- Email: [email protected]
The Prognostic Value of Patient-Reported Outcome Data in Patients With Colorectal Hepatic Metastases Who Underwent Surgery
Original Study
The Prognostic Value of Patient-Reported Outcome Data in Patients With Colorectal Hepatic Metastases Who Underwent Surgery Jonathan R. Rees,1,2 Myrddin Rees,3 Angus G.K. McNair,1 Lang’o Odondi,1 Chris Metcalfe,1 Timothy John,3 Fenella K. Welsh,3 Jane M. Blazeby1,2 Abstract Quality of life and survival after liver resection for colorectal cancer liver metastases were studied. Patients who reported poorer global quality and greater weight loss at baseline had worse long-term survival. These findings should be used to inform patients and influence treatment decisions. Introduction: Patient-reported outcomes (PROs) are critical to evaluate clinically effective treatments and evidence suggests that PROs might predict survival. The prognostic value of PROs in patients with isolated liver metastases from colorectal cancer (CRC) who undergo surgery is unclear. In this study we investigated whether baseline PROs are prognostic in this patient group. Patients and Methods: From April 2004 to May 2007, consecutive patients who underwent curative resection of CRC liver metastases completed the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire (QLQ)-C30 and QLQ-LMC21 questionnaires before surgery. Patients were followed until death or data were censored on April 9, 2012. Cox proportional hazards models were used to assess the effect of PROs on survival controlling for predefined clinical covariates. Models were simplified using a backwards stepwise approach and model utility appraised using the Harrell C and Somers D statistics and bootstrap methods. Results: Two hundred thirty-two patients underwent liver resection and 101 (43.5%) survived 5 years. Multivariate analysis controlling for relevant clinical covariates showed that a 10-point improvement in baseline global quality of life scores was associated with a 54% improvement in survival (hazard ratio [HR], 0.46; 95% confidence interval [CI], 0.33-0.63; P < .001), and a clinically significant weight loss was associated with 75% worse survival (HR, 1.75; 95% CI, 1.20-2.55; P ¼ .004). Smaller effects were noted for worsening abdominal pain, taste problems, and fatigue (30%-38% poorer survival). Results of bootstrap resampling suggested that global health and weight loss most reliably predicted survival. Conclusion: Results of this study demonstrated that patients who reported worse baseline global quality of life and increased weight loss before liver resection for CRC liver metastases had significantly poorer survival. These findings if externally validated might be used to inform patients, and could also influence treatment planning and advise follow-up strategies and supportive care. Clinical Colorectal Cancer, Vol. -, No. -, --- ª 2015 Elsevier Inc. All rights reserved. Keywords: Cancer, Colon, Liver, Patient-reported outcome, Prognosis
Introduction 1
Centre for Surgical Research, School of Social and Community Medicine, University of Bristol, Bristol, United Kingdom 2 Division of Surgery, Head and Neck, University Hospitals Bristol NHS Foundation Trust, Bristol, United Kingdom 3 Department of Hepatobiliary Surgery, Hampshire Hospitals Foundation Trust, Basingstoke, Hampshire, United Kingdom Submitted: Apr 23, 2015; Accepted: Jul 20, 2015 Address for correspondence: Jonathan R. Rees, MSc, PhD, FRCS, School of Social and Community Medicine, Canynge Hall, 39 Whatley Rd, Bristol BS8 2PS, United Kingdom Fax: þ44 117 9287235; e-mail contact: [email protected]
1533-0028/$ - see frontmatter ª 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.clcc.2015.07.003
Patient-reported outcomes (PROs) are increasingly recognized to have prognostic value.1 Studies demonstrate the association in locally advanced and metastatic solid cancers such as esophageal,2 gastric,3 and lung4 tumors and in hematological malignancies such as leukemia,5 lymphoma,6 and myeloma.7 Particular PRO domains are prognostic across a range of cancer sites, including role and physical function,8-10 symptoms of fatigue,11 appetite loss,10,12 and pain.9,13 In contrast, reports in localized breast cancer12,14,15 and nonmetastatic prostate cancer have not shown PROs to be prognostic. This might be because localized disease does not exert systemic
Clinical Colorectal Cancer Month 2015
-1
Prognostic Role of PROs After Liver Surgery effects on a patient’s physiology and therefore does not affect selfreported PROs. Furthermore, because individuals with early disease might have very good long-term survival, other comorbidities that occur after the diagnosis of early cancer might have a greater influence on PROs and survival.1 An additional group of patients with malignant disease are those with isolated liver metastases that arise from colorectal cancer (CRC) suitable for surgery. They are reported to have survival rates of 40% at 5 years.16 On the basis of the hypotheses that PROs are prognostic in advanced cancer and not in localized disease, the role of the prognostic value of PROs in patients with isolated liver metastases from CRC is uncertain. Although the disease is confined to the abdominal cavity it has spread via the portal venous circulation to the liver and its systemic effects might still influence PROs and these changes might be relevant to prognosis. However, although the trajectory of PROs after liver surgery for CRC metastases has recently been described17 the prognostic value of PROs in patients who undergo liver surgery for CRC metastases is unknown. Therefore the aim of this study was to examine associations between baseline PRO values before hepatic surgery for CRC metastases and long-term survival.
QLQ-C30 (version 3.0) and the colorectal liver metastases specific module, the QLQ-LMC21.19,20 The QLQ-C30 is a reliable and valid tool that is used to assess functional domains and symptoms relevant to patients with cancer. The validated QLQ-LMC21 module contains 4 scales that are used to assess nutritional problems, activity and/ or vigor problems, pain, and emotional function, and 9 singlesymptom items. Responses from the QLQ-C30 and the QLQLMC21 questionnaires were linearly transformed to scores from 0 to 100.21 High scores represented better quality of life (better function) for the functional scales and, in contrast, high scores represented worse symptoms (more problems) for the symptom scales. Baseline PRO assessment occurred in the 4 weeks before surgery and further assessments at 3, 6, and 12 months after surgery were performed. Time windows of approximately 3 weeks were applied for each assessment. Completion of questionnaires was arranged using the telephone and performed in hospital at baseline and using mail during follow-up. A single telephone reminder was made if required and on return of questionnaires any missing data were completed in a telephone call. Details of patients who died were established from records held at Hampshire Hospitals NHS Trust and direct contact with each patients’ general practitioner and confirmed with a fax that noted the cause of death.
Patients and Methods Data Source and Study Design
Data Analyses
Between April 2004 and May 2007 consecutive patients who underwent standard curative resection of CRC liver metastases completed PRO measures before surgery. Eligible were patients with a histological diagnosis of colonic or rectal cancer and hepatic metastases, who were able to understand the language and content of the questionnaires and complete them with minimal or no assistance. Participation in another questionnaire study that would interfere with this protocol or the presence of a concurrent malignancy (excluding basal cell carcinoma of the skin) were exclusion criteria. Patients were observed until death or the end of the study period when they were censored (April 9, 2012), those lost to follow-up were excluded from the analysis. Participants provided written informed consent and ethical approval was obtained for the research (National Research Ethics Service Committee North East—County Durham and Tees Valley 12/NE/0332).
