Is rapid on-site evaluation during bronchoscopy useful?

Rapid on-site evaluation (ROSE) of cytologic material obtained with fine-needle aspiration procedures was long used in the evaluation of lesions located in deep-seated organs/structures throughout the body. The thorax was no exception to the rule, as ROSE was used especially to assist the clinicians in the diagnosis of lesions that are not directly visible at a standard bronchoscopic airways examination, such as those located in the mediastinum or in the pulmonary parenchyma.

Unfortunately, studies reported different results on the influence of ROSE during bronchoscopy, and its use in different interventional pulmonology centers is extremely variable. Interestingly, a critical literature review shows that the results associated with the use of ROSE in this setting were significantly different according to two different periods, namely the period of the observational trials (1990–2010) and the period of the randomized trials (2011–2013).

The ‘era’ of the observational trials

The early studies aimed at evaluating the value of ROSE suggested that on-site review was highly beneficial, as it was found to increase the specimen’s adequacy [1] and/or the diagnostic yield [1–3] of the procedure (conventional transbronchial needle aspiration [TBNA]) whose samples were evaluated in real-time. ROSE, in particular, was shown to improve the diagnostic yield, which was the outcome that was cared for most, by a rate ranging from 25 to 46% [1–3]. The usefulness of ROSE was undoubted since 2005, when Baram et al. published another observational trial which, for the first time, failed to show an increase of the diagnostic yield or the specimen’s adequacy for the patients undergoing ROSE, yet suggested that on-site review allowed to avoid additional sampling without loss in diagnostic yield [4]. The results of this study prompted a critical re-evaluation of the clinical research regarding ROSE, which led to the identification of several problems limiting the reliability of the results from the above studies [5,6].

The first key problem of the observational trials is the extremely high risk of selection bias. None such studies, in fact, set a priori parameters for the allocation of the patients in the ROSE or no-ROSE arms, a practice which makes it impossible to rule out that more complex cases were allocated to the ROSE arm or vice versa [1–4]. The success rate of conventional TBNA is influenced by a number of factors besides ROSE, such as size and location of lymph nodes, experience of the examiner, needle type used, underlying disease, prevalence of the disease being ascertained [7]. In the absence of a method of patients’ allocation able to sterilize the confounding effect of the above factors, that is randomization, it is basically impossible to attribute reliably to the on-site review the observed results. Furthermore, the study population of these trials was often extremely heterogeneous, as it included patients undergoing bronchoscopy for lymphadenopathy alone, peripheral lesions alone, and both [1–4]. As the success rate of bronchoscopy in the setting of lymphadenopathy and peripheral lesions is, per se, very different, ROSE could theoretically have a different utility in these different clinical settings.

The ‘era’ of the randomized trials

In the past few years, three randomized trials have been published that significantly contributed to shed light on the true role of ROSE during bronchoscoscopy, in particular for the diagnosis o lymphadenopathy.

The first randomized trial was aimed at evaluating the usefulness of ROSE in clinically unselected patients with lymphadenopathy at CT [8]. The study was powered to detect a difference of 25% between the ROSE and no-ROSE arms. The diagnostic yield was the primary outcome, whereas specimen’s adequacy, number of biopsy sites and complications rate of bronchoscopy were the designed secondary outcomes. Interestingly, neither diagnostic yield nor specimen’s adequacy was significantly different in the two study arms. The possibility to avoid biopsy from additional targets without loss in diagnostic yield was the most important benefit of the use of ROSE, as it was associated with a significant reduction of the complication rate of bronchoscopy. Patients with multiple targets available for biopsy (e.g., lung nodule and lymphadenopathy), in fact, were more likely to terminate bronchoscopy after the initial hilar/mediastinal TBNA, therefore avoiding the sampling of concomitant parenchymal lesion. This aspect related to the use of ROSE is certainly of great potential importance, as forceps biopsy, and in particular transbronchial biopsy, raises the risk of major complications several-fold.

Yarmus et al. performed a similar, yet smaller randomized trial of ROSE for diagnosis of lymphadenopathy with conventional TBNA [9]. Even though the study was statistically powered to detect a large, and probably unrealistic difference in diagnostic yield between the ROSE and no-ROSE arms (71 vs 37%), the results were similar to those obtained by Trisolini et al. Diagnostic yield and specimen’s adequacy were not statistically different between the study arms, whereas a trend toward reduction of transbronchial lung biopsies was seen in the ROSE arm.

