Article Text
PDF
Abstract
Background While therapeutic bronchoscopy has been used to treat malignant central (CAO) airway obstruction for >25 years, there are no studies quantifying the impact of therapeutic bronchoscopy on long-term quality-adjusted survival.
Methods We conducted a prospective observational study of consecutive patients undergoing therapeutic bronchoscopy for CAO. Patients had follow-up at 1 week and monthly thereafter until death. Outcomes included technical success (ie, relief of anatomic obstruction), dyspnoea, health-related quality of life (HRQOL) and quality-adjusted survival.
Results Therapeutic bronchoscopy was performed on 102 patients with malignant CAO. Partial or complete technical success was achieved in 90% of patients. At 7 days postbronchoscopy, dyspnoea improved (mean ∆Borg-day-7=−1.8, 95% CI −2.2 to −1.3, p<0.0001) and HRQOL improved (median prebronchoscopy 0.618 utiles, 25%–75% IQR 0.569 to 0.699, mean ∆utility-day-7+0.047 utiles, 95% CI +0.023 to 0.071, p=0.0002). Improvements in dyspnoea and HRQOL were maintained long-term. Compared with the prebronchoscopy baseline, HRQOL per day of life postbronchoscopy improved (mean ∆utility-long-term+0.036 utiles, 95% CI +0.014 to 0.057, p=0.002). Median quality-adjusted survival was 109 quality-adjusted life-days (QALDs) (95% CI 74 to 201 QALDs). Factors associated with longer quality-adjusted survival included better functional status, treatment-naïve tumour, endobronchial disease, less dyspnoea, shorter time from diagnosis to bronchoscopy, absence of cardiac disease, bronchoscopic dilation and receiving chemotherapy.
Conclusions Therapeutic bronchoscopy improves HRQOL as compared with baseline, resulting in approximately a 5.8% improvement in HRQOL per day of life. The risk-benefit profile in these carefully selected patients was very favourable.
Trial registration number Results; NCT03326570.
- bronchoscopy
- lung cancer
Statistics from Altmetric.com
Key messages
What is the key question?
The long-term benefit of therapeutic bronchoscopy for malignant central airway obstruction has never been quantified in terms of quality-adjusted life-years (QALYs).
Knowing the magnitude of the long-term benefits and what factors influence this are essential for patient management since it is central to risk-benefit assessments and determination of cost-effectiveness.
What is the bottom line?
Therapeutic bronchoscopy results in a 5.8% improvement in health-related quality of life per day of life, which is roughly 1.5 times the benefit seen with treatment of malignant pleural effusions over the long-term.
Why read on?
This manuscript quantifies the magnitude and variability of the impact of therapeutic bronchoscopy on patients with malignant central airway obstruction as measured in QALYs and identifies the factors that influence this; this information can help inform proper patient selection.
Malignant central airway obstruction (CAO) occurs when tumours occlude the trachea, the main bronchi, the bronchus intermedius or a lobar bronchus. The clinical consequences of malignant CAO include dyspnoea and decreased health-related quality of life (HRQOL). Prior studies have demonstrated that therapeutic bronchoscopy for malignant CAO is associated with good short-term outcomes, as measured by relief of dyspnoea and improvements in HRQOL.1–14
However, while there is a single study looking at short-term changes in utility following therapeutic bronchoscopy,3 and there are studies of quality-adjusted survival in other areas of interventional pulmonology such as malignant pleural effusions15 and bronchoscopic treatments of emphysema,16 there are to our knowledge no long-term studies of quality-adjusted survival beyond 30 days in patient treated with therapeutic bronchoscopy for malignant CAO. Quality-adjusted survival, usually expressed as quality-adjusted life-years (QALYs), is essential for cost-effectiveness analysis and has been recognised as the most important indicator of the effectiveness of healthcare interventions, as reflected by the guidelines of the National Institute of Health and Clinical Excellence, the US Public Health Service and the Agency for Healthcare Research and Quality.17–21
The goal of this study was to quantify long-term quality-adjusted survival following therapeutic bronchoscopy for malignant CAO. Our secondary objective was to identify factors associated with quality-adjusted survival. Our hypothesis was that therapeutic bronchoscopy would result in sustained significant long-term improvements in utility and dyspnoea as compared with baseline.