This study was designed to test the hypotheses that patients with better physical function, less pain, and better global quality of life at baseline assessment will have improved survival.1,22
Clinical Assessment and Treatment Details Computed tomography (CT) and magnetic resonance imaging with liver-specific contrast were used to stage the hepatic metastases and pulmonary staging was undertaken with CT scan of the thorax. Patients with high recurrence risk disease were further staged using laparoscopy with or without laparoscopic ultrasound and selective use of CT positron emission tomography. All procedures were performed by specialist surgeons (F.K.W., T.J., and M.R.) with general anesthesia using standardized nomenclature of the International HepatoPancreato-Biliary Association and latterly facilitated by the use of surgical stapling devices.18 Patients were considered inoperable if the liver disease could not be safely excised in its entirety or extrahepatic disease was identified and these patients were excluded from the study.
2
-
Model-Building. For each model, time 0 was the date of curative surgery, which was within 4 weeks of the baseline PRO assessment. Cox proportional hazards regression23 was used to explore unadjusted associations between baseline PROs and survival (model 1). This model was adjusted (model 2) for age, primary colorectal tumor nodal status, differentiation status of the primary colorectal tumor, largest size of the liver metastases, total number of liver metastases, presence of extrahepatic locally resectable disease, carcinoembryonic antigen levels, and presence of a histologically involved tumor resection margin. The clinical covariates were identified according to a literature review24 and expert consultation and defined a priori. Changes of 10 points on a 0 to 100 scale were considered clinically relevant and hazard ratios (HRs) presented for a 10-point change in PRO score on a continuous scale.
Patient-Reported Outcomes
Model Simplification. Backwards stepwise Cox proportional hazards regression, with a significance level for variable removal of .125 to maximize data inclusion25 was undertaken. Prognostic PRO variables identified using this approach were incorporated into a final simplified model (model 3) after comparison of the three models constructed (Supplemental Table 1). In the final model, variables with a P < .01 were considered relevant to account for multiple testing and reduce the risk of overfitting of data.26 The proportional hazards assumption was assessed by examining proportional hazards plots and the corresponding global test.27
Patient-reported outcomes were assessed using the European Organization for Research and Treatment of Cancer (EORTC) core questionnaire (Quality of Life Questionnaire [QLQ]), the EORTC
Model Validation. To evaluate the performance of the selected final model, Harrell concordance c-statistic was used. This discrimination
Clinical Colorectal Cancer Month 2015
Jonathan R. Rees et al index provides a measure of the ability of a model to distinguish PROs that are prognostic to survival from those that are not prognostic. The index is also expressed in terms of the Somers d rank correlation where d ¼ 2c 1.28,29 When c ¼ 1 (or d ¼ 1) this implies perfect concordance between the predicted outcome from the model under test and the observed outcome from the data set. In contrast, if c ¼ 0.5 (or d ¼ 0), this indicates random discrimination and any associations between predicted outcomes from the prognostic model and observed outcomes from the data set are due to chance. Models derived from this structured approach were tested using a randomly generated test set and compared using linear combinations
of coefficients after estimation.30 Values of c and d indices are reported, where larger values of these indices (c and d) indicate better prediction of outcome by the selected model. Bootstrap methods with 1000 iterations were used to assess stability of the selected model where reported bootstrap inclusion fractions provided the frequency with which a particular variable is selected over the 1000 iterations.31 It was assumed that the observed data effectively represented the true population distribution and “random resampling with replacement” of the observed data set was undertaken using samples with the same number of observations as the original data set.
Table 1 Clinical and Sociodemographic Details of Patients Who Underwent Liver Resection for Colorectal Liver Metastases, Patients Who Survived in LTFU, Patients Who Did Not Have a Resection, and Patients Who Died Hepatic Resection of Colorectal Metastases (n [ 232)a
Hepatic Resection of Colorectal Metastases Alive at LTFU (n [ 101)
Deceased (n [ 131)
Mean Age, Years (SD)
63.1 (8.5)
71.0 (10.0)
70.1 (10.3)
Male Sex, n (%)
162 (69.8)
70 (69.3)
90 (68.7)
Less than compulsory/compulsory
109 (47.0)
49 (48.5)
60 (45.8)
Postcompulsory school education
118 (50.9)
50 (49.5)
68 (51.9)
5 (2.1)
2 (2.0)
3 (2.3)
Employed (full, part, self-employed)
106 (45.7)
43 (42.6)
63 (48.1)
Retired
118 (50.9)
56 (55.4)
62 (47.3)
8 (3.4)
2 (2.0)
6 (4.6)
40 (17.2)
13 (12.9)
27 (20.6)
192 (82.8)
88 (87.1)
104 (79.4)
Characteristic
Highest Level of Education, n (%)
Unknown Current Employment Status, n (%)
Other/unknown Cohabitation, n (%) Alone With family or other adults Unknown
0 (0)
0 (0)
0 (0)
Type of Previous Colorectal Surgery, n (%) Hemicolectomy Anterior resection
76 (32.8)
31 (30.7)
45 (34.4)
115 (49.6)
58 (57.4)
57 (43.5)
Abdominoperineal resection
11 (4.7)
Other/unknown
30 (12.9)
0 (0)
11 (8.4)
12 (11.9)
18 (13.7)
Stage of Colorectal Primary, n (%) I
12 (5.2)
5 (5.0)
7 (5.4)
II
37 (15.9)
16 (15.8)
21 (16.0)
III
67 (28.9)
29 (28.7)
38 (29.0)
IV
97 (41.8)
42 (41.6)
55 (42.0)
Unknown
19 (8.2)
9 (8.9)
10 (7.6)
Detection of Liver Metastases, n (%) Synchronous with primary tumor
54 (23.3)
20 (19.8)
34 (26.0)
Routine follow-up (asymptomatic)
143 (61.6)
65 (64.4)
78 (59.5)
Symptomatic
19 (8.2)
8 (7.9)
11 (8.4)
Other/unknown
16 (6.9)
8 (7.9)
8 (6.1)
ASA Grade, n (%) 1/2
195 (84.1)
87 (86.1)
108 (82.4)
3/4
32 (13.8)
12 (11.9)
20 (15.3)
5 (2.1)
2 (2.0)
3 (2.3)
Unknown
Abbreviations: ASA ¼ American Society of Anesthesiology; LTFU ¼ long-term follow-up. a For 9 patients who were planned for liver resection no surgery was undertaken because of extensive disease and the patients were excluded from the study.
Clinical Colorectal Cancer Month 2015
-3
Prognostic Role of PROs After Liver Surgery Results Patients Of the 241 patients enrolled, 9 had unresectable disease and were excluded, and 232 entered the study (Table 1). Baseline data were collected from all patients before surgery17 and 101 (43.5%) were alive at study censoring with a median follow-up of 7.5 years (range, 6.4-8.1 years). Clinical and demographic details of the 232 study participants are shown in Table 1.
