Microbe frequencies
We analyzed the patterns of infection with common respiratory agents in a well-defined population of military recruits. The use of highly sensitive multiplex PCR diagnostics allowed an accurate characterization of the spectrum of organisms contained in non-acute and acute samples.
The data indicate co-circulation of several different viral agents, and high frequency of bacterial colonization in both non-acute and acute samples. Up to one third of respiratory illness cases among army personnel are reportedly caused by viral or bacterial infections [
6]. The gathering of individuals from diverse geographic locations and the crowded living conditions increase the risk of microbe transmission in these settings [
7]. Illnesses are usually self-limiting, although the emergence of highly virulent strains can lead to high morbidity and mortality [
8].
Streptococcus bacteria, adenoviruses, coronaviruses and influenza are among the most widely distributed microbes in the military environment, and are implicated in > 50% of febrile illness cases reported at military medical facilities [
6]. We identified each of these organisms in one or more samples. For most of these microbes, overall detection frequencies were comparable in non-acute and acute samples, although influenza B and coronavirus 229 were more commonly identified among acute specimens. Other infectious agents commonly circulating among military personnel include
H. influenzae, rhinovirus, and, to a lesser extent, parainfluenza, RSV, and
L. pneumophila, although their presence does not necessarily imply the occurrence of clinical symptoms [
9‐
11].
H. influenzae and rhinoviruses were the most frequently detected organisms in our population in both non-acute and acute samples. We detected parainfluenza and
L. pneumophila, but we did not find RSV in any of our samples.
Clinical relevance
For individuals developing URI during follow-up, illness etiology could not be unequivocally determined. Among acute samples, Hi-B was the most frequently detected organism. It was the sole agent identified in 12% of acute specimens, while it was co-detected with other microbes in > 50% of acute samples. However, colonisation with Hi-B was also common among non-acute baseline samples, where it was detected alone or in combination with other microbes in 40.5% and 43.3% of specimens, respectively.
For organisms rarely detected among asymptomatic individuals but frequently found in acute samples, a causal association may be more likely. For instance, influenza B was detected in none of the non-acute, but 9.7% of acute samples. Similarly, the proportion of both Hi-nonB- and rhinovirus-positive samples was significantly lower among non-acute specimens collected at baseline compared to acute samples. However, > 85% of acute samples positive for Hi-non B, rhinovirus or influenza B were also positive for one or more additional microbe, so that a causal association could not be determined. Some agents, such as Hi-non B or adenovirus, were most frequently detected in non-acute samples collected at the end of follow-up, possibly indicating post-infectious shedding or persistent infection at sub-clinical levels.
In the clinical setting, overlapping clinical presentations and poor capabilities to determine the etiology of respiratory illnesses often lead to inappropriate treatment with broad-spectrum antibiotics [
12]. This might occur even more frequently in the military setting, where molecular diagnostic tools are usually inaccessible [
6]. Since a considerable fraction of respiratory illnesses is caused by viruses, the unsubstantiated use of antibiotics is particularly problematic, because it can lead to negative health outcomes and promote the development of antimicrobial resistance [
3]. Studies evaluating the impact of multiplex diagnostic procedures on patient management report inconsistent results. In the outpatient setting, access to rapid molecular diagnostic tools for respiratory pathogens significantly reduced antibiotic prescription rates for patients presenting with respiratory illness [
13]. However, these findings were not confirmed in the hospital setting. PCR-based testing failed to reduce hospital admissions and duration of hospital stay in patients with acute respiratory infection [
14,
15]. Although molecular diagnostic tools may help to differentiate bacterial and viral respiratory agents, it is unlikely that antibacterial treatment would be terminated based on the mere presence of viral agents in an acute respiratory sample, especially considering the high rates of bacterial co-infection [
16].
Microbial load
Quantitative or semi-quantitative diagnostic tools can potentially help define clinically significant pathogen densities, and have proven highly valuable to understand the dynamics of diarrheal disease [
17] and to improve the management of gastrointestinal illnesses [
18]. Among acute diarrhea patients, quantitative amplification of norovirus RNA from fecal samples can help determine pathogen load thresholds that effectively distinguish between causal association and sub-pathogenic carriage [
19]. Similarly, rotavirus load correlates with disease severity among children with gastroenteritis [
20]. Because of the crucial role of microbial replication in viral pathogenesis, the value of pathogen load quantitation could be most clearly established for gastrointestinal illnesses of viral etiology, although some evidence is available for bacterial infections as well. For instance, microbial load of enteropathogenic
E. coli is significantly higher among children with diarrhea compared to control subjects, especially when enteropathogenic
E. coli is the sole agent identified [
21].
