Acute lower respiratory illnesses (ALRI) are associated with substantial morbidity and mortality in sub-Saharan Africa¹. Understanding etiology of respiratory illnesses in resource-limited countries may support clinicians with patient management and provide a framework for decision makers when prioritizing public health interventions. Recently, with availability of multi-pathogen molecular diagnostic methods, and increasing recognition of viruses in cases of ALRI, the use of nasopharyngeal (NP) and oropharyngeal (OP) swab specimens to identify pathogens associated with ALRI has increased^2–4. However, NP/OP swabs may not accurately reflect etiologic agents from the lower respiratory tract⁵. Sputum specimens, on the other hand, have been considered a more representative sample of the lower respiratory tract⁵, but have not been extensively explored, especially in tropical countries.

A multisite study in the United States compared pathogen yield between NP/OP and sputum specimens and argued for the use of high-quality sputum specimens for greater pathogen yield⁶. In contrast, another multisite study in low-middle income countries, observed that the pathogen yield from NP/OP and sputum specimens were similar². Given the paucity of data and the difference in observations in these two recent studies, and that most data from tropical countries were derived from paediatric populations, we aimed to investigate adults living in a tropical, low-income setting. Here we describe the performance of NP/OP swabs against sputum specimens for detecting respiratory pathogens among adult patients hospitalized with ALRI using the TaqMan Array Card (TAC).

From March 2014 through July 2015, we tested 294 pairs of NP/OP swabs and sputum specimens using TAC. On average, NP/OP swabs and sputum specimens were collected 29 (Standard Deviation [SD] = 14) and 38 (SD = 31) hours after admission, respectively.

Most patients (198/294; 67%) were between ages 18–49 years, and 81/294 (28%) were male. Fever was either self-reported or measured (≥38 °C) in 40% of the patients, and malaria was diagnosed in 19% of the patients. More than half the patients (179/294; 61%) had underlying chronic illness; 39% and 22% for HIV and other chronic conditions respectively. Of the 294 patients, 116 (39%) sought care in <5 days from respiratory illness onset, 99 (34%) sought care 5–10 days and 79 (27%) sought care after 10 days. Of all cases, 88 (30%) were hospitalized for >7 days and 24 (8%) died in hospital (Table 1). Of the 24 who died, 9 (38%) were HIV-positive, 3 (13%) had malaria and 3 (13%) had other chronic illnesses.Table 1

Characteristics of patients hospitalized with acute respiratory illness at Siaya County Referral Hospital, March 2014–July 2015 (Total patient population N = 294).

Of the 294 pairs of specimens tested, 248 bacterial and 136 viral pathogens were detected from 141 and 115 NP/OP specimens respectively whereas 213 bacterial and 47 viral pathogens were detected from 141 and 45 sputum specimens respectively. NP/OP specimens were more likely to test positive for any viral pathogen compared to sputum specimens, p < 0.01. There was no difference in bacterial pathogen detection in NP/OP vs. sputum specimens, and the pathogen specific agreements ranged between 85–100% (Table 2).Table 2

Comparison of respiratory pathogen yields in nasopharyngeal/oropharyngeal (NP/OP) vs sputum specimens, March 2014–July 2015.

Among the bacterial pathogens detected, Streptococcus pneumoniae and Moraxella catarrhalis were more significantly present in NP/OP specimens compared to sputum specimens. However, Klebsiella pneumoniae and Mycobacterium tuberculosis were more significantly present in the sputum specimen compared to the NP/OP specimen. Among the viral pathogens, Respiratory Syncytial Virus (RSV), Rhinovirus, and Human Coronavirus 2 (NL63) were commonly detected in NP/OP than sputum specimens (Table 2). When stratified by HIV status, malaria infection, prolonged hospitalization and duration from onset to admission, there was not much difference in pathogen yield. Although, Mycobacterium tuberculosis (detected in sputum) was associated with HIV positive, malaria negative, and severe cases, and with patients admitted ≥5 days from illness onset. Also, patients with malaria were more likely to test positive for Pseudomonas aeruginosa from sputum compared to NP/OP specimens, p = 0.02 (Supplementary Tables S1–S4). Streptococcus pneumoniae, Haemophilus influenzae and Moraxella catarrhalis were the most co-detected bacterial pathogens and rhinovirus was the most co-detected viral pathogen (Supplementary Tables S5 and S6).

