Chapter 66: Controversies: Invasive Versus Noninvasive Strategy for Diagnosing Respiratory Failure

The etiology of acute respiratory failure (ARF) among different patient populations is highly variable.¹ In general, pulmonary infections are the leading cause of ARF followed by heart failure, exacerbation of chronic obstructive pulmonary disease (COPD), and sepsis.² Early and appropriate diagnostic strategies are vital for the initial choice of therapy and subsequent treatment decisions. With advances in medicine, aggressive treatments have been introduced to achieve the highest possible cure which in turn, has resulted in the concomitant rise in the incidence of life-threatening toxic and infectious complications, particularly involving the lungs.

There are many practical questions surrounding the diagnostic work-up of critically ill patients with ARF. What is the utility of noninvasive testing? Is empiric treatment enough? What are the risks and benefits of invasive diagnostic procedures? Does invasive testing result in improved outcomes? To answer these questions, a review of the literature has produced mixed results. The apparent dilemma between the need to identify the cause of ARF and complications associated with invasive procedures may have created this uncertainty. To this end, there is some data in hematology and oncology patients which can only be extrapolated to other groups of immunocompromised hosts for whom further testing guided by clinical and epidemiologic data may reveal unsuspected diagnoses. In this chapter, we discuss the utility of noninvasive and invasive testing for diagnosing ARF and provide a diagnostic algorithm based on the best available data (Figure 66–1).

This approach consists of obtaining chest imaging studies, cardiac biomarkers (eg, B-type natriuretic peptide) and echocardiography to exclude cardiogenic pulmonary edema, and serologic and microbiologic studies of sputum, nasopharyngeal (NP) aspirates, blood, and urine to diagnose infection. In addition, newer molecular techniques are being implemented along with conventional methods in order to identify specific pathogens not only faster but more accurately without exposing patients to additional risks.

The most commonly performed noninvasive tests for diagnosing ARF are shown in Table 66–1.

Important causes of ARF such as sepsis, aspiration pneumonia, diffuse alveolar hemorrhage, eosinophilic pneumonia, transfusion-related lung injuries and respiratory failure associated with cancer and other immunocompromised states are diagnosed clinically using a constellation of radiographic and other ancillary investigations.

The clinical presentation of suspected pulmonary infections is usually nonspecific. This is particularly important in immunocompromised patients who are prone to develop pulmonary infections with certain pathogens such as herpes simplex virus, Cytomegalovirus (CMV), Pneumocystis jirovecii (PCP), Mycobacterium tuberculosis, and Aspergillus species and hence require specific diagnosis and treatment as opposed to empiric therapy.

Gram stain of respiratory secretions, available within a few hours, may help in narrowing or broadening the antimicrobial spectrum but lacks sensitivity and specificity. Similarly, the yield of conventional stains for the diagnosis of other infections such yeast, other fungi, Aspergillus, M tuberculosis, and PCP pneumonia have been disappointing and slow which has led to a greater reliance on polymerase chain reaction (PCR)-based techniques. The latter has provided improved insight into the biology and genomic structure of the pathogens.

Azoulay et al³ performed a multicenter randomized controlled trial (RCT) comparing noninvasive testing (of sputum, induced sputum, NP aspirates, serum and urine as well as chest imaging and echocardiography) with an invasive strategy involving fiberoptic bronchoscopy with bronchoalveolar lavage (FOB-BAL) in 219 hematology and oncology patients with ARF. Bacterial infection was the primary etiology of ARF followed by infections with viruses, yeasts and molds, or Pneumocystis. Noninvasive testing provided the diagnosis in 71% of cases and resulted in a change in management in 44% of patients compared to only 34% in those who underwent FOB-BAL. In 20% of cases, no diagnosis was made with either strategy. Additionally, the time to diagnosis of bacterial, non-Aspergillus, and viral infection was shorter and available quicker with noninvasive testing as compared to invasive testing. However, the diagnosis of PCP took longer in the noninvasive group (3 [-7] days vs 1 [0-1] day). In both groups, a specific diagnosis was established in 80% of cases. These diagnostic rates were significantly higher than those reported in the earlier observational study conducted by the same authors suggesting the improved utility and advances in noninvasive testing in recent years.⁴