Very recently, Oki et al. published the first randomized trial aimed at evaluating the role of ROSE during endobronchial ultrasound-guided TBNA (EBUS-TBNA) for diagnosis of lymph node metastasis from lung cancer [10]. Keeping into account the disappointing results of the randomized studies of ROSE in the setting of conventional TBNA [8,9], as well as the very high success rate of EBUS-TBNA regardless of the use of ROSE [11], the authors calculated that thousands of patients in each arm would be required to demonstrate the superiority of EBUS-TBNA plus ROSE vs EBUS-TBNA in terms of diagnostic yield. As a consequence, they designated the need for additional procedures (EBUS-TBNA for lesions other than the main target lesion and/or transbronchial biopsies of lung nodules) as primary outcome. Indeed, patients enrolled in this trial were significantly more likely to avoid additional diagnostic procedures if they were in the ROSE arm (11 vs 57%, p < 0.001). Diagnostic sensitivity and accuracy (secondary outcome measures) were not different in the two arms, but the study was clearly underpowered to detect differences regarding this outcome measures, as previously explained.

In conclusion, the randomized trials demonstrated that the true clinical advantage associated with the use of ROSE is the possibility to avoid biopsy from additional targets without loss in diagnostic yield, which in turn can reduce the complication rate of bronchoscopy. They also suggest that ROSE influences neither diagnostic yield nor specimen’s adequacy of conventional and EBUS-TBNA specimens. This is certainly not surprising if one considers the high adequacy rate and diagnostic yield of these procedures, especially that of EBUS-TBNA, regardless of ROSE. A recent meta-analysis on the influence of ROSE on fine-needle aspiration adequacy across all anatomic body locations showed, in fact, that the improvement associated with the use of ROSE was very low when on-site review was coupled with diagnostic procedures which have, per se, high success rates [12].

The experience of the clinical research on ROSE during bronchoscopy demonstrates the importance of confirming with high-quality, randomized trials the hypotheses that can arise from observational studies.

An urgent call for new studies on ROSE during broncoscopy!

There are several aspects related to the use of ROSE, some of which are briefly cited below, that need further evaluation.

The ability to modify in real-time the sampling plan based on the results of on-site review is one such aspect. It is not infrequent, when using conventional or EBUS-TBNA in the setting of mediastinal diagnosis/staging of advanced lung cancer, to retrieve an amount of malignant cells that is enough for an accurate pathologic typing of the malignancy, but is not sufficient for a full mutation analysis (EGFR, K-ras, ALK) to be performed. In cases like this, the diagnostic procedure would be ‘technically’ diagnostic, but a repeat bronchoscopy would still be required to establish the most appropriate treatment in the era of targeted oncological therapy. Studies aimed at evaluating whether or not ROSE could improve the rate of lung cancer patients, submitted to EBUS-TBNA, who get both pathologic diagnosis and full mutation analysis with the first bronchoscopy would be very important.

Most of the studies regarding ROSE during bronchoscopy were performed in expert centers where both the interventional pulmonologists and the pathologists had large experience with the diagnostic technique being used (conventional or EBUS-TBNA). It would be extremely useful to verify whether or not the results obtained in these studies would be reproduced in less expert settings, as well as to verify to which extent ROSE can be useful in the setting of education/training of interventional pulmonology fellows.

Even studies that proved the cost–effectiveness of ROSE demonstrated that it is time-consuming for the pathologist [13]. Preliminary studies recently assessed options to overcome this problem. Khurana et al. evaluated the possible role of telecytopathology, and suggested that it is as accurate as on-site review, while sparing the involved pathologist the time to and back from the endoscopy suite [14]. Bonifazi et al. demonstrated that a pulmonologist, after a short yet intensive training period, can perform ROSE of conventional TBNA from lymphadenopathy with an 81% interobserver agreement with the pathologist taken as gold standard [15].

Conclusion

ROSE, when used for the bronchoscopic diagnosis of lymphadenopathy, allows avoiding additional sampling without loss in diagnostic yield and reduces the complication rate of the procedure. Several aspects of great potential clinical, educational and organizational importance regarding ROSE during bronchoscopy require further evaluation. These aspects include the possibility to modify the sampling strategy in real-time, the value for education/training of pulmonary fellows, the role for diagnosis of peripheral lesions, the cost–effectiveness, the implementation of options aimed at reducing the ‘timewastage’ of pathologists performing ROSE and the ability of pulmonologists to perform ROSE for assessment of specimen’s adequacy themselves.

Financial & competing interests disclosure

The author have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.