Advantages of a pooled analysis is that it makes use of data from multiple clinical trials and allows one to assess more generalized and consistent relationships across trials rather than individual trials. For the Uniscale analysis, the large patient sample (n = 355) consisted of a homogeneous population of advanced NSCLC from five prospective NCCTG clinical trials with sufficient follow-up allowing for an adequately powered analysis.
The main finding of the present analysis is that pretreatment patient self-reported QOL, measured by Uniscale, is an independent prognostic factor for OS and this prognostic association remained significant in the presence of other factors including PS, age, gender, BMI, Hgb, PLT, and ANC in patients with advanced NSCLC. Univariately, the hazard of death of patients with Uniscale lower than the CDS (≤50) is twice as high as that of patients with Uniscale >50. After adjusting for age, gender, PS, BMI, Hgb, PLT, and ANC, the strength of this association remained similar. This study showed no significant prognostic association between the total scores of multi-item indices (LCSS and FACT-L) and OS in patients with advanced NSCLC.
There are some limitations to this analysis. First, the phase III first-line maintenance therapy trial, 972451, contributed 47% of patients included in the Uniscale analysis. The median survivals are different between first-line and first-line maintenance treatment, thus the strength of the prognostic impact of baseline Uniscale might need further elaboration. Second, two trials enrolling elderly patients (i.e., age ≥65 years) were included. However, stratifying by trial and including age as a continuous covariate in the multivariate models accounted for some, if not all, of these heterogeneity issues. Third, the retrospective nature of this analysis made the exploration of all factors that might have a potential impact on patient survival not possible as they were not collected. For example, comorbidity score, which captures pre-existing comorbidity conditions and has been indicated to have a significant impact on survival,33
could not be explored in this analysis as data on comorbidity score was not collected on any of the trials. Finally, the sample sizes for the LCSS and FACT-L analyses were limited, thus compromising the ability to do a detailed exploration of these assessments on OS. In addition, only summated total scores, not subscales, of these two tools were explored in this analysis.
This study confirmed the prognostic importance of pretreatment QOL on OS. Although research on prognostic factors of QOL in patients with advanced NSCLC has provided conflicting results regarding different QOL tools and different subscales within one instrument, the basic idea that pretreatment QOL is an important predictor for survival is consistent across most studies. EORTC QLQ-C30 is the most commonly used QOL instrument in the research of QOLs prognostic value on survival in advanced NSCLC. Several studies have demonstrated that the global QOL score assessed with QLQ-C30 is an independent prognostic factor for survival in advanced NSCLC.12,14,15,20
Recently, in a cohort of 391 patients, Efficace et al.13
found that patient self-reported pain and dysphagia measured by QLQ-C30, in addition to gender and PS, were independent significant predictors of survival. An earlier study conducted by Herndon et al.21
had similar finding in that pain, not global QOL, was prognostic for survival. The inconsistency of evidence described above in terms of the prognostic impact of global QOL using EORTC QLQ-C30 could reflect possible multicollinearity that contributed to model instability when global QOL was included in the final multivariate model with other subscales. 13,22
The conflicting results could also be due to different cut-offs used to categorize factors, different selection of factors other than QOL, and even analysis methodology preferences.
The multi-item indices explored in this analysis, LCSS and FACT-L, are not significantly associated with survival outcome in patients with advanced NSCLC in the univariate setting. The limited research of prognostic impact of QOL on survival using LCSS and FACT-L suggests that the total summated QOL scores or some subscales assessed with these tools may independently predict survival in patients with advanced NSCLC.33–35
Changes of LCSS scores from baseline were shown to be associated with efficacy outcomes in a recently reported study.36
The present analysis only explored the total summated scores of LCSS and FACT-L, and further research is needed for the subscales in each tool which might show different features. In addition, because LCSS does not contain many of the important components of the QOL, it may be a limited measurement of QOL.17
One obstacle for routine integration of QOL in clinical trials is due to the increased patient burden. Compared with the multi-item QOL indices, single-item Uniscale is simple and easy to implement and has the inherent advantage of minimal patient burden, especially for those seriously ill patients. However, single-item indices do demonstrate greater variability in the overall range of scores than the multiple-item indices.19,37
Despite this limitation, Uniscale has been shown to be a valid, reliable, and sensitive tool to assess overall QOL and may measure something “broader” than symptom-specific multi-item indices.38
Huschka et al.39
observed that Uniscale actually detects a clinical significant decline better over time in an NCCTG pooled analysis. These merits make Uniscale an attractive stratification factor in advanced NSCLC clinical trials.