The sensitivity of nasal swab and NP aspirate is comparable for all respiratory viruses except for the detection of respiratory syncytial virus due to the inherent lability and relatively lower viral load present in the nasopharynx.⁵ Although serologic testing for CMV IgM and IgG may be used to screen for primary CMV infection or exposure, this is not used for the diagnosis of active CMV infection. Viral load testing in serum, tissue specimens, or BAL fluid using both quantitative nucleic acid amplification testing (QNAT) and antigenemia testing are currently the cornerstone for diagnosis and monitoring for CMV infection and disease.⁶ Real-time QNAT testing for CMV is now the standard of care given its better precision, broader linear range, faster turnaround time, higher throughput, and less risk of contamination compared with conventional PCR tests. In contrast, the antigenemia test is labor intensive, lacks a standardized cutoff value, and the assay performance diminishes when the absolute neutrophil count is less than 1000/mm³.⁶

Streptococcus pneumoniae urinary antigen has sensitivity of up to 78% in nonbacteremic patients and 80% in bacteremic cases and a specificity of more than 90% in adults. However, there have been false-positive reports of pneumococcal infection rates ranging from 21% to 54% in children with NP carriage and no evidence of disease.⁷ This limitation may be overcome with quantitative real-time PCR testing of NP swab samples to estimate NP colonization density and using cutoff values more than or equal to 8000 copies/mL, which improved the sensitivity to 82% and specificity to 92% for distinguishing pneumococcal pneumonia from asymptomatic colonization.⁸ Similarly, detection of Legionella pneumophila antigen in urine is rapidly available and requires no specialized laboratory equipment. Sensitivity rates of 73% and 77%, respectively, were reported for the new Meridian TRU Legionella assay and BINAX urinary antigen test. Additionally, the sensitivity of the Meridian TRU Legionella test increased to 81% after 60 minutes of incubation.⁹

PCP cannot be cultured and hence diagnosis relies on visualization of the fungus on microscopic examination of respiratory specimens. The sensitivity of microscopy varies according to the staining technique and type of respiratory sample used. In one study, the sensitivity and specificity of Grocott–Gomori methenamine silver stain for PCP were 79.4% and 99.2%, respectively.¹⁰ A serum assay for beta glucan, a cell wall component of most pathogenic fungi, is currently being used for PCP diagnosis with sensitivity and specificity rates of 100% and 96.4%, respectively, using a cut off value of 100 pg/mL.¹¹ In a select group of patients the yield of induced sputum-PCR assay for detection of PCP has been shown to be excellent with sensitivity of 100% and specificity of 90% which would make FOB-BAL or lung biopsy practically unnecessary.¹² Of note, respiratory specimens should be collected before or immediately after initiation of therapy because the sputum-PCR assay for detection of PCP may turn negative quickly. A negative PCR may allow for the discontinuation of anti-PCP therapy.¹²

Both direct microscopy and cultures are insensitive methods to diagnose aspergillosis. Galactomannan is an Aspergillus-specific polysaccharide which is released during Aspergillus growth at the site of infection. This antigen can be detected by quantitative serum galactomannan index and is used as a biomarker of the disease. In addition, galactomannan antigen has a strong correlation with clinical outcome; thus the test is currently repeated frequently during the course of treatment. In contrast, the presence and magnitude of the galactomannan index in BAL fluid does not have any mortality implication.¹³ In a study of 500 patients, the diagnostic yield of acid-fast bacilli (AFB) smear, cultures and PCR of induced sputum for M. tuberculosis was increased consistently with repeated induction from 64%, 70%, and 89% for AFB smear, and from cultures and PCR, respectively, to 98%, 100%, and 100% on fourth induction.¹⁴ Current amplification techniques can detect M. tuberculosis and rifampin resistance directly from clinical specimens in approximately 2 hours with high sensitivity compared to the standard method.¹⁵