Association Between Baseline PROs and Survival The unadjusted HR for the association between PROs and survival suggest that global quality of life, physical function, constipation, weight loss, and taste have potential prognostic relevance
(Table 2). The proportional hazards assumption was not violated by this model (c2 ¼ 13.74; P ¼ .98). A Cox proportional hazard model, adjusted for the prespecified clinical covariates (Table 3) demonstrated that a 10-point improvement in quality of life was associated with a 55% improvement in survival (HR, 0.45; 95% confidence interval [CI], 0.31-0.65; P < .001). Weaker associations were seen between survival and a number of symptoms; with less severe (by 10 points) constipation (HR, 0.80; 95% CI, 0.65-1.00; P ¼ .046) and diarrhea (HR, 0.79; 95% CI, 0.63-1.00; P ¼ .044) each associated with a 20% improvement in survival. In contrast, poorer (by 10 points) sexual function (HR, 1.16; 95% CI, 1.03-1.32; P ¼ .013) and taste (HR, 1.47; 95% CI, 1.06-2.05; P ¼ .02) were associated with
Table 2 Unadjusted HR With 95% CI for Associations Between a 10-Point Increase in PRO Scores From the EORTC QLQ-C30 and QLQ-LMC21 Questionnaires and Survival in 232 Patients Treated With Curative Surgery for CRC Liver Metastases (Model 1)
Variable
Mean PRO Score Before Surgery (95% CI)
Model 1 HR
95% CI
P
QLQ-C30 Scalesa Global health
76.3 (73.8-78.7)
0.69
(0.54-0.89)
.004
Physical function
88.0 (85.9-90.0)
1.41
(1.06-1.89)
.019
Role function
80.7 (77.2-84.2)
0.91
(0.76-1.07)
.255
Emotional function
79.8 (77.2-82.3)
1.16
(0.92-1.45)
.219
Cognitive function
87.2 (85.0-89.4)
1.10
(0.90-1.35)
.341
Social function
78.5 (75.0-81.9)
1.06
(0.88-1.28)
.544
QLQ-C30 Scales and Itemsb Fatigue Nausea and vomiting
21.8 (19.1-24.6)
1.10
(0.84-1.44)
.472
3.4 (9.6-14.8)
0.78
(0.54-1.12)
.178
Pain
12.2 (2.0-4.7)
1.08
(0.91-1.29)
.374
Dyspnoea
11.0 (8.3-13.6)
1.01
(0.88-1.16)
.912
Insomnia
22.3 (18.8-25.9)
1.01
(0.88-1.15)
.911
7.6 (5.1-10.2)
1.01
(0.72-1.40)
.974
Appetite loss Constipation
9.7 (7.4-12.1)
0.82
(0.70-0.96)
.016
Diarrhea
12.2 (9.3-15.0)
0.92
(0.79-1.07)
.268
Financial difficulty
13.5 (10.2-16.8)
0.94
(0.83-1.07)
.354
QLQ-LMC21 Scales and Itemsb Eating problems
7.6 (5.5-9.7)
1.14
(0.76-1.72)
.532
Activity/vigor problems
29.6 (26.3-33.0)
0.83
(0.65-1.06)
.128
Abdominal pain
14.4 (12.1-16.7)
1.07
(0.87-1.32)
.539
Anxiety problems
19.4 (15.7-23.2)
1.20
(0.97-1.50)
.095
Nutritional issues
11.6 (8.5-14.7)
1.51
(1.09-2.11)
.015
Taste problems
34.6 (31.5-37.7)
1.30
(1.09-1.55)
.004
6.8 (4.6-8.9)
0.98
(0.84-1.15)
.834
Sore mouth/tongue
8.5 (5.8-11.3)
1.06
(0.91-1.24)
.437
Tingling in fingers
15.8 (12.5-19.1)
1.00
(0.91-1.09)
.947
5.5 (3.4-7.6)
0.00
e
e
Contact with friends
10.2 (7.4-13.1)
0.94
(0.77-1.15)
.563
Talking about feelings
31.1 (26.5-37.7)
0.94
(0.79-1.12)
.488
Sexual function
30.6 (25.1-36.1)
1.06
(0.98-1.15)
.123
Dry mouth
Jaundice
Abbreviations: CRC ¼ colorectal cancer; EORTC ¼ European Organization for Research and Treatment of Cancer; HR ¼ hazard ratio; PRO ¼ patient-reported outcome. a Higher score indicates better function. b Higher score indicates worse symptom or problem.
4
-
Clinical Colorectal Cancer Month 2015
Jonathan R. Rees et al Table 3 Adjusteda HR With 95% CI for Model 1 for 10-Point Change in Baseline PRO Scores From the EORTC QLQ-C30 and QLQ-LMC21 Questionnaires and Survival in 232 Patients Treated With Curative Surgery for CRC Liver Metastases (Model 2)
Variable
Mean PRO Score Before Surgery (95% CI)
Model 2 HR
95% CI
P
QLQ-C30 Scales Global health
76.3 (73.8-78.7)
0.45
(0.31-0.65)
<.001
Physical function
88.0 (85.9-90.0)
1.14
(0.76-1.70)
.535
Role function
80.7 (77.2-84.2)
0.96
(0.77-1.19)
.684
Emotional function
79.8 (77.2-82.3)
1.29
(0.91-1.82)
.157
Cognitive function
87.2 (85.0-89.4)
1.26
(0.91-1.74)
.159
Social function
78.5 (75.0-81.9)
1.16
(0.88-1.53)
.292
QLQ-C30 Scales and Items Fatigue Nausea and vomiting
21.8 (19.1-24.6)
1.02
(0.71-1.47)
.898
3.4 (9.6-14.8)
0.82
(0.52-1.31)
.410
Pain
12.2 (2.0-4.7)
1.16
(0.89-1.51)
.271
Dyspnoea
11.0 (8.3-13.6)
0.97
(0.77-1.23)
.823
Insomnia
.212
22.3 (18.8-25.9)
0.89
(0.74-1.07)
Appetite loss
7.6 (5.1-10.2)
0.85
(0.53-1.38)
.518
Constipation
9.7 (7.4-12.1)
0.80
(0.65-1.00)
.046
Diarrhea
12.2 (9.3-15.0)
0.79
(0.63-1.00)
.044
Financial difficulty
13.5 (10.2-16.8)
0.99
(0.81-1.21)
.910
QLQ-LMC21 Scales and Items Eating problems
7.6 (5.5-9.7)
1.31
(0.75-2.29)
.349
Activity/vigor problems
29.6 (26.3-33.0)
0.72
(0.51-1.03)
.075
Abdominal pain
14.4 (12.1-16.7)
1.38
(0.95-2.00)
.095
Anxiety problems
19.4 (15.7-23.2)
1.19
(0.84-1.70)
.329
Nutritional issues
11.6 (8.5-14.7)
1.52
(0.87-2.66)
.141
Taste problems
.020
34.6 (31.5-37.7)
1.48
(1.06-2.05)
Dry mouth
6.8 (4.6-8.9)
0.90
(0.71-1.15)
.400
Sore mouth/tongue
8.5 (5.8-11.3)
1.18
(0.85-1.63)
.323
Tingling in fingers
15.8 (12.5-19.1)
1.04
(0.88-1.22)
.634
5.5 (3.4-7.6)
0.00
e
Contact with friends
10.2 (7.4-13.1)
0.93
(0.67-1.28)
.636
Talking about feelings
31.1 (26.5-37.7)
0.99
(0.76-1.28)
.915
Sexual function
30.6 (25.1-36.1)
1.17
(1.03-1.32)
.013
Jaundice
e
Abbreviations: CRC ¼ colorectal cancer; EORTC ¼ European Organization for Research and Treatment of Cancer; HR ¼ hazard ratio; PRO ¼ patient-reported outcome. a Adjusted for predefined clinical covariates.
poorer survival of 16% to 47%. There was no evidence to suggest that the model violated the proportional hazards assumption (c2 ¼ 36.32; P ¼ .407).
vigor; HR, 1.31; 95% CI, 1.15-1.45; P ¼ .001), which were associated with 30% to 38% greater mortality. There was no evidence to suggest that the model violated the proportional hazards assumption (c2 ¼ 10.17; P ¼ .86).
