Pre-symptomatic (Baseline) computed tomography predicts invasive pulmonary aspergillosis in high-risk adult haemato-oncology patients.

Abstract

Invasive pulmonary aspergillosis (IPA) adversely influences outcome in patients undergoing haematopoietic stem cell transplantation (HSCT) (Marr et al, 2002; Kontoyiannis et al, 2010). Early diagnosis is key but the majority of diagnoses of proven IPA are made at autopsy (Chamilos et al, 2006). Biopsy confirmation is desirable but seldom undertaken in clinical practice. In the absence of histopathological proof, haematologists frequently rely on clinical suspicion and findings on computed tomography (CT). The typical CT findings of IPA include the presence of nodules (with or without a characteristic “halo” of ground-glass opacification) and peripheral foci of consolidation (Greene et al, 2007). Air-crescents and cavitation are late features which tend to coincide with neutrophil recovery (Caillot et al, 2001). The revised European Organization for Research and Treatment of Cancer and the Mycoses Study group (EORTC/MSG) definitions of invasive fungal disease have been welcomed and emphasize the role of imaging in diagnosing IPA (De Pauw et al, 2008). However, these definitions were designed for clinical trials. CT scanning is a key component of these definitions but little is known about the prevalence of pre-symptomatic or baseline CT (BCT) abnormalities and their value in IPA diagnosis. The present study prospectively investigates the prevalence, nature and utility of CT abnormalities among asymptomatic patients prior to initiation of chemotherapy, immunosuppressive therapy (IST) or HSCT, and is a secondary analysis of a previously reported study (Ceesay et al, 2015). Adult haematology patients admitted for intensive chemotherapy or HSCT were recruited within 24 h of hospital admission and volumetric high-resolution BCT was performed within 72 h, but before initiating chemotherapy. Patients were excluded if they were receiving antifungal treatment or secondary prophylaxis (Data SI). Two hundred and three patients were recruited and followed-up for a median of 556 (range 12–730) days. BCT was performed in all but a single patient in whom planned allogeneic HSCT was cancelled (Fig. 1A). All CT examinations were independently reviewed by two experienced thoracic radiologists. Individual CT signs were defined according to the Fleischner Society terminology (Hansell et al, 2008) and abnormalities were categorised as EORTC/MSG-recognised or non-EORTC/MSG-recognised (Data SII). Discrepant observations were resolved by agreement. The predictive role of BCT signs for the diagnosis of IPA was tested using multivariate Cox proportional-hazards model using the bootstrap strategy for variable selection (Data SIII). Inter-observer agreement was moderate or good (Data SIV). Patients with normal and abnormal BCT had similar demographics and underlying haematological diagnoses (Table 1). Among the 198 evaluable patients BCT was abnormal in 72 (36%) patients: 19 (10%) had EORTC/MSG-recognised signs and 53 (27%) had non-EORTC/MSG-recognised signs. The predominant EORTC/MSG-recognised signs were nodules (Table SI). Among the 19 patients with EORTC/MSG-recognised signs three developed proven mould not otherwise specified and six had probable IPA (Table SII), a cumulative incidence of proven/probable IPA of 47·4% (9/19). This group predominantly comprised of patients with myelodysplastic syndrome/acute myeloid leukaemia admitted for allogeneic HSCT and received itraconazole as primary prophylaxis. All except one patient had previously received 1–3 cycles of chemotherapy before recruitment. The median time from BCT to development of proven/probable IPA was 14 (range 7–62) days. Consolidation was the commonest non-EORTC/MSG sign among the 53 patients with these abnormalities (Table SI). Consolidation can occur in IPA (De Pauw et al, 2008) but is non-specific (Greene et al, 2007). Nine of these patients developed proven/probable IPA (Table SII), a cumulative incidence of proven/probable IPA of 17% (9/53). On multivariate analysis, the final model showed that patients with an abnormal BCT had a significantly higher risk of IPA [adjusted hazard ratio (HR), 2·52; 95% confidence intervals (CI), 1·27–5·03; Fig. 1B]. The increased risk of IPA was primarily confined to those with EORTC/MSG-recognised signs (adjusted HR, 4·67; 95% CI, 2·04–10·75) and was less evident in patients with non-EORTC/MSG-recognised signs (adjusted HR, 1·30; 95% CI, 0·87–2·15) (Fig. 1C). Early CT scanning has been advocated for the diagnosis of IPA (Caillot et al, 2001), but, to date, there are no robust data on the prevalence and predictive value of pre-symptomatic BCT. In this study, we found both EORTC/MSG and non-EORTC/MSG-recognised signs on BCT in 72/198 (36%) patients. These abnormalities were significantly predictive of development of IPA during chemotherapy/transplantation in patients with EORTC/MSG-recognised signs. Our patients with baseline abnormalities were at the pre-symptomatic stage with no clinical evidence of sepsis. Moreover, they had similar demographic and clinical characteristics to those with normal scans and C-reactive protein and neutrophil counts were not significantly different (Table 1). Fever-driven management strategies like ours depend on failure to respond to broad-spectrum antibacterial therapy for ≥96 h to trigger the diagnostic work-up for IPA. Although fever is nonspecific for IPA (De Pauw et al, 2008), persistent fever despite broad-spectrum antibacterial treatment remains the most frequent reason for starting empirical antifungal therapy, an approach still endorsed by consensus guidelines (Patterson et al, 2016). Patients who are otherwise well with no fever but abnormal CT findings present clinical challenges for the treating physician. The four patients with EORTC/MSG-recognised signs who were excluded due to fever within 24-h of BCT scan raise an intriguing possibility of occult IPA. None were on any regular antipyretics or corticosteroids, which may mask inflammatory response. Positive galactomannan or (1-3)-β-d-glucan before the onset of fever is recognised (Maertens et al, 2005) but abnormal CT preceding other clinical signs of infection and mycological evidence of IPA is not. The high proportion of baseline abnormalities found in this study may seem surprising but it is worth noting that this is a heavily pre-treated cohort (Table 1) that had received previous chemotherapy, often outside the tertiary centre. The baseline lesions may therefore, represent residual IPA that might not have been recognised locally or recognised but not reported to the tertiary centre. That two experienced observers with interest in thoracic imaging independently evaluated BCT scans is a strong feature of this study. Our results indicate that inter-observer agreement for EORTC/MSG-recognised signs was good. However, the findings also underscore the importance of radiologists' experience for the reading of thoracic CT studies in neutropenic patients with suspected IPA. In summary, our data has shown that abnormalities are present on BCT in over one-third of pre-symptomatic patients in this cohort. More importantly, the presence of EORTC/MSG-recognised signs on BCT is predictive of future IPA. BCT may be an important component in the management of haematology patients at high-risk of IPA. MMC and SRD designed the study and wrote the manuscript. MMC collected and analysed the data. SRD and JC performed independent CT review. AP designed the study and critically reviewed manuscript. LB, MS, JW and GJM contributed to the study design and reviewed manuscript. We are grateful to J Peter Donnelly for critical review of the manuscript and to Abdel Douiri for statistical advice. Table SI. Radiological signs in patients with baseline CT abnormalities. Table SII. Final invasive fungal disease diagnoses according to baseline CT appearances. Figure S1. (A) Baseline and (B) diagnostic CT scans of a study patient. Data SI. Clinical protocol. Data SII. Imaging studies. Data SIII. Statistical analysis. Data SIV. Inter-observer variability & Diagnostic lesions with respect to baseline abnormalities. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.