Early molecular detection methods are currently under development. With regards to bacterial pathogens, multiplex amplification assays might include frequently involved microorganisms. Despite the obvious advantages of PCR-based methods (rapidity, sensitivity, convenience), several limitations including the need of cut-off levels to distinguish between colonization and infection, the inability to detect living viruses from prolonged harmless shedding, lack of information about the antibiotic susceptibility and the high cost of these tests still limits their widespread use in clinical decision making. Moreover, studies measuring clinical outcomes and mortality benefits are lacking at this time. These limitations in diagnostic testing with noninvasive strategies have led to the development of antimicrobial guidelines for empiric treatment.¹⁶ At this time, PCR-based testing can supplement rather than replace culture-based methods for pathogens where antibiotic resistance is a concern.

This approach relies on FOB-BAL and/or protected brush specimens (PBS), transbronchial biopsy (TBB) and lung biopsy (open or video assisted). FOB-BAL and/or PBS permit collection of distal airway secretions with minimal oropharyngeal contamination which is inevitable during specimen collection.

The reported diagnostic yield of FOB in different studies has been variable. This is attributed to the heterogeneity of the patient population, severity of illnesses, regional differences, timing, and techniques of FOB-BAL and/or PBS.^17,18 A summary of these studies is shown in Table 66–2.

In a prospective study of 101 immunocompromised patients with suspected pneumonia, 80% of whom had human immunodeficiency virus infection and/or acquired immunodeficiency syndrome, hematologic and solid organ malignancies, chronic steroid use and neutropenia, FOB-BAL/TBB had a general diagnostic yield of 52% but 76% in infectious diseases.¹⁹ Mycobacterial infections accounted for one third of infections followed by bacterial pathogens and PCP. In contrast, the diagnostic yield of FOB in immunocompetent patients is lower. In one study of nonselected patients, the diagnostic yield of FOB was 53% in immunocompromised patients versus only 19% in immunocompetent patients suggesting that the need of invasive testing in the latter group must be established based on clinical importance and risks of the procedure.²⁰ The yield was further decreased if patients received antibiotics prior to bronchoscopy. Complications were seen in 5 patients (12%) with a higher proportion of minor self-limited bleeding (n = 3), pulmonary edema (n = 1), and cardiac arrhythmias (n = 1).²⁰

In a retrospective study of 501 patients who underwent hematopoietic stem cell transplantation, the mortality rate was lower in the early (< 4 days) FOB-BAL group (6%) versus 14% to the late FOB group (> 5 days). The overall diagnostic yield for clinically significant pathogens was 55% and was highest (75%) if the FOB-BAL was performed with 24 hours.²¹ Similarly, a multicenter RCT in patients with suspected hospital-acquired pneumonia (HAP) where an invasive strategy (direct examination of FOB-BAL or PSB samples and quantitative cultures) was utilized compared with patients managed clinically, showed that the invasive approach was associated with reduction in mortality on day 14 (16% vs 25%), decreased organ dysfunction score on day 3 and 7, less use of antibiotics, mean number of antibiotic-free days (7 ± 7 vs 11± 9 days).²²

Reliance on semiquantitative cultures from endotracheal aspirates which may not distinguish a true pathogen from a colonizer, can lead to more or broader antibiotic use and potential delay in the diagnosis of extrapulmonary infection. Thus, semiquantitative cultures of tracheal aspirates cannot be used reliably as quantitative cultures to define the presence of pneumonia and the need for antibiotic therapy.¹⁶ However, it is recommended that patients with community-acquired pneumonia should be investigated for specific pathogens only if it would alter the standard (empiric) management decisions when the presence of such pathogen is suspected based on the clinical and epidemiologic clues.¹⁶ For example, such alteration in therapy is likely to benefit patients with psittacosis, tularemia, endemic fungi, M. tuberculosis or antibiotic resistance issues.