Methods
This was a single-centre prospective observational study of consecutive patients undergoing therapeutic bronchoscopy for malignant central CAO from 1 September 2011 to 17 June 2014. Informed consent was obtained from all patients. Data were entered using definitions and quality control checks as previously described for the AQuIRE database but the data were captured using REDCap electronic data capture tools hosted at the University of Texas MD Anderson Cancer Center.22–25
Patients undergoing therapeutic flexible or rigid bronchoscopy for malignant CAO were included. CAO was defined as occlusion ≥50% of the trachea, mainstem bronchi, bronchus intermedius or lobar bronchus. Patients had systematic follow-up at 1 week, 1 month and monthly thereafter and were followed until death. Outcomes included: technical success (ie, relief of anatomic obstruction), dyspnoea, HRQOL/quality-adjusted survival and survival time.
Technical success was defined as reopening the airway lumen to >50% of normal. If all stenotic airways were reopened this was considered complete success. If one or more occluded airways were reopened but any segment remained occluded this was classified as partial success. If no airways were reopened this was classified as failure. Dyspnoea was measured using the Borg score. HRQOL was measured using the SF-6D as previously described.3 26 The SF-6D provides a means to estimate a preference-based single index measure for health using general population data (see online supplement). Single index measures such as the SF-6D generate measures of utility for particular health states that range from 0 to 1 utiles. Zero utiles represents death and one is perfect health. Quality-adjusted survival can be thought of as the area under the curve with utility on the vertical axis and time on the horizontal axis. We express quality-adjusted survival in quality-adjusted life-days (QALDs). Since the focus was on long-term outcomes and quality-adjusted survival, we report complications in the online supplement using the AQuIRE registry definitions.27
Supplementary file 1
Statistical analysis
To assess the short-term impact of therapeutic bronchoscopy on HRQOL and dyspnoea, we used paired t-tests to compare baseline versus day 7 utilities (derived from SF-6D) and Borg scores. Day 7 measured from the study entry bronchoscopy. We used a generalised linear model to identify variables associated with these outcomes. The dependent variable was the pairwise difference between day 7 postprocedure scores and baseline values (∆utility-day-7 for HRQOL and ∆Borg-day-7 for dyspnoea). We conducted a responder analysis for these outcomes as well (see online supplement).
The primary outcome was quality-adjusted survival and HRQOL. Long-term quality-adjusted survival was expressed as QALDs. We used a previously described method to calculate QALDs for each patient (see online supplement).15 The last observation carried forward (LOCF) method was specified a priori for imputation of missing utility measures. A secondary analysis used subsequent observations to impute missing utility measures (ie, carrying backward). There was no significant difference between results, so we report LOCF results. The Kaplan-Meier product-limit method was used to estimate median QALDs. Patients lost to follow-up or alive at study completion were censored. Cox proportional hazards models were fit to determine the association of variables with quality-adjusted survival (see online supplement). Time-varying covariates were not used for this outcome.
To compare HRQOL over the long-term versus baseline, we defined each patient’s mean HRQOL per day of life postbronchoscopy as that patient’s QALD/survival time. This represents the patient’s time-weighted mean utility experienced postbronchoscopy—so what was that patient’s utility on an average day. We calculated the long-term change in utility (∆utility-long-term) by subtracting each patient’s baseline utility from their mean HRQOL per day of life postbronchoscopy. We recognise that this is not a perfect measure of therapeutic impact, since patients who develop new clinical problems postbronchoscopy that are non-lethal could live a long time with lower utility, which would tend to drag down ∆utility long-term. We therefore also plotted utility over time stratified by survival duration (<14, 15–60, 61–180, 181+ days).
For time to death, we used extended Cox models with time-dependent covariates to represent repeated measures such as utility and Borg scores as well as clinical events such as lower respiratory tract infections (see online supplement).