Final Adjusted and Simplified Model for Associations Between PROs and Survival (Model 3)
Model Validation
Hazard ratios from the stepwise backwards Cox proportional hazards regression are shown in Table 4. A 10-point higher global quality of life was associated with a 54% better survival (HR, 0.46; 95% CI, 0.33-0.63; P < .001) and a 10-point higher reported weight loss was associated with 75% greater mortality (HR, 1.75; 95% CI, 1.20-2.55; P ¼ .004). Smaller effects were seen for 10-point greater abdominal pain (HR, 1.38; 95% CI, 1.08-1.77; P ¼ .01), taste problems (HR, 1.30; 95% CI, 1.08-1.57; P ¼ .005), and fatigue (activity and/or
The c index value for the model incorporating solely PROs was 0.71; when all clinical and patient-reported covariates were incorporated the index increased to 0.79 and was reduced to 0.78 for the simplified model obtained using backward stepwise procedures (Supplemental Table 2 in the online version). Bootstrap resampling showed that global health (89.5% of samples), constipation (86.2% of samples), and weight loss (64.8% of samples) were reliably (> 60% of bootstrap samples) included in the internal validation (Table 4).
Clinical Colorectal Cancer Month 2015
-5
Prognostic Role of PROs After Liver Surgery Table 4 Simplified Model Incorporating Potentially Relevant Prognostic PROs With Adjusteda HRs Describing the Association of a 10Point (Clinically Relevant) Change in Baseline PRO Score and Survival in 232 Patients Treated With Curative Surgery for CRC Liver Metastases Proportion of Bootstrap Resamples With Effect at a Significance Level of £.05
Model 3 Variable
HR
95% CI
P
%
N
QLQ-C30 Scales Global health
0.46
(0.33-0.63)
<.001
89.5
895
Emotional function
1.20
(0.99-1.47)
.068
34.2
342
Constipation
0.75
(0.62-0.89)
.001
86.2
862
Diarrhea
0.88
(0.75-1.03)
.123
17.4
174
Insomnia
0.88
(0.75-1.03)
.110
18.0
180
QLQ-C30 Items
QLQ-LMC21 Scales and Items Weight loss
1.75
(1.20-2.55)
.004
64.8
648
Abdominal pain
1.38
(1.08-1.77)
.010
44.1
441
Taste
1.30
(1.08-1.57)
.005
56.3
563
Activity/vigor problems
1.31
(1.15-1.45)
.001
56.0
560
Sex function
1.13
(1.04-1.23)
.006
26.4
264
The proportion of bootstrap resamples (n ¼ 1000) for which prognostic PRO variables identified by the final model (model 3) were selected. Abbreviations: CRC ¼ colorectal cancer; HR ¼ hazard ratio; PRO ¼ patient-reported outcome. a Adjusted for predefined clinical covariates.
Discussion Results of this exploratory study suggest that differences in baseline PROs in selected domains including symptoms might be prognostic in patients who undergo surgery with curative intent for CRC liver metastases. Improved quality of life, less constipation, and less diarrhea were associated with better survival, and worse problems with weight loss, abdominal pain, and taste; fatigue (activity and vigor) and worse problems with sex life were associated with poorer survival. Internal validation using bootstrap resamples highlighted that global health (89.5% of samples); constipation (86.2% of samples), and weight loss (64.8% of samples) were the most stable predictors of outcome. These findings suggested that in addition to patient and tumor characteristics, self-reported measures of quality of life and symptoms have prognostic value. Therefore, clinicians should be alerted to patients who at baseline have poorer global quality of life scores or particular issues with constipation and weight loss because they might be at increased risk of disease recurrence and/or death and these are findings that should be considered when initial treatment and later follow-up are planned. The prognostic value of PROs in patients who undergo curative liver surgery for CRC has not previously been described, although results of a single study in patients with hepatitis B and unresectable or metastatic hepatocellular carcinoma10 have suggested that appetite loss, and physical and role function might be prognostic. However, these findings should be viewed with caution because they might not be directly relevant to patients with resectable CRC liver metastases. Global quality of life and weight loss have, however, been suggested to be prognostic in a number of other tumor types. Studies in localized esophagogastric cancer,32 lung cancer and lymphoma,6 but not localized breast cancer12 suggest a prognostic role for global quality of life, and in patients with metastatic disease, global quality of
6
-
Clinical Colorectal Cancer Month 2015
life might be prognostic in breast carcinoma,33 CRC,34 lung cancer,35 and prostate malignancy.9 In addition, weight loss has been associated with poor prognosis across a range of cancer patients who receive chemotherapy36 and results of specific work in metastatic lung cancer also suggest that this symptom has prognostic value.37 The reasons that PROs have prognostic significance are less clear. It is possible they might be more sensitive indicators of disease progression responding to subtle physiological changes perhaps driven by micrometastatic disease or early tumor recurrence. These changes, reflecting subclinical disease and hypothetically mediated by proinflammatory agents such as interleukin (IL) 2, IL-6, and tumor necrosis factor a might influence PROs before clinically detectable disseminated disease becomes apparent.38-40 Further possible explanations for the predictive value of PROs include the suggestion that patients who report better PROs might have more positive lifestyle behaviors. These individuals putatively have greater physiological reserve and therefore experience longer survival. In addition, it is suggested that patients who describe better PROs are recognized to have greater adherence to treatment regimens. The reasons are unclear but might reflect that these individuals have fewer comorbidities and fewer reactions to previous therapy, meaning that their exposure to potentially curative therapy is maximized and their survival potentially improved.1 Considering weight loss in cancer patients, possible etiologies are multiple; tumors might directly impede gut function or suppress appetite, and many therapies interfere with appetite, taste, and gut function, and are often emetogenic.37 Furthermore, systemic inflammation caused by tumors might also drive weight loss. Weight loss has also been associated with poorer survival but the mechanisms for this effect are also unclear although increases in proinflammatory cytokines and growth factors that drive metabolic
Jonathan R. Rees et al disturbance and lead to cachexia and tumor growth are thought to be important.40 In this study we described a large (n ¼ 232) well characterized cohort and used validated generic and disease PRO instruments in a prospective design. Missing PRO data were rare but participants in this study were from a single center, and those with unresectable liver disease did have a lower proportion of stage III and IV primary colorectal tumors although the absolute numbers were very small; therefore care must be exercised when considering the generalizability of these findings. The analysis approach used did result in multiple testing, however, this limitation might be offset by the application of a stringent significance level at P .01. Furthermore, model simplification methods resulted in a small decrease in the Harrell c statistic when the model incorporating all covariates considered clinically relevant (model 2) was compared with the final model (model 3) generated by the backward stepwise selection approach. This phenomenon reflected the necessary compromise struck between model simplicity (by reducing the number of variables) and prognostic ability. However the final model (model 3) did have sufficient events per covariate for unbiased estimation and accurate CI coverage probabilities, potentially offsetting the consequences of model simplification.41 Confounding might also arise from the more frequent future use of chemotherapy in this setting.
