Pneumocystis pneumonia (PCP), caused by Pneumocystis jirovecii, is an important opportunistic infection in HIV-infected children [1, 2]. The incidence of PCP in developed countries has declined since the introduction of highly active anti-retroviral therapy and use of chemoprophylaxis. However, PCP remains a major cause of hospitalization and mortality in HIV-infected children in low or middle income countries, [1, 3–5] with reported incidence rates of 10 - 49%, [1, 3, 6] and in-hospital case-fatality rates of 20 - 63% [1, 3, 4, 6]. Apart from HIV infection, there are other factors that predispose children to developing PCP including malnutrition, other immune deficiencies or HIV exposure. Untreated, the case fatality rate in children with PCP approximates 100% [1, 3, 4, 6]. However, diagnosis can be difficult as clinical and radiological findings are non-specific. Therefore, a rapid, accurate laboratory diagnosis is important for timely use of appropriate medication.

Detection of P. jirovecii is hampered by the lack of a sustainable in-vitro culture method [7]. Standard laboratory diagnostic methods are microscopic examination of a lower respiratory tract sample with Gomori Grocott's methenamine silver nitrate stain or the more sensitive immunofluorescence assay (IF) [8] on bronchoalveolar lavage (BAL) or induced sputum (IS) specimens. Using microscopy, the yield from IS has been reported to be similar compared to that from BAL [9]. However, sputum induction in children is not widely performed, requires staff trained to do the procedure and may result in clinical deterioration or nosocomial transmission of respiratory pathogens. Diagnosis using a non-invasive sample such as a nasopharyngeal aspirate (NPA) is, therefore, desirable.

The clinical sensitivity from a NPA has been reported to be low and variable when microscopy is utilised [4, 5, 10]. The development of polymerase chain reaction (PCR) techniques has provided a more reliable diagnostic method. In adult studies, PCR is as specific as and more sensitive than microscopy for diagnosis when performed on respiratory specimens, including oral washes [7, 11–21]. In a study of oropharyngeal washes from HIV-infected adult patients, P. jirovecii DNA-amplification had a sensitivity of 44% using a nested PCR protocol compared to trans-bronchial biopsy, [16] increasing to 90% when touch-down real-time PCR was utilised [7, 11]. Real-time PCR also allows for quantification of the organism load and with application of cutoff values, could improve the specificity by distinguishing between colonization and infection [13].

The aims of this study were (1) to compare a real-time quantitative PCR assay with IF for the diagnosis of PCP in children and (2) to evaluate the reliability of PCR for the diagnosis of PCP in upper compared to lower respiratory tract secretions.

This study found that PCR had a much higher sensitivity than the current standard diagnostic test of IF, with a more than 5 fold increased detection in upper and lower samples using PCR. In addition, an URT specimen was as reliable as a LRT specimen for PCR confirmed diagnosis. To our knowledge, this is the first study in children that demonstrates that P. jirovecii can be reliably detected using PCR on an upper respiratory tract specimen. The increase in diagnostic yield using PCR occurred in both upper and LRT specimens, but was especially marked in upper respiratory samples. This is important as obtaining a LRT specimen using BAL or IS in children can be difficult whereas obtaining a NPA is much easier, and carries a lower risk of nosocomial transmission of infection, thus making the diagnosis of PCP feasible in many health-care settings, including in primary care. The detection of organism in an URT sample may occur due to extension of infection to the upper airways in severe disease or by organisms that are propelled to the upper airway during coughing [16].

Using current methods of detection with staining, PCP may therefore be substantially under-diagnosed in children. This is supported by the low rates of laboratory confirmation of approximately 20% reported in clinical studies of children with suspected PCP [1, 3]. When IF on LRT samples was used as the reference standard, the sensitivity of the PCR assay was 100%.

A further advantage of PCR is the rapid turn-around time of the test, with results obtainable within hours. Children with PCP often have a rapidly progressive course, thus timely diagnosis would allow early initiation of treatment in those not on treatment, with potential reduction in mortality. The usual management of suspected PCP is empiric treatment with co-trimoxazole and corticosteroids in children who are hypoxic. Rapid confirmation of diagnosis would therefore also be useful for guiding corticosteroid therapy. This is especially relevant given the concerns about use of corticosteroids in other infections such as CMV pneumonitis or TB in children, without specific antimicrobial therapy for these pathogens [23, 24]. The clinician may also consider discontinuation of co-trimoxazole therapy in patients with negative PCR results and contemplate alternative diagnoses.

The specificity of PCR for PCP is more difficult to determine as the organism cannot be cultured and there is no method to distinguish disease from asymptomatic colonization. The large number of IF-negative, PCR-positive samples may therefore represent children with PCP or with pneumocystis colonization. As all children were treated for PCP based on their clinical presentation, it was impossible to use response to therapy to distinguish active disease from colonization. The organism has been reported to colonize the airways of adults with chronic obstructive airway disease, with these individuals forming the reservoir for infection of susceptible hosts [25]. Moreover, positive PCR results have been reported in BAL samples from adults without clinical PCP, HIV, or other infections [26]. However, all children in the study had an underlying risk factor for PCP and were severely ill with a clinical presentation consistent with PCP. In addition most children were under 6 months of age, the highest risk period for primary PCP rather than colonization. It is, therefore, likely that a positive PCR result in these children represents disease.

To further investigate the specificity of PCR for PCP samples from 30 control children with confirmed viral LRT illness who improved clinically without therapy for PCP, were tested. In contrast to the detection rate in cases (110, 55%), only a minority (2, 6%) of control children were PCR positive. The 2 control positives may represent primary infection which is very common in children in the first years of life, and may manifest as a viral-like illness. Another possible explanation may be laboratory contamination. However, this is unlikely as the laboratory layout allows only unidirectional work-flow with strict separation of all steps in the PCR process. The real-time PCR was performed in a closed system, with detection occurring on the same platform thus minimizing the risk of contamination.

An alternative strategy for differentiating colonization from disease is by applying cut-off values to quantitative PCR results. Huggett et al were able to achieve a specificity of 96% in BAL specimens from adult patients by using a quantitative PCR for HSP70, with appropriate cut-off values [27]. Unfortunately, the quantitative results obtained in our study could not be accurately interpreted as data was not normalized by use of a reference gene as an internal control. Normalization may control for variations in the extraction process, patient variability and inter-sample variability by accounting for the dilution factor in the BAL fluid. However, there is controversy about the choice of the most appropriate reference gene [28]. If normalization can be reliably performed, it may allow for more accurate estimation of colonization rates and of cut-off values to distinguish disease thus improving the specificity of the assay. Further work in this area is needed.

Potential difficulties in achieving widespread implementation of PCR for PCP diagnosis include cost and the need for a specialized molecular laboratory, equipment, and personnel. However, much progress has been made in developing closed molecular amplification platforms for the diagnosis of TB at point-of-care facilities. Given the scale of the HIV pandemic and the substantial mortality associated with PCP particularly in children, a similar approach for the diagnosis of PCP, should be undertaken.

In summary, this real-time PCR-based assay has a much higher sensitivity than current diagnostic tests and allows for the use of a NPA sample for reliable detection of P. jirovecii in children. PCR on a NPA should replace the traditional diagnostic method of IF or silver stain performed on LRT samples. The possibility of making a reliable diagnosis on a NPA represents an important advance in confirming PCP in children. Widespread implementation of this technology is needed especially in high burden HIV and resource limited settings.