For multivariate models, all variables that had a p value <0.20 on univariate analyses were considered as candidate variables. In the absence of data on important predictors, we used backward selection to retain only variables whose p values were <0.05. P values <0.05 were considered to be significant; all tests were two-sided. Predictive models used only information available preprocedure, explanatory models used all available information.
Sample size was based on HRQOL. Assuming a baseline utility of 0.690 utiles and an improvement in utility to 0.723 utiles following therapeutic bronchoscopy with an SD of the difference of 0.1, then to have 90% power to demonstrate a difference of +0.033 utiles with an alpha of 0.05 requires 99 patients.3 We estimate 10% of patients that consented would prove to have lesions that were not amenable to bronchoscopic intervention, so enrolment was capped at 110 patients. All statistical analyses were performed using STATA V.14.2 (StataCorp, College Station, Texas, USA).
Results
Patients
A total of 102 patients had interventions. Baseline characteristics are shown in table 1. Partial or complete technical success was achieved in 90% of patients (see online supplement). Of the 102 patients, 75 died during the study period, 8 were eventually lost to follow-up (censored) and 17 were alive at study completion (censored). There were 915 scheduled follow-up measurements for dyspnoea (Borg) and utility (SF-6D) (average 9 measurements of each per patient). Follow-up data were missing in 159 (17%) time points. Of the 102 patients, 61 had one or more missing follow-up data points.
Patient and procedural characteristics
Impact on dyspnoea
Of the 102 patients, 96 had baseline Borg scores, one was on a ventilator, two were emergent, three were missed. The analysis for dyspnoea was based on these 96 patients. Therapeutic bronchoscopy resulted in a significant improvement in dyspnoea at 7 days post the study entry bronchoscopy (mean ∆Borg-day-7=−1.8, 95% CI −2.2 to −1.3, p<0.0001). Note that lower scores indicate less dyspnoea, so if ∆Borg-day-7 is negative dyspnoea decreased. In the multivariate predictive model, higher baseline Borg was associated with greater improvements in dyspnoea (table 2). A one unit increase in baseline Borg was associated with a 0.53 unit decrease in ∆Borg-day-7 (p<0.001).
Factors associated with change in dyspnoea (Borg score day-7 minus Borg baseline)*
In a responder analysis, 58 of 96 patients (60%) had a clinically significant improvement in dyspnoea, 34 (35%) stayed the same and 4 (4%) worsened. In the predictive multivariate responder analysis, only higher baseline Borg (p<0.001) was associated with a clinically significant improvement in dyspnoea (see online supplement e-table 1). In the explanatory multivariate responder analysis, higher baseline Borg, receiving prior cancer treatment and use of the microdebrider were associated with clinically significant improvement in dyspnoea.
This improvement in dyspnoea was sustained over time for most patients (figure 1A). Patients who died within 14 days of bronchoscopy (n=4) did not experience dyspnoea relief. For most patients, dyspnoea initially dramatically improved and this was maintained for most of their lives (figure 1B). In some patients, towards the end of life, there was a recurrence of dyspnoea.
Changes in dyspnoea over time for the entire cohort. (A) Boxes represent medians and IQR of Borg scores among patients alive at that time point. Note that as patients die, they no longer contribute, so median dyspnoea of those alive appears to increase (ie, sick patients with dyspnoea die and no longer contribute data as we move to the right). To adjust for this, we can stratify by survivor group (bottom panel). (B) Median Borg scores over time by survivor group.
Impact on HRQOL
HRQOL improved from prebronchoscopy to postbronchoscopy (median prebronchoscopy 0.618 utiles, 25%–75% IQR 0.569 to 0.699; mean ∆utility-day-7+0.047 utiles, 95% CI +0.023–0.071, p=0.0002). In multivariate predictive analysis, higher baseline Borg score, not having COPD and having no prior therapeutic bronchoscopies were associated with higher ∆utility-day-7 (ie, more benefit) (table 3). In explanatory multivariate analysis, greater improvements in dyspnoea (ie, more negative ∆Borg-day-7), not having COPD and having no prior therapeutic bronchoscopies were associated with higher ∆utility-day-7. A one unit increase in ∆Borg-day-7 (ie, more dyspnoea) resulted in a decrease in ∆utility-day-7 of 0.021 utiles (p<0.001).