The present analysis has confirmed the prognostic impact of previously identified prognostic factors on OS in the univariate settings. Our findings of female gender and good PS as favorable prognostic factors univariately are consistent with previous studies.5,7,13,40
BMI (underweight) was identified to be significantly prognostic for worse survival in the present analysis multivariately. This confirmed the findings of Mandrekar et al.6
in a pooled analysis of NCCTG trials (underweight versus normal weight, HR of OS = 1.77, 95% CI 1.30 –2.40).
In the presence of baseline QOL assessed by Uniscale, the prognostic effect of PS, Hgb, PLT, and ANC were not prominent. The inconsistency of the gender effect on OS between the non-PS-stratified and the PS-stratified multivariate models suggests that the prognostic impact of gender on OS may vary across different levels of PS. Conflicting data have been reported regarding whether PS and gender are significant predictors of survival in the presence of QOL. Langendijk et al.12
observed that PS became nonsignificant in the presence of QOL. Montazeri et al.20
found that both PS and gender’s effect became nonsignificant in the presence of QOL, whereas Efficace et al.13
reported that PS, gender, and QOL were all retained in the final multivariate model. PS is primarily a measure of patient ambulatory ability per se, and thus has limited scope in measuring patient overall well-being compared with Uniscale. Further analysis of our data indicated that baseline PS and Uniscale were correlated (χ2
< 0.001). Together with the finding that the effect of PS became nonsignificant in the multivariate model, QOL assessed by single-item Uniscale might cover the scope of PS in prediction for OS. Of note, however, is that PS 1 and PS 2 were grouped together in this analysis because of a small proportion of PS 2 patients. This might have masked the well-accepted independent negative association between PS 2 and OS.7
The precise prognostic value of baseline Hgb levels on OS remains unclear in advanced NSCLC. There are studies demonstrating baseline Hgb levels as independent prognostic factor for superior survival.5,6,33,41
The Radiation Therapy Oncology Group study’s finding was consistent with ours that Hgb level lost its significance when QOL was added to the model.16
However, the Radiation Therapy Oncology Group study was based on patients with locally advanced NSCLC receiving radiotherapy. Our study also identified ANC as a significant predictor for OS univariately. This may signal some underlying issues, for example, a greater tumor burden (e.g., increasing neutrophils via antibody-dependent cell-mediated cytotoxicity), antichemotactic activities of some mediators produced by the tumor cells. Mandrekar et al. and Sculier et al. reported similar results that white blood cell count was associated with a poor survival outcome.6,7
Similarly, high pretreatment PLT count showed a trend toward worse OS univariately, although this effect was not statistically significant. Thrombocytosis has been recognized as a paraneoplastic symptom and has been shown to predict for shorter survival in lung cancer.33,42
In general, however, there has not been a consistent effort in investigating the prognostic values of complete blood count in advanced NSCLC.
In conclusion, the present analysis demonstrated that pretreatment patient QOL measured by Uniscale is a significant prognostic factor for OS in advanced NSCLC independent of other known factors such as PS, age, gender, BMI, and some laboratory parameters. Based on the current work and a few others published in the literature, it can be concluded that pretreatment patient self-reported QOL is an independent and an important prognostic factor in advanced NSCLC. Pretreatment QOL can prospectively identify patient subgroups with survival more divergent than the survival advantages associated with currently available therapies. This highlights the need to routinely integrate QOL in advanced NSCLC clinical trials either by including it as a stratification factor or by appropriately adjusting for it in the analysis for a proper interpretation of data from trials. Our results also have a broader scope in that a simple, quick, and convenient QOL assessment, Uniscale, can provide clinically meaningful information regarding patient survival.