In this study, tobit regression indicated significantly lower microbial load in non-acute relative to acute samples for rhinovirus, HI-nonB, and
S. pneumoniae. However, due to a substantial overlap in Ct-value distributions, it was not possible to identify a Ct-value threshold indicating causality for any of these organisms. Previous studies assessing the association of viral load with clinical symptoms of respiratory infections reported similar findings. Mean viral load for rhinovirus and six additional viruses was significantly higher in upper respiratory tract aspirates from children with pneumonia compared to healthy controls, but the overlap in viral load distribution was substantial [
22]. In pediatric patients, high rhinovirus load was associated with the presence of lower respiratory tract symptoms [
23,
24], but a threshold for clinical relevance could only be determined if rhinovirus was the sole agent identified [
24]. Additional studies reported a correlation between microbial load and occurrence or severity of respiratory symptoms for RSV [
25], bocavirus [
26], and human metapneumovirus (HMPV) [
27,
28], although these findings were inconsistent [
29,
30] or conditional on the presence of the virus as a single microbe [
31]. We did not detect any significant association between microbial load and clinical manifestations for viruses other than rhinovirus.
For both
H. influenzae and
Streptococcus species, previous studies reported a significant correlation of bacterial densities with clinical manifestations of disease [
32]. In young patients with acute respiratory tract infection,
S. pneumoniae load fluctuated with symptom incidence and resolution [
33]. Among children hospitalized with pneumonia, median nasopharyngeal
S. pneumoniae load was substantially higher compared to healthy controls [
32]. Pneumococcal density was also associated with severity of symptoms [
34] and increased duration of children’s hospital stay [
35]. Similar associations were observed in pneumonic adults, although the correlation was not significant in this population [
36].
The association between microbial load and clinical manifestations may depend on specific pathogen-host interactions. If pathogenesis is primarily related to microbial replication, a stronger correlation between microbial load and illness magnitude may be observed [
37]. If clinical manifestations are largely attributable to host immune defences or bacterial toxins, the correlation with microbial load may not be obvious [
37]. Temporal variations in microbial load may also play an important role if the quantity of nucleic acid is significantly more abundant at the time and location of pathology [
30,
33]. In acute respiratory illness patients, high bacterial colonization densities are often associated with the presence of viral co-infections [
38], and clinical manifestations may vary depending on specific co-infection patterns [
39].
The ecology of respiratory pathogens is also likely to be influenced by the living conditions in military settings. Mixing of individuals from diverse backgrounds living in close-quarters with high levels of inter-personal contact increases the potential for introduction and spread of multiple microbes in this population, which could account for the broad range of organisms and co-detections in this study.
Strengths and limitations
We analysed both non-acute and acute samples from a closely monitored population in a semi-closed, longitudinal setting. The study population was well-defined and relatively homogeneous with regards to demographics and living conditions. However, our findings may not be applicable to populations with different socio-demographic characteristics and populations outside the military environment, such as cohorts of children among whom the impact of respiratory infections may be greater.
The frequent co-detection of multiple respiratory agents and the failure to distinguish between viable and dead organisms, or microbes that colonize the host at sub-pathogenic levels, may prevent the unambiguous interpretation of test results [
2]. A positive result may indicate illness aetiology, asymptomatic colonisation, post-infectious shedding, or an incipient infection. Therefore, Ct-values may not always be a reliable surrogate for infectious load.
Samples from only two out of six cohorts were tested by real-time PCR. Although there might be bias from seasonal effects, these are usually less pronounced in the tropics. Given the relatively low frequencies of viral detection, a larger sample size and a longer follow-up may have captured a more precise picture of infection patterns in this population. This study was also limited to the detection of organisms contained in the respiratory panel. We cannot exclude the presence of additional organisms in our specimens. In addition, the data were obtained from throat and nasal swab samples, but our findings may not apply to nasopharyngeal or sputum specimens. Finally, the quality and quantity of material obtained through nose and throat swabs may differ significantly among subjects, and the success of PCR-based methods also depends on the availability of intact genome sequences and the absence of random mutations.