Factors associated with change in utility (utility at day-7 minus baseline)*
In a responder analysis, 51 of the 102 patients (50%) had a clinically significant improvement in utility at 7 days, 33 (32%) stayed the same and 18 (18%) worsened (see online supplement e-table 2). In a multivariate predictive analysis, only having fewer lung segments occluded (p=0.05) was associated with having a clinically significant improvement in dyspnoea. In an explanatory multivariate analysis, older age (p=0.03), greater improvements in dyspnoea (p=0.01) and not having COPD (p=0.03) were associated with having a clinically significant improvement in dyspnoea.
Median quality-adjusted survival was 109 QALDs (95% CI 74 to 201 QALDs). In a predictive multivariate Cox model (table 4), factors associated with longer quality-adjusted survival included better functional status, treatment-naïve tumour, endobronchial disease and less dyspnoea at baseline. In an explanatory multivariate Cox model, these same factors were associated with longer quality-adjusted survival. In addition, shorter time from cancer diagnosis to therapeutic bronchoscopy, absence of cardiac disease, bronchoscopic dilation and receiving chemotherapy postintervention were associated with longer quality-adjusted survival.
Factors associated with quality-adjusted survival time
Given that therapeutic bronchoscopy is primarily palliative, long-term impact on HRQOL per day of life is an important measure. Median postbronchoscopy HRQOL per day of life was 0.655 utiles per day (25%–75% IQR 0.590 to 0.741). Compared with the prebronchoscopy baseline, HRQOL per day of life postbronchoscopy improved (mean ∆utility-long-term+0.036 utiles, 95% CI +0.014 to 0.057, p=0.002). Evaluating utility changes over time stratified by survivor group (figure 2), we can see that for most survivor groups the gains in utility postbronchoscopy were maintained for some time, but as patients near the end of their lives utility starts to decrease, which in turn drags down HRQOL per day of life. Patients that died within 14 days (n=4) were the exception; these patients had very low functional status to begin with and did not experience meaningful gains in utility. But for the majority, the immediate gains in utility postbronchoscopy were maintained for a significant amount of time.
Changes in median utility over time by survivor group.
In multivariate predictive analysis (table 5), factors associated with greater ∆utility-long-term included higher baseline Borg score and having fewer lung segments occluded. In a multivariate explanatory analysis, higher baseline Borg score, having fewer lung segments occluded and ∆utility-day-7 were associated with a higher ∆utility-long-term. In univariate explanatory analysis, 59% of the observed variability in ∆utility-long-term was explained by the linear relationship between ∆utility-long-term and ∆utility-day-7 (ie, adjusted R2=0.59).
Factors associated with mean benefit per day of life measured in utility
Survival
Median survival was 179 days (95% CI 107 to 271). The factors associated with survival in the predictive Cox model (table 6) were the same as those associated with quality-adjusted survival. When we limited the explanatory model to the same covariates as used for quality-adjusted survival, the factors identified were again largely the same as with quality-adjusted survival. The one difference was that cardiovascular disease was associated with decreased survival but not associated with decreased quality-adjusted survival.
Factors associated with time to death, predictive Cox model
We then built extended Cox models to assess the impact of variables that vary over time on survival, such as lower respiratory tract infections, dyspnoea, HRQOL and whether or not a stent was in place (table 7). Taking into consideration these time-varying covariates, we found better baseline performance status as measured by Zubrod, higher HRQOL (utility) and receiving chemotherapy following therapeutic bronchoscopy were associated with improved survival. Conversely, lower respiratory tract infections were associated with increased mortality risk.