FOB-BAL is also favored in the presence of noninfectious pulmonary infiltrates causing ARF since as described previously, noninfectious disorders explained most of the missed diagnoses.^3,19 In addition, transbronchial or surgical open lung biopsy (OLB) may be considered in select patients to obtain a definitive diagnosis and guide therapy when it is believed to make a substantial impact in the management and improvement in the prognosis. These include patients with suspected cryptogenic organizing pneumonia, fungal lung infections, and acute exacerbation of a chronic interstitial lung disease, vasculitis, or disseminated cancer. In a prospective study of 100 nonselected patients with ARF of unknown etiology after a negative BAL, OLB provided a contributive result in 78% of patients. A change in the management was associated with improved survival to 67% compared to 14% when OLB did not help in management. The most common diagnoses obtained by OLB were fibrosis associated with infections (55%), followed by pulmonary infections (42%) (CMV, herpes, M. tuberculosis, invasive aspergillosis) and miscellaneous diagnoses such as systemic lupus erythematosus, bronchoalveolar carcinoma, intraalveolar hemorrhage, drug toxicity, and carcinomatous lymphangitis.²³

FOB is commonly associated with alteration in gas exchange; however, hypoxemia is usually mild and transient.²⁴ The severity of hypoxemia during bronchoscopy may vary depending on the patient’s comorbid conditions, indications of the procedure and the type of bronchoscopic interventions performed. The risk of hypoxemia may be further decreased when using noninvasive ventilation and/or target controlled sedation during bronchoscopy which have both been shown to not only limit discomfort but also reduce the subsequent respiratory worsening.²⁵ In various studies, the risk of FOB-BAL has been proven to be small with no effect on outcome.^3,26 In few selected patients, the diagnostic yield improves if FOB-BAL is performed in the ICU within 24 hours of admission; thereafter, this benefit seems to fade with less favorable risk/benefit ratio.³

A multicenter prospective observational study evaluated the safety of FOB in critically ill nonintubated patients with hypoxemic ARF due to a variety of causes including immunodeficiency (37%), atelectasis (29%), healthcare-associated pneumonia (27%), acute diffuse pneumonia (27%), community-acquired pneumonia (12%), hemoptysis (3%), suspected malignancy (3%), and chronic diffuse infiltrative pneumonia (1%). The patients had a PaO₂/FiO₂ ratio of less than 300 and were on noninvasive ventilation or receiving oxygen supplementation of more than 8 L/min. FOB provided the diagnosis in 59% of cases with subsequent change in the management in 51% of patients. Factors independently associated with the need of invasive ventilatory support were COPD and cancer but not the extent of radiologic opacities, PaO₂/FiO₂ ratio, BAL, or injected BAL volume. The overall incremental risk associated with bronchoscopy is small and has no effect on mortality. It is generally considered safe with arterial oxygen saturation more than 90% on FiO₂ of less than 0.5 and PaO₂ more than 60 mm Hg.

Recent advances in the sensitivity and specificity of noninvasive diagnostic investigations may change the risk/benefit ratio in favor of conservative noninvasive testing in patients with ARF. However, a subset of patients may still benefit from the addition of FOB-BAL when performed safely in addition to noninvasive testing, particularly when protective procedural strategies such as noninvasive ventilation and/or targeted controlled sedation are used. It is hoped that further refinement of molecular techniques and their enhanced diagnostic yield will decrease the need for invasive procedures in the future. Current evidence suggests that noninvasive and invasive strategies both offer benefits and complement each other. The main consideration in deciding which approach to take is the influence any result would have on management. Noninvasive tests are generally useful in determining the accurate diagnosis in immunocompromised patients. On the other hand, FOB-BAL and other invasive techniques have an advantage especially in the setting of suspected M. tuberculosis, PCP, malignancy and interstitial lung diseases. Invasive testing should be reserved for immuncompetent patients only when conservative techniques are likely to miss the diagnosis. We conclude that an individualized approach based on available clinical information and epidemiologic data should facilitate the relevant use of noninvasive and/or invasive testing for diagnosing ARF.