of the remaining contacts who remained uninfected ranged from 18% to 70% in the different studies. The microbiological status and settings were different, as was treatment duration before exposure [52, 78, 79, 81, 93]. The Madras study found that fewer contacts (both TST positive and negative) of patients treated at home developed TB compared with contacts of patients treated at a sanatorium [79]. Another important finding was that most of the contacts who developed TB did so within the first 3 months (irrespective of whether the index case was treated at home or in the sanatorium). The other four studies had consistent results [52, 78, 81, 93].

In summary, the aforementioned studies do not provide convincing evidence that patients undergoing effective treatment can infect their contacts, regardless of smear or culture status. The studies by Gunnels et al. [52], Riley [53, 54] and Dharmadhikari et al. [55] clearly demonstrate the rapid effect of adequate treatment on infectiousness.

Which tools do we have to monitor treatment and assess infectiousness?

Sputum smear microscopy and culture are the classic surrogate markers for monitoring treatment and assessing infectiousness [94], and form the most common basis of decisions on the patient's contagiousness before effective treatment starts [95]. Early bactericidal activity is still used as an important measure in controlled clinical trials [94].

In a situation of failure (e.g. when the treatment is not effective, smear and culture do not convert, and clinical and radiological indicators deteriorate) the patient is likely to be still infectious.

Colony numbers (in smear cultures) and time or cycles to positivity in molecular detection can also be used to assess the treatment response in clinical trials, but not for assessing infectiousness unless providing evidence of treatment failure [96].

Radiology (number and extension of pulmonary cavities) can also contribute to treatment success monitoring.

Although pharmacokinetics/pharmacodynamics and therapeutic drug monitoring are not considered treatment monitoring tools, they represent additional resources to ensure adequate dosing, thus aiding effective treatment. They have been recently recommended, for the first time, in the North American TB guidelines [75, 97]. Genetic tests, which are core tools for diagnosis, cannot be used to monitor infectiousness due the high rate of false-positive results (caused by the genetic material of dead bacilli) [98–100].

In acknowledging the limitations of sputum smear as an initial diagnostic tool and to guide treatment decisions, Behr et al. [50] suggested that interpretations of infectiousness need to be guided by clinical suspicion and the HIV serostatus of contacts. Their study also revealed that treatment was prescribed 12 days later to sputum smear-negative patients than to those positive.

Fortún et al. [101] measured the time to sputum smear and culture conversion (34 and 38 days after the start of treatment, respectively), and found that 53% of patients underwent culture conversion within the first 4 weeks of treatment. Thus, the time to sputum smear and culture conversion is useful to evaluate treatment effectiveness, but not the patient's infectiousness. Hernández-Garduño et al. [102] identified high colony counts, lung cavities and a prolonged symptomatic period as factors associated with persistent culture positivity in treated cases, in addition to drug resistance.

In summary, monitoring infectiousness is simply ensuring that treatment, based on DST, appears to be effective based on clinical response and, eventually, by follow-up culture.

Bacterial load/disease site

Patients with extensive cavitary lung disease and high bacterial load are highly infectious and remain infectious for longer.

Nutrition/immunity status

The roles of nutrition and immunity in combating TB are well known since the sanatoria era. Interventions to improve nutrition and to restore immunity when necessary (i.e. antiretroviral treatment for TB/HIV co-infected patients) are likely to help [103]. However, there is no evidence to support their role in transmission.

People-centred approach

Enhancing treatment adherence (DOT/video-observed treatment) can improve the patient's response to treatment [35, 104]. Although plausible (adherence is likely to reduce infectiousness, given the core role of effective treatment), no evidence is currently available to support a relationship between a people-centred approach and the treatment response.

Immunotherapy/host-directed therapies

New approaches in this area have been proposed, but there is insufficient evidence on their clinical utility [103].

Bactericidal potentialities of treatment regimens

The bactericidal capacity of the anti-TB regimen in use is likely to reduce the infectiousness period, although there is no experimental evidence to support this. Bactericidal drugs can reduce the number of colonies, but this may not be the only mechanism involved in reducing infectiousness (e.g. the bacilli growing in culture may be not infectious) [76, 105].

Improved treatment programmes

All efforts to improve the quality of treatment (e.g. high-quality drugs, correct doses and regimens, and use of therapeutic drug monitoring) are likely to be useful in reducing infectiousness, although there is no formal evidence to support this.

Implementation of WHO guidance

Sustainable implementation of administrative, environmental and personal protection measures to reduce the duration and degree of infectiousness are prerequisites for health units admitting MDR-TB patients (see later for details) [8, 36, 104, 106].

Active screening

The administrative approach of active screening for undetected TB and drug resistance followed by prompt effective therapy is most likely to result in sustainable reductions of transmission.

Criteria for hospital admission

A proportion of cases still need to be admitted for medical reasons, including severe cases, life-threatening conditions, comorbidities, psychiatric problems and adverse drug reactions, and for social or logistical reasons (table 5) [129].

Features of a hospital admitting TB and MDR/XDR-TB patients

A patient admitted to hospital should never represent a source of transmission to other patients, health staff and visitors. Therefore, the hospital needs to offer adequate infection control measures in addition to core services (table 6) [1, 2].

The WHO suggests that people deemed to be at a low risk of rifampicin-resistant TB and MDR-TB should be placed in single rooms, and that those at a high risk should ideally be accommodated in a negative-pressure room while rapid diagnostic tests are urgently performed until effective treatment starts [104].