Factors associated with time to death, multivariate explanatory extended Cox models
Discussion
Therapeutic bronchoscopy for malignant CAO is essentially a palliative intervention. While therapeutic bronchoscopy may extend life in some patients with malignant CAO, such as those on mechanical ventilation due to CAO, for the majority of patients the goal is relief of dyspnoea and improvement in HRQOL. In deciding whether therapeutic bronchoscopy is likely to benefit a given patient, it is therefore necessary to consider the probability of technical success, the magnitude of the impact that technical success is likely to have on dyspnoea, whether or not that benefit is likely to be maintained over time and how that in turn will impact long-term HRQOL and quality-adjusted survival. The potential benefits in long-term HRQOL must then be weighed against the risks of intervention. To the best of our knowledge, this study is the first to quantify long-term quality-adjusted survival following therapeutic bronchoscopy for malignant CAO and to identify the factors associated with it. Median quality-adjusted survival was 109 QALDs (95% CI 74 to 201 QALDs). Factors associated with improved quality-adjusted survival included better baseline performance status, less dyspnoea at baseline, having a treatment-naïve tumour, endobronchial disease, shorter time from cancer diagnosis to intervention and receiving follow-up chemotherapy. Therapeutic bronchoscopy resulted in immediate improvements in dyspnoea and utility. Importantly, these improvements over baseline were maintained long-term such that HRQOL per day of life improved compared with prebronchoscopy levels.
Previous investigators have demonstrated improvements in dyspnoea and quality of life.1–14 However, these prior studies cannot fully inform clinical decision making because the benefits of intervention must be weighed against the risks but the trade-offs involved have different units.3 For example, how much risk does technical success warrant, when technical success is measured in relief of luminal narrowing and risk might be measured as pneumothorax or death? Dyspnoea is a more relevant patient-centred outcome, but how much risk of death does a clinically significant one unit change in Borg score warrant?
So benefit needs to be quantified in units that allow physicians to compare the benefits with the risks involved. This requires use of generic single-index measures of HRQOL such as the SF-6D that can be used to generate utilities (see the online supplement). The AQuIRE registry study of therapeutic bronchoscopy was the first to do this for short-term outcomes, reporting a mean baseline HRQOL of 0.65 utiles, a ∆utility-day-30 of +0.023 utiles and a ∆Borg-day-30 of −0.9.3 In AQuIRE, for every one unit increase in ∆Borg-day-30, ∆utility-day-30 changed by −0.020 utiles (ie, more dyspnoea, lower HRQOL). In this study, baseline HRQOL was 0.618 utiles, ∆utility-day-7 was +0.047 utiles, ∆Borg-day-7 was −1.8 and for every one unit increase in ∆Borg-day-7, ∆utility-day-7 changed by −0.021 utiles. So in terms of the short-term impact of therapeutic bronchoscopy on dyspnoea and HRQOL, our findings are similar to AQuIRE.3 The present findings are also consistent with those in the literature, which demonstrate that for non-small cell lung cancer, utilities for health states for cancer-related outcome range from 0.70 (responding to therapy) to 0.47 (progressive disease) to 0.33 (end of life).28
The effect size observed in this study for short-term outcomes was somewhat greater for change in utility and dyspnoea than in AQuIRE, but the relationship between ∆Borg and ∆utility was similar. Part of the difference between studies may be due to the use of day-7 outcomes versus day-30 outcomes. We chose day-7 outcomes as a measure of the immediate impact of therapeutic bronchoscopy rather than the previously published day-30 based on what we learnt from AQuIRE. In AQuIRE, we found that many patients received additional treatments (eg, chemotherapy) within 30 days, which confounds the assessment of the impact of bronchoscopy on utility, since changes could be due to factors other than bronchoscopy. So the increased effect size observed in this study for relatively short-term outcomes is likely a combination of differences in timing of measurement, centre-level variability and patient selection.
However, while knowledge of the determinants of short-term outcomes following therapeutic bronchoscopy for malignant central airway obstruction is useful, it is not sufficient. We need to be able to quantify long-term benefits as well. One of the key questions unanswered by AQuIRE was whether short-term improvements in dyspnoea and utility are maintained over time and how much do they impact quality-adjusted survival. This study adds to the existing body of knowledge by quantifying the long-term quality-adjusted survival of patients undergoing therapeutic bronchoscopy. Our findings suggest that long-term HRQOL per day of life following therapeutic bronchoscopy was significantly better than the prebronchoscopy baseline (∆utility-long-term+0.036 utiles per day of life).