Conclusion Results of this study provide the first evidence to suggest that patients who report worse baseline global quality of life, more constipation, and greater problems with weight loss before potentially curative liver resection for CRC liver metastases have significantly poorer survival. External validation of these findings is necessary and must demonstrate that use of PRO data with current clinical predictors of outcome significantly increase the ability of the proposed model to predict outcome. If these criteria are achieved then the findings might be used in clinical practice to inform patients, influence treatment planning including surgery, neoadjuvant and/or adjuvant therapy, and supportive care, and also help determine follow-up strategies.
Clinical Practice Points The prognostic value of PROs in patients who undergo liver
resection for CRC metastases is unknown. Patients who report poorer global quality of life and greater
weight loss at baseline have worse long-term survival. Clinicians should counsel patients regarding the potential effect
that poorer global quality of life and greater weight loss have on survival. They should further ensure that psychological, physical, and optimal nutritional support is available to individuals who undergo liver resection for CRC metastases.
Disclosure The authors have stated that they have no conflicts of interest.
Supplemental Data Supplemental tables accompanying this article can be found in the online version at http://dx.doi.org/10.1016/j.clcc.2015.07.003.
References 1. Gotay CC, Kawamoto CT, Bottomley A, Efficace F. The prognostic significance of patient-reported outcomes in cancer clinical trials. J Clin Oncol 2008; 26:1355-63. 2. Blazeby JM, Brookes ST, Alderson D. Prognostic value of quality of life scores in patients with oesophageal cancer. Br J Surg 2000; 87:362-73. 3. Blazeby JM, Metcalfe C, Nicklin J, Barham CP, Donovan J, Alderson D. Association between quality of life scores and short-term outcome after surgery for cancer of the oesophagus or gastric cardia. Br J Surg 2005; 92:1502-7. 4. Barney BJ, Wang XS, Lu C, et al. Prognostic value of patient-reported symptom interference in patients with late-stage lung cancer. Qual Life Res 2013; 22: 2143-50. 5. Efficace F, Lo-Coco F. Using patient-reported health status to improve prognostic assessment in patients with acute myeloid leukemia: current challenges and future applications. Haematologica 2013; 98:7-9. 6. Jerkeman M, Kaasa S, Hjermstad M, Kvaloy S, Cavallin-Stahl E. Health-related quality of life and its potential prognostic implications in patients with aggressive lymphoma: a Nordic Lymphoma Group Trial. Med Oncol 2001; 18:85-94. 7. Dubois D, Dhawan R, van de Velde H, et al. Descriptive and prognostic value of patient-reported outcomes: the bortezomib experience in relapsed and refractory multiple myeloma. J Clin Oncol 2006; 24:976-82. 8. Efficace F, Bottomley A, Coens C, et al. Does a patient’s self-reported healthrelated quality of life predict survival beyond key biomedical data in advanced colorectal cancer? Eur J Cancer 2006; 42:42-9. 9. Sullivan PW, Nelson JB, Mulani PM, Sleep D. Quality of life as a potential predictor for morbidity and mortality in patients with metastatic hormonerefractory prostate cancer. Qual Life Res 2006; 15:1297-306. 10. Yeo W, Mo FK, Koh J, et al. Quality of life is predictive of survival in patients with unresectable hepatocellular carcinoma. Ann Oncol 2006; 17:1083-9. 11. Brown PD, Maurer MJ, Rummans TA, et al. A prospective study of quality of life in adults with newly diagnosed high-grade gliomas: the impact of the extent of resection on quality of life and survival. Neurosurgery 2005; 57:495-504, discussion 495e504. 12. Efficace F, Therasse P, Piccart MJ, et al. Health-related quality of life parameters as prognostic factors in a nonmetastatic breast cancer population: an international multicenter study. J Clin Oncol 2004; 22:3381-8. 13. Efficace F, Bottomley A, Smit EF, et al. Is a patient’s self-reported health-related quality of life a prognostic factor for survival in nonesmall-cell lung cancer patients? A multivariate analysis of prognostic factors of EORTC study 08975. Ann Oncol 2006; 17:1698-704. 14. Coates AS, Hurny C, Peterson HF, et al. Quality-of-life scores predict outcome in metastatic but not early breast cancer. International Breast Cancer Study Group. J Clin Oncol 2000; 18:3768-74. 15. Tross S, Herndon J 2nd, Korzun A, et al. Psychological symptoms and disease-free and overall survival in women with stage II breast cancer. Cancer and Leukemia Group B. J Natl Cancer Inst 1996; 88:661-7. 16. Nordlinger B, Sorbye H, Glimelius B, et al. Perioperative FOLFOX4 chemotherapy and surgery versus surgery alone for resectable liver metastases from colorectal cancer (EORTC 40983): long-term results of a randomised, controlled, phase 3 trial. Lancet Oncol 2013; 14:1208-15. 17. Rees J, Blazeby J, Fayers P, et al. Patient-reported outcomes after hepatic resection of colorectal cancer metastases. J Clin Oncol 2012; 30:1364-70. 18. Strasberg SM, Belghiti J, Clavien PA. Terminology of liver anatomy and resections. HPB (Oxford) 2000; 2:333-9. 19. Blazeby JM, Fayers P, Conroy T, et al. Validation of the European Organization for Research and Treatment of Cancer QLQ-LMC21 questionnaire for assessment of patient-reported outcomes during treatment of colorectal liver metastases. Br J Surg 2009; 96:291-8. 20. Kavadas V, Blazeby JM, Conroy T, et al. Development of an EORTC diseasespecific quality of life questionnaire for use in patients with liver metastases from colorectal cancer. Eur J Cancer 2003; 39:1259-63. 21. Fayers P, Aaronson N, Bjordal K, Groenvold M, Curran D, Bottomley A. EORTC QLQ-C30 Scoring Manual. 2nd ed. Brussels: EORTC Publications; 2002. 22. Quinten C, Martinelli F, Coens C, et al. A global analysis of multitrial data investigating quality of life and symptoms as prognostic factors for survival in different tumor sites. Cancer 2014; 120:302-11. 23. Kirkwood BR, Sterne JA. Essential Medical Statistics. 2nd ed. Malden, MA: Blackwell Science; 2003. 24. Rees M, Tekkis PP, Welsh FK, O’Rourke T, John TG. Evaluation of long-term survival after hepatic resection for metastatic colorectal cancer: a multifactorial model of 929 patients. Ann Surg 2008; 247:125-35. 25. Kaszycki P, Fedorovych D, Ksheminska H, Babyak L, Wojcik D, Koloczek H. Chromium accumulation by living yeast at various environmental conditions. Microbiol Res 2004; 159:11-7. 26. Babyak MA. What you see may not be what you get: a brief, nontechnical introduction to overfitting in regression-type models. Psychosom Med 2004; 66: 411-21. 27. Grambsch PM. TTM: proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994; 81:515-26. 28. Harrell FE Jr, Lee KL, Califf RM, Pryor DB, Rosati RA. Regression modelling strategies for improved prognostic prediction. Stat Med 1984; 3:143-52. 29. Somers RH. A new asymmetric measure of association for ordinal variables. Am Sociol Rev 1962; 27:799-811.
