Routine chest x-rays in intensive care units: a systematic review and meta-analysis

Physicians often order routine daily antero-posterior chest x-rays (CXRs) for patients in intensive care units (ICUs) due to concerns about the severity of cardiopulmonary illness and complexity of medical interventions [1] and for detection of complications associated with indwelling devices, such as endotracheal tubes and central venous catheters. The frequency of complications, such as device malpositioning or pneumothoraces, has led some guidelines to recommend routine CXRs for all patients with acute cardiopulmonary problems or receiving mechanical ventilation [2]. Advantages of routine CXRs may include prompt detection and thus earlier treatment of clinically unsuspected abnormalities, documentation of disease progression or response to therapy, and educational value for trainees [3, 4]. In contrast, a restrictive strategy limits CXRs to specific clinical indications, such as a change in clinical status or following certain procedures. Arguments for adopting a restrictive approach include variable interpretation of CXRs depending on clinician and patient factors, low incidence of clinically unsuspected abnormalities, potential harm arising from unnecessary treatment of minor or false positive findings, cost, radiation exposure and adverse events arising from repositioning of the patient to obtain the CXR [5, 6].

Our objective was to systematically review the available literature evaluating the effect on clinical outcomes of abandoning routine CXRs in favor of a more restrictive approach.

We conducted our study following recommendations from the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement [7] and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) Group [8].

We identified only three trials that evaluated the effects on clinical outcomes in critically ill adults and children of a restrictive strategy of obtaining CXRs only when abnormalities are clinically suspected compared to routine daily CXRs. We found no differences in mortality when results of trials were combined, or when results of an additional six before-after observational studies were also considered. Similarly, duration of mechanical ventilation and ICU and hospital lengths of stay were similar for both CXR strategies.

Strengths of our systematic review include a comprehensive search strategy and broad inclusion criteria, with consideration of both RCTs and quasi-RCTs and before-after observational studies. Although we searched electronic biobliographic databases with no restriction on publication date, studies included in the systematic review were all published in the last 15 years and, therefore, are likely representative of current ICU practices. Our study is limited by small sample size, moderate methodological quality and heterogeneity of included studies. Only three trials evaluated the safety of a restrictive versus routine CXR strategy; the combined sample size of these trials would only have sufficient power to detect an implausibly large difference in our primary outcome. Despite the absence of statistical heterogeneity in our analyses restricted to trials, all of which excluded patients mechanically ventilated for < 48 hours, there is likely substantial remaining clinical heterogeneity related to severity of illness and surgical status, for example.

Several studies have evaluated a restrictive CXR strategy with hypotheses that costs and radiation exposure would be reduced. The restrictive strategy can substantially reduce the number of CXRs and thus may reduce costs [18, 22] but the anticipated cost-savings have varied greatly [29, 31, 42]. With respect to radiation exposure, levels measured at ICU nursing stations have been shown to be below the maximum permissible for non-occupational workers [64, 65]. Although exposure to patients from a single CXR is felt to pose minimum risk, the cumulative exposure is thought to be significant due to the high frequency of this test [66]. In contrast, the main short-term risk of a restrictive CXR strategy in critically ill patients is the potential for a missed or delayed diagnosis, leading to delayed treatment and thus additional morbidity and mortality. We were unable to estimate the number of missed diagnoses or delays in management of complications in the restrictive CXR group because only two studies [15, 22] had a surveillance strategy for these patients.

The major design challenge for studies of different CXR ordering strategies is to overcome ascertainment bias, which may lead to more complications (including clinically non-apparent complications) being detected in the routine CXR group if surveillance CXRs are not obtained in the restrictive group. Complications may thus be underestimated in the restrictive CXR group. These study challenges can be overcome by obtaining surveillance CXRs (results blinded to clinicians) in the restrictive group to determine the delay between radiographically apparent and clinically detected complications. Other design challenges include ensuring blinding of those adjudicating radiological and clinical outcomes subject to ascertainment bias (for example, pneumonia) and standardization of definitions of complications. Furthermore, establishing non-inferiority for clinical endpoints in a RCT comparing CXR strategies requires a large sample size and modifications to the usual approach to hypothesis-testing [67]. An alternative approach would be to design a non-inferiority trial powered on the surrogate outcome of delayed or missed diagnoses. The heterogeneity of patients studied to date underscores the importance of a non-inferiority trial that tests the hypothesis that the restrictive CXR strategy does not increase harm and that is adequately powered to examine pre-specified subgroup effects. Finally, even if trials demonstrate the non-inferiority of the restrictive CXR strategy, ICUs implementing this strategy should have immediate access to clinicians who can reliably identify patients requiring urgent CXRs, interpret images that are promptly obtained and processed, and act rapidly on any findings requiring intervention [68].

A recent study using a web-based Delphi technique explored common indications for CXRs in critically ill patients, as determined by 82 French intensivists [69]. Consensus was reached that CXRs should be considered routinely after certain procedures (for example, insertion of endotracheal tube, subclavian or internal jugular central venous catheter, pulmonary artery catheter, temporary transvenous pacing lead, chest tube) and during invasive mechanical ventilation for acute respiratory distress syndrome or status asthmaticus. There was also consensus that CXRs were unnecessary in patients with an existing nasogastric tube for enteral feeding or with an existing central venous catheter previously documented to be correctly positioned in the superior vena cava. However, there was no consensus on the need for routine CXRs in intubated patients. The clinical impact and safety of implementing these consensus recommendations have not been evaluated, but the findings of our review suggest no glaring safety concerns with a restrictive CXR strategy in critically ill patients while underscoring the need for additional large randomized trials to establish definitively the non-inferiority of this approach.

Pooling the results of three trials alone or combined with six before-after observational studies that evaluated the safety of abandoning routine CXRs in favor of a more restrictive approach did not detect harm. However, given the small overall sample size and infrequent systematic evaluation of missed findings in the restrictive strategy group, the safety of abandoning routine CXRs in critically ill patients remains uncertain and mandates further investigation.