This change in long-term HRQOL represents the aggregate of a multimodality approach, not just bronchoscopy (see online supplement e-table 3). However, much of the observed improvement is probably due to a sustained reduction in dyspnoea. This reduction in dyspnoea is initially achieved with bronchoscopy and is subsequently maintained over time. This is supported by our finding that the short-term impact of therapeutic bronchoscopy on dyspnoea (∆Borg-day-7) was strongly associated with short-term HRQOL changes (∆utility-day-7). Short-term term HRQOL changes were also strongly associated with long-term changes in HRQOL as measured by ∆utility-long-term (table 5, adjusted R2=0.59 for ∆utility-day-7 impact on ∆utility-long-term). Since changes in utility from prebronchoscopy to 7 days postbronchoscopy are most likely due to the effects of therapeutic bronchoscopy rather than other treatments (eg, chemotherapy), the finding that ∆utility-day-7 explains 59% of the variance in long-term HRQOL suggests that the short-term benefits of therapeutic bronchoscopy are maintained over time and help drive improvements in long-term HRQOL.
Comparing therapeutic bronchoscopy as part of a multimodality approach with radiation therapy (RT) alone for malignant CAO is difficult because improvements in RT dosing and delivery limit the value of historical comparisons. In addition, there is very little data on RT for malignant CAO. However, historical data from when therapeutic bronchoscopy was not available can provide some useful insights provided we keep these limitations in mind.4 A study from the Mayo Clinic found that laser resection for malignant CAO improved survival (40% mortality at 7 months and 72% at 1 year) as compared with historical controls that received radiation (76% mortality at 4 months and 100% mortality by 7 months).29 The Cleveland Clinic published data on Nd:YAG laser/RT versus historical controls receiving RT and found better survival with bronchoscopy when the airway obstruction was emergent (267 vs 150 days, p=0.04).30 In the same series, taking both emergent and non-emergent cases, there was no survival difference (YAG 304 days vs RT 253 days, p=0.17). A study of mechanically ventilated patients with malignant central airway obstruction found that those with successful reopening of the airway did better and received additional chemo-RT after successful extubation.31 Unfortunately, none of these studies measured quality-adjusted survival.
We believe that quantifying the impact of therapeutic bronchoscopy on dyspnoea and quality-adjusted survival can help inform clinical decision making. We found that the impact of bronchoscopic intervention is maximal early on (ie, ∆utility-day-7+0.047 utiles), but as the disease progresses and side effects set in HRQOL is impacted (∆utility-long-term+0.036 utiles per day of life). A conservative assumption regarding therapeutic bronchoscopy is that it does not extend duration of life at all, it does not prevent any postobstructive pneumonias, and its only benefit is that it improves HRQOL by alleviating dyspnoea related to CAO obstruction. The initial ∆utility-day-7 provides an upper limit estimate of the impact of therapeutic bronchoscopy on HRQOL based on these assumptions. Over the long-term, this estimate may be too high—time and disease will mitigate the beneficial effects, so ∆utility-long-term provides a more conservative assessment of benefit. Since this is not a randomised controlled trial, we cannot provide a more precise assessment of the impact of therapeutic bronchoscopy as compared with the alternative of not doing bronchoscopy. While such a randomised controlled trial would be informative, it would not be ethical, and as such we believe using ∆utility-long-term provides a practical if not perfect estimate of the effect of therapeutic bronchoscopy on HRQOL. Given these limitations, what we can say with confidence is that HRQOL per day of life postbronchoscopy is on average better than it was prebronchoscopy and this effect is maintained over time.
Because physicians do not often deal with utility, it can be difficult for clinicians to interpret how large an effect size this represents. So for clinical practice it can be helpful to put utility numbers into clinical context in order to calibrate judgement. So how much does a change of +0.036 utiles per day of life represent? First, let us remember that mean baseline utility was 0.618 utiles, so this represents a 5.8% improvement. Since we are evaluating quality-adjusted survival, this improvement in HRQOL is equivalent to living 5.8% longer if HRQOL were to remain at baseline levels.