Clinical Colorectal Cancer Month 2015
-7
Prognostic Role of PROs After Liver Surgery 30. Bacon SL, Watkins LL, Babyak M, et al. Effects of daily stress on autonomic cardiac control in patients with coronary artery disease. Am J Cardiol 2004; 93:1292-4. 31. Royston P, Sauerbrei W. Multivariable Model-Building: A Pragmatic Approach to Regression Analysis Based on Fractional Polynomials for Modelling Continuous Variables. Chichester, England, Hoboken, NJ: John Wiley; 2008. 32. Chau I, Norman AR, Cunningham D, Waters JS, Oates J, Ross PJ. Multivariate prognostic factor analysis in locally advanced and metastatic esophago-gastric cancerepooled analysis from three multicenter, randomized, controlled trials using individual patient data. J Clin Oncol 2004; 22:2395-403. 33. Winer EP, Berry DA, Woolf S, et al. Failure of higher-dose paclitaxel to improve outcome in patients with metastatic breast cancer: cancer and leukemia group B trial 9342. J Clin Oncol 2004; 22:2061-8. 34. Maisey NR, Norman A, Watson M, Allen MJ, Hill ME, Cunningham D. Baseline quality of life predicts survival in patients with advanced colorectal cancer. Eur J Cancer 2002; 38:1351-7. 35. Maione P, Perrone F, Gallo C, et al. Pretreatment quality of life and functional status assessment significantly predict survival of elderly patients with advanced nonesmall-cell lung cancer receiving chemotherapy: a prognostic analysis of the
8
-
Clinical Colorectal Cancer Month 2015
36. 37.
38. 39. 40. 41.
multicenter Italian lung cancer in the elderly study. J Clin Oncol 2005; 23: 6865-72. Dewys WD, Begg C, Lavin PT, et al. Prognostic effect of weight loss prior to chemotherapy in cancer patients. Eastern Cooperative Oncology Group. Am J Med 1980; 69:491-7. Sánchez-Lara K, Turcott JG, Juárez E, et al. Association of nutrition parameters including bioelectrical impedance and systemic inflammatory response with quality of life and prognosis in patients with advanced non-small-cell lung cancer: a prospective study. Nutr Cancer 2012; 64:526-34. Blazeby JM, Brookes ST, Alderson D. The prognostic value of quality of life scores during treatment for oesophageal cancer. Gut 2001; 49:227-30. Djarv T, Metcalfe C, Avery KN, Lagergren P, Blazeby JM. Prognostic value of changes in health-related quality of life scores during curative treatment for esophagogastric cancer. J Clin Oncol 2010; 28:1666-70. Abramovitch R, Marikovsky M, Meir G, Neeman M. Stimulation of tumour growth by wound-derived growth factors. Br J Cancer 1999; 79:1392-8. Vittinghoff E, McCulloch CE. Relaxing the rule of ten events per variable in logistic and Cox regression. Am J Epidemiol 2007; 165:710-8.
Jonathan R. Rees et al Supplemental Table 1 Cox Regression Models for the Association of a 10-Point (Clinically Relevant) Change in Baseline PRO Score and Survival in 232 Patients Treated With Curative Surgery for CRC Liver Metastases Variable
Difference in Coefficients of HR (95% CI)
PROs and Clinical Covariates in Model 2 Compared With Prognostically Relevant PROs and Clinical Covariates in Model 3 PROs and Clinical Covariates in Model 2 Compared With PROs in Model 1 Prognostically Relevant PROs and Clinical Covariates in Model 3 Compared With PROs in Model 1
P
0.06 (0.06 to 0.18)
.354
0.04 (0.04 to 0.12) 0.01 (0.11 to 0.08)
.312 .760
Comparison of coefficient of HR using linear combinations of coefficients after estimation in a randomly generated test set comparing unadjusted (PROs: model 1), adjusted (PROs and clinical covariates: model 2) and simplified (prognostically relevant PROs and clinical covariates: model 3), suggesting no loss in model predictive utility. Abbreviations: CRC ¼ colorectal cancer; HR ¼ hazard ratio; PRO ¼ patient-reported outcome.
Supplemental Table 2 Somers D and Harrell C Statistics for Unadjusted (Model 1), Adjusted (Model 2), and Simplified (Model 3) Cox Regression Models for the Association of a 10-Point (Clinically Relevant) Change in Baseline PRO Score and Survival in 232 Patients Treated With Curative Surgery for CRC Liver Metastases Concordance Metric Variable PROs: Model 1 PROs and Clinical Covariates: Model 2 Prognostically Relevant PROs and Clinical Covariates: Model 3
Somers D
Harrell C
0.42 0.58 0.57
0.71 0.79 0.78
Abbreviations: CRC ¼ colorectal cancer; PRO ¼ patient-reported outcome.
Clinical Colorectal Cancer Month 2015
- 8.e1
The Prognostic Value of Patient-Reported Outcome Data in Patients With Colorectal Hepatic Metastases Who Underwent Surgery Jonathan R. Rees,1,2 Myrddin Rees,3 Angus G.K. McNair,1 Lang’o Odondi,1 Chris Metcalfe,1 Timothy John,3 Fenella K. Welsh,3 Jane M. Blazeby1,2 Abstract Quality of life and survival after liver resection for colorectal cancer liver metastases were studied. Patients who reported poorer global quality and greater weight loss at baseline had worse long-term survival. These findings should be used to inform patients and influence treatment decisions. Introduction: Patient-reported outcomes (PROs) are critical to evaluate clinically effective treatments and evidence suggests that PROs might predict survival. The prognostic value of PROs in patients with isolated liver metastases from colorectal cancer (CRC) who undergo surgery is unclear. In this study we investigated whether baseline PROs are prognostic in this patient group. Patients and Methods: From April 2004 to May 2007, consecutive patients who underwent curative resection of CRC liver metastases completed the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire (QLQ)-C30 and QLQ-LMC21 questionnaires before surgery. Patients were followed until death or data were censored on April 9, 2012. Cox proportional hazards models were used to assess the effect of PROs on survival controlling for predefined clinical covariates. Models were simplified using a backwards stepwise approach and model utility appraised using the Harrell C and Somers D statistics and bootstrap methods. Results: Two hundred thirty-two patients underwent liver resection and 101 (43.5%) survived 5 years. Multivariate analysis controlling for relevant clinical covariates showed that a 10-point improvement in baseline global quality of life scores was associated with a 54% improvement in survival (hazard ratio [HR], 0.46; 95% confidence interval [CI], 0.33-0.63; P < .001), and a clinically significant weight loss was associated with 75% worse survival (HR, 1.75; 95% CI, 1.20-2.55; P ¼ .004). Smaller effects were noted for worsening abdominal pain, taste problems, and fatigue (30%-38% poorer survival). Results of bootstrap resampling suggested that global health and weight loss most reliably predicted survival. Conclusion: Results of this study demonstrated that patients who reported worse baseline global quality of life and increased weight loss before liver resection for CRC liver metastases had significantly poorer survival. These findings if externally validated might be used to inform patients, and could also influence treatment planning and advise follow-up strategies and supportive care. Clinical Colorectal Cancer, Vol. -, No. -, --- ª 2015 Elsevier Inc. All rights reserved. Keywords: Cancer, Colon, Liver, Patient-reported outcome, Prognosis
Introduction 1
Centre for Surgical Research, School of Social and Community Medicine, University of Bristol, Bristol, United Kingdom 2 Division of Surgery, Head and Neck, University Hospitals Bristol NHS Foundation Trust, Bristol, United Kingdom 3 Department of Hepatobiliary Surgery, Hampshire Hospitals Foundation Trust, Basingstoke, Hampshire, United Kingdom Submitted: Apr 23, 2015; Accepted: Jul 20, 2015 Address for correspondence: Jonathan R. Rees, MSc, PhD, FRCS, School of Social and Community Medicine, Canynge Hall, 39 Whatley Rd, Bristol BS8 2PS, United Kingdom Fax: þ44 117 9287235; e-mail contact: [email protected]
1533-0028/$ - see frontmatter ª 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.clcc.2015.07.003
Patient-reported outcomes (PROs) are increasingly recognized to have prognostic value.1 Studies demonstrate the association in locally advanced and metastatic solid cancers such as esophageal,2 gastric,3 and lung4 tumors and in hematological malignancies such as leukemia,5 lymphoma,6 and myeloma.7 Particular PRO domains are prognostic across a range of cancer sites, including role and physical function,8-10 symptoms of fatigue,11 appetite loss,10,12 and pain.9,13 In contrast, reports in localized breast cancer12,14,15 and nonmetastatic prostate cancer have not shown PROs to be prognostic. This might be because localized disease does not exert systemic