The median C_T values shown in box-and-whiskers plot were comparable for NP/OP and sputum specimens for all bacterial targets (Fig. 1). This was also observed for most viral pathogens, except that a higher concentration of viruses (corresponding to lower C_T values) was observed for influenza B, RSV and human metapneumovirus in the NP/OP specimens compared to sputum specimens. The median C_T values were similar in the NP/OP and sputum specimens for the RNP target.Figure 1

Box- Whiskers plot of threshold cycle (Ct) values for bacterial pathogens (panel A) and viral pathogens (panel B), detected by TaqMan Array Cards.

We found good agreement (85–100%) in the detection of respiratory pathogens between NP/OP swabs and sputum specimens when using TAC. There was no overall benefit of using sputum over NP/OP swab specimens for etiologic assessment of respiratory illness, except that we found improved detection of Mycobacterium tuberculosis and Klebsiella pneumoniae from sputum specimens compared to NP/OP. When considering viral pathogens, rhinovirus, RSV, and coronavirus (HCoV NL63) were more often identified in NP/OP swabs. The use of molecular tests in NP/OP and sputum specimens warrants further investigation as a tool to identify etiology of ALRI in hospitalized adults.

Although previous pneumonia related studies have observed more bacteria yield from sputum^11,12, arguing that this would be the ideal specimen for identification of lower respiratory tract infection etiology. We did not observe substantial differences in pathogen detection from NP/OP vs. sputum specimens. Mycobacterium tuberculosis, was mostly detected in sputum specimens, however, one case was detected in the NP/OP swab and not found in the sputum specimen. Though others suggested that this could happen in cases of extra-pulmonary tuberculosis¹³, we found that this patient was positive by GeneXpert (data not shown) which may suggest some shortfalls of testing for Mycobacterium tuberculosis in sputum using TAC.

Jeong et al. argued that as viruses play an important role in lower respiratory tract infections, sputum would be the best choice for etiologic diagnosis of virus-associated respiratory illness. The authors looked at adults who were hospitalized with severe acute respiratory illness, although the difference in yield from the two types of specimens was not impressive (15% increase in viral detection using sputum). Moreover, they only collected NP swabs which could have led to less representation of respiratory viruses with tropism to the oropharynx¹⁴. Wolff et al. also observed the detection of more viruses in sputum, where C_T values were earlier for almost all pathogens compared to that found in NP/OP swabs⁶, leading the authors to argue for the benefit of sputum use in diagnosis of lower respiratory tract infections. However this position was not supported in the recent work done by Thea et al.², where they looked at children <5 years hospitalized with ALRI and pneumonia, or by our study done among adults. Perhaps in tropical countries, the likelihood of multi-pathogen detection may be higher than that seen in temperate countries, for children and adults alike, which may merit further investigation.

Both Wolff et al. and Thea et al. used high quality induced-sputum specimens in their studies but acknowledged that contamination would still be possible^2,6. Furthermore, Thea et al. noted that the yield of pathogens from high quality induced sputum was similar to that of low quality induced sputum suggesting that specimen collected from the upper respiratory tract (i.e., NP/OP) could be used when investigating ALRI. The biggest challenge in these studies, including ours, is the interpretation of multi-pathogen detection, particularly when trying to attribute a specific pathogen as the cause of an episode of ALRI. Many bacterial pathogens detected from respiratory specimens could reflect carriage, being part of the normal microbiome, and not necessarily associated with the disease process. Similarly, viral pathogens detected in respiratory samples could indicate recent past infections. The early C_T values from the upper tract could indicate higher pathogen load and be associated with pneumonia¹⁵, which would suggest that NP/OP swabs could be used when investigating lower ALRI etiology if interpretation is combined with C_T values. However, establishing causality can be challenging even when considering organism density (here we used C_T values as proxy). The use of radiographic images, epidemiologic data, clinical course of illness, and the possibility to investigate lower respiratory tract specimens collected aseptically could be important tools to add to the investigation of ALRI/pneumonia.

A main limitation in our study is that direct comparison of our findings with those from the literature is complicated due to varying case definitions, severity of cases investigated, age of patients, study settings (temperate vs. tropical, less vs. more resourced countries), and laboratory tests used. Secondly, we did not use induced sputum collection as we worked with a population of adult patients, most of whom were accustomed to providing sputum specimen for tuberculosis testing. Moreover, as used by Thea et al.², we considered <10 squamous epithelial cells per low-power field to determine specimen quality and observed that RNP C_T values were similar for NP/OP and sputum specimens, giving us confidence that results from sputum specimens were credible.

In conclusion, our findings suggest that there is no clear advantage in using sputum specimens over NP/OP swabs to investigate the etiology of ALRI using molecular diagnostics. Interpretation of multi-pathogen molecular laboratory tests, however, needs further investigation, especially in tropical settings where co-detections seem very common.