Another way to gain clinical context is to compare this effect with the effect of more common interventions in similar populations. However, comparing the impact of interventions between very different populations using QALYs is not always valid (see online supplement).32–35 Care must be taken when choosing a comparator population and intervention. Given these limitations, the most analogous population for which there is data is probably patients with malignant pleural effusions (MPE).15 Both populations have metastatic cancer, both have similar life expectancies and in both cases the palliative intervention is aimed at dyspnoea relief. Unlike malignant CAO, MPE is a very common problem for many pulmonologists, so a comparison can provide context to the utility measures being reported. In a prospective observational study of 266 patients undergoing indwelling pleural catheter (IPC) placement for MPE, HRQOL at 1 month post-IPC increased compared with baseline by +0.023 utiles.15 So as a rough approximation, the mean change in HRQOL immediately following therapeutic bronchoscopy at day-7 is double that seen with drainage of a MPE (+0.047 vs +0.023 utiles, respectively). Over the long-term, the impact of bronchoscopy is roughly one-and-a-half times as much as the improvement seen with an IPC.
This improvement in HRQOL must be weighed against the risks. While there were no fatal procedural complications in this study, the overall rate of fatal procedural complications reported in AQuIRE was 0.5%.27 The overall risk-benefit ratio of therapeutic bronchoscopy is very favourable if we use 0.5% as our estimate of fatal procedural complications and consider ∆utility-long-term as roughly a 5.8% improvement over baseline.
Although these findings are useful, it is important to recognise the limitations of the data. This was a single-centre prospective observational study. The generalisability of the results to other centres may be limited. There is significant heterogeneity between centres in terms of methods of therapeutic bronchoscopy, complication rates and 30-day mortality.3 27 The impact of therapeutic bronchoscopy observed in this study may vary from that of other centres depending on patient referral patterns, patient selection and bronchoscopic techniques. In addition, as with any long-term longitudinal study, imputation for missing data can be a problem. We prespecified the LOCF method based on prior studies,15 but the LOCF methodology can be prone to bias under certain conditions (see online supplement).36 37 However, because we sampled so frequently (ie, monthly) the intrapatient variation between months in utility was low, such that whether we used an LOCF or a carry-backwards imputation method, this did not change the results. Note that the limitations of imputation in this study only apply to the long-term outcomes, not the day-7 outcomes.
In conclusion, this study supports and extends the findings of the AQuIRE group. The data suggest that when baseline dyspnoea is more severe there is greater potential for therapeutic bronchoscopy to have a meaningful impact on utility.3 The impact of bronchoscopic interventions, especially when done as part of multimodality treatment, are such that the improvements in dyspnoea and utility are likely to be sustained for most patients. While immediate and long-term complications are significant, given the observed rates in this trial and other large series,1–14 the risk-benefit ratio can be considered favourable, provided the malignant CAO is causing significant dyspnoea that is contributing to the patient’s poor HRQOL. Future studies using more formal decision analysis techniques will be needed to further clarify the decision thresholds and trade-offs involved. The magnitude of the change in HRQOL following therapeutic bronchoscopy is roughly equivalent to a 5.8% improvement in quality-adjusted survival.
Acknowledgments
The authors would like to thank Dr Xiudong Lei, PhD, biostatistics department for her help with power calculations and design and Dr Liang Li for his help with imputation analysis.
References
Footnotes
Contributors DEO was the principal investigator (PI) for this study and was involved in project oversight, organisation, data collection and auditing, statistical analysis and manuscript writing. PO did the statistical analysis and was involved in manuscript writing. HBG, LD, CAJ, GAE and RFC were involved in data collection and auditing, and manuscript writing. LN contributed to data collection, auditing and review.
Funding This study was funded by NIH/NCI biostatistic core (P30CA016672), biostatistics core, at the University of Texas, MD Anderson Cancer Center and Institutional Research Grant from MD Anderson Cancer Center.
Competing interests None declared.
Patient consent Obtained.
Ethics approval Institutional review board approval was obtained from The University of Texas MD Anderson IRB Committee 4.
Provenance and peer review Not commissioned; externally peer reviewed.