Clinical Colorectal Cancer Month 2015
-1
Prognostic Role of PROs After Liver Surgery effects on a patient’s physiology and therefore does not affect selfreported PROs. Furthermore, because individuals with early disease might have very good long-term survival, other comorbidities that occur after the diagnosis of early cancer might have a greater influence on PROs and survival.1 An additional group of patients with malignant disease are those with isolated liver metastases that arise from colorectal cancer (CRC) suitable for surgery. They are reported to have survival rates of 40% at 5 years.16 On the basis of the hypotheses that PROs are prognostic in advanced cancer and not in localized disease, the role of the prognostic value of PROs in patients with isolated liver metastases from CRC is uncertain. Although the disease is confined to the abdominal cavity it has spread via the portal venous circulation to the liver and its systemic effects might still influence PROs and these changes might be relevant to prognosis. However, although the trajectory of PROs after liver surgery for CRC metastases has recently been described17 the prognostic value of PROs in patients who undergo liver surgery for CRC metastases is unknown. Therefore the aim of this study was to examine associations between baseline PRO values before hepatic surgery for CRC metastases and long-term survival.
QLQ-C30 (version 3.0) and the colorectal liver metastases specific module, the QLQ-LMC21.19,20 The QLQ-C30 is a reliable and valid tool that is used to assess functional domains and symptoms relevant to patients with cancer. The validated QLQ-LMC21 module contains 4 scales that are used to assess nutritional problems, activity and/ or vigor problems, pain, and emotional function, and 9 singlesymptom items. Responses from the QLQ-C30 and the QLQLMC21 questionnaires were linearly transformed to scores from 0 to 100.21 High scores represented better quality of life (better function) for the functional scales and, in contrast, high scores represented worse symptoms (more problems) for the symptom scales. Baseline PRO assessment occurred in the 4 weeks before surgery and further assessments at 3, 6, and 12 months after surgery were performed. Time windows of approximately 3 weeks were applied for each assessment. Completion of questionnaires was arranged using the telephone and performed in hospital at baseline and using mail during follow-up. A single telephone reminder was made if required and on return of questionnaires any missing data were completed in a telephone call. Details of patients who died were established from records held at Hampshire Hospitals NHS Trust and direct contact with each patients’ general practitioner and confirmed with a fax that noted the cause of death.
Patients and Methods Data Source and Study Design
Data Analyses
Between April 2004 and May 2007 consecutive patients who underwent standard curative resection of CRC liver metastases completed PRO measures before surgery. Eligible were patients with a histological diagnosis of colonic or rectal cancer and hepatic metastases, who were able to understand the language and content of the questionnaires and complete them with minimal or no assistance. Participation in another questionnaire study that would interfere with this protocol or the presence of a concurrent malignancy (excluding basal cell carcinoma of the skin) were exclusion criteria. Patients were observed until death or the end of the study period when they were censored (April 9, 2012), those lost to follow-up were excluded from the analysis. Participants provided written informed consent and ethical approval was obtained for the research (National Research Ethics Service Committee North East—County Durham and Tees Valley 12/NE/0332).
This study was designed to test the hypotheses that patients with better physical function, less pain, and better global quality of life at baseline assessment will have improved survival.1,22
Clinical Assessment and Treatment Details Computed tomography (CT) and magnetic resonance imaging with liver-specific contrast were used to stage the hepatic metastases and pulmonary staging was undertaken with CT scan of the thorax. Patients with high recurrence risk disease were further staged using laparoscopy with or without laparoscopic ultrasound and selective use of CT positron emission tomography. All procedures were performed by specialist surgeons (F.K.W., T.J., and M.R.) with general anesthesia using standardized nomenclature of the International HepatoPancreato-Biliary Association and latterly facilitated by the use of surgical stapling devices.18 Patients were considered inoperable if the liver disease could not be safely excised in its entirety or extrahepatic disease was identified and these patients were excluded from the study.
2
-
Model-Building. For each model, time 0 was the date of curative surgery, which was within 4 weeks of the baseline PRO assessment. Cox proportional hazards regression23 was used to explore unadjusted associations between baseline PROs and survival (model 1). This model was adjusted (model 2) for age, primary colorectal tumor nodal status, differentiation status of the primary colorectal tumor, largest size of the liver metastases, total number of liver metastases, presence of extrahepatic locally resectable disease, carcinoembryonic antigen levels, and presence of a histologically involved tumor resection margin. The clinical covariates were identified according to a literature review24 and expert consultation and defined a priori. Changes of 10 points on a 0 to 100 scale were considered clinically relevant and hazard ratios (HRs) presented for a 10-point change in PRO score on a continuous scale.
Patient-Reported Outcomes
Model Simplification. Backwards stepwise Cox proportional hazards regression, with a significance level for variable removal of .125 to maximize data inclusion25 was undertaken. Prognostic PRO variables identified using this approach were incorporated into a final simplified model (model 3) after comparison of the three models constructed (Supplemental Table 1). In the final model, variables with a P < .01 were considered relevant to account for multiple testing and reduce the risk of overfitting of data.26 The proportional hazards assumption was assessed by examining proportional hazards plots and the corresponding global test.27
Patient-reported outcomes were assessed using the European Organization for Research and Treatment of Cancer (EORTC) core questionnaire (Quality of Life Questionnaire [QLQ]), the EORTC
Model Validation. To evaluate the performance of the selected final model, Harrell concordance c-statistic was used. This discrimination
Clinical Colorectal Cancer Month 2015
Jonathan R. Rees et al index provides a measure of the ability of a model to distinguish PROs that are prognostic to survival from those that are not prognostic. The index is also expressed in terms of the Somers d rank correlation where d ¼ 2c 1.28,29 When c ¼ 1 (or d ¼ 1) this implies perfect concordance between the predicted outcome from the model under test and the observed outcome from the data set. In contrast, if c ¼ 0.5 (or d ¼ 0), this indicates random discrimination and any associations between predicted outcomes from the prognostic model and observed outcomes from the data set are due to chance. Models derived from this structured approach were tested using a randomly generated test set and compared using linear combinations
of coefficients after estimation.30 Values of c and d indices are reported, where larger values of these indices (c and d) indicate better prediction of outcome by the selected model. Bootstrap methods with 1000 iterations were used to assess stability of the selected model where reported bootstrap inclusion fractions provided the frequency with which a particular variable is selected over the 1000 iterations.31 It was assumed that the observed data effectively represented the true population distribution and “random resampling with replacement” of the observed data set was undertaken using samples with the same number of observations as the original data set.
Table 1 Clinical and Sociodemographic Details of Patients Who Underwent Liver Resection for Colorectal Liver Metastases, Patients Who Survived in LTFU, Patients Who Did Not Have a Resection, and Patients Who Died Hepatic Resection of Colorectal Metastases (n [ 232)a
Hepatic Resection of Colorectal Metastases Alive at LTFU (n [ 101)
Deceased (n [ 131)
Mean Age, Years (SD)
63.1 (8.5)
71.0 (10.0)
70.1 (10.3)
Male Sex, n (%)
162 (69.8)
70 (69.3)
90 (68.7)
Less than compulsory/compulsory
109 (47.0)
49 (48.5)
60 (45.8)
Postcompulsory school education
118 (50.9)
50 (49.5)
68 (51.9)
5 (2.1)
2 (2.0)
3 (2.3)
Employed (full, part, self-employed)
106 (45.7)
43 (42.6)
63 (48.1)
Retired
118 (50.9)
56 (55.4)
62 (47.3)
8 (3.4)
2 (2.0)
6 (4.6)
40 (17.2)
13 (12.9)
27 (20.6)
192 (82.8)
88 (87.1)
104 (79.4)
Characteristic
Highest Level of Education, n (%)
Unknown Current Employment Status, n (%)
Other/unknown Cohabitation, n (%) Alone With family or other adults Unknown
0 (0)
0 (0)
0 (0)
Type of Previous Colorectal Surgery, n (%) Hemicolectomy Anterior resection
76 (32.8)
31 (30.7)
45 (34.4)
115 (49.6)
58 (57.4)
57 (43.5)
Abdominoperineal resection
11 (4.7)
Other/unknown
30 (12.9)
0 (0)
11 (8.4)
12 (11.9)
18 (13.7)
Stage of Colorectal Primary, n (%) I
12 (5.2)
5 (5.0)
7 (5.4)
II
37 (15.9)
16 (15.8)
21 (16.0)
III
67 (28.9)
29 (28.7)
38 (29.0)
IV
97 (41.8)
42 (41.6)
55 (42.0)
Unknown
19 (8.2)
9 (8.9)
10 (7.6)
Detection of Liver Metastases, n (%) Synchronous with primary tumor
54 (23.3)
20 (19.8)
34 (26.0)
Routine follow-up (asymptomatic)
143 (61.6)
65 (64.4)
78 (59.5)
Symptomatic
19 (8.2)
8 (7.9)
11 (8.4)
Other/unknown
16 (6.9)
8 (7.9)
8 (6.1)
ASA Grade, n (%) 1/2
195 (84.1)
87 (86.1)
108 (82.4)
3/4
32 (13.8)
12 (11.9)
20 (15.3)
5 (2.1)
2 (2.0)
3 (2.3)
Unknown
Abbreviations: ASA ¼ American Society of Anesthesiology; LTFU ¼ long-term follow-up. a For 9 patients who were planned for liver resection no surgery was undertaken because of extensive disease and the patients were excluded from the study.
Clinical Colorectal Cancer Month 2015
-3
Prognostic Role of PROs After Liver Surgery Results Patients Of the 241 patients enrolled, 9 had unresectable disease and were excluded, and 232 entered the study (Table 1). Baseline data were collected from all patients before surgery17 and 101 (43.5%) were alive at study censoring with a median follow-up of 7.5 years (range, 6.4-8.1 years). Clinical and demographic details of the 232 study participants are shown in Table 1.
Association Between Baseline PROs and Survival The unadjusted HR for the association between PROs and survival suggest that global quality of life, physical function, constipation, weight loss, and taste have potential prognostic relevance
(Table 2). The proportional hazards assumption was not violated by this model (c2 ¼ 13.74; P ¼ .98). A Cox proportional hazard model, adjusted for the prespecified clinical covariates (Table 3) demonstrated that a 10-point improvement in quality of life was associated with a 55% improvement in survival (HR, 0.45; 95% confidence interval [CI], 0.31-0.65; P < .001). Weaker associations were seen between survival and a number of symptoms; with less severe (by 10 points) constipation (HR, 0.80; 95% CI, 0.65-1.00; P ¼ .046) and diarrhea (HR, 0.79; 95% CI, 0.63-1.00; P ¼ .044) each associated with a 20% improvement in survival. In contrast, poorer (by 10 points) sexual function (HR, 1.16; 95% CI, 1.03-1.32; P ¼ .013) and taste (HR, 1.47; 95% CI, 1.06-2.05; P ¼ .02) were associated with
Table 2 Unadjusted HR With 95% CI for Associations Between a 10-Point Increase in PRO Scores From the EORTC QLQ-C30 and QLQ-LMC21 Questionnaires and Survival in 232 Patients Treated With Curative Surgery for CRC Liver Metastases (Model 1)
Variable
Mean PRO Score Before Surgery (95% CI)
Model 1 HR
95% CI
P
QLQ-C30 Scalesa Global health
76.3 (73.8-78.7)
0.69
(0.54-0.89)
.004
Physical function
88.0 (85.9-90.0)
1.41
(1.06-1.89)
.019
Role function
80.7 (77.2-84.2)
0.91
(0.76-1.07)
.255
Emotional function
79.8 (77.2-82.3)
1.16
(0.92-1.45)
.219
Cognitive function
87.2 (85.0-89.4)
1.10
(0.90-1.35)
.341
Social function
78.5 (75.0-81.9)
1.06
(0.88-1.28)
.544
QLQ-C30 Scales and Itemsb Fatigue Nausea and vomiting
21.8 (19.1-24.6)
1.10
(0.84-1.44)
.472
3.4 (9.6-14.8)
0.78
(0.54-1.12)
.178
Pain
12.2 (2.0-4.7)
1.08
(0.91-1.29)
.374
Dyspnoea
11.0 (8.3-13.6)
1.01
(0.88-1.16)
.912
Insomnia
22.3 (18.8-25.9)
1.01
(0.88-1.15)
.911
7.6 (5.1-10.2)
1.01
(0.72-1.40)
.974
Appetite loss Constipation
9.7 (7.4-12.1)
0.82
(0.70-0.96)
.016
Diarrhea
12.2 (9.3-15.0)
0.92
(0.79-1.07)
.268
Financial difficulty
13.5 (10.2-16.8)
0.94
(0.83-1.07)
.354
QLQ-LMC21 Scales and Itemsb Eating problems
7.6 (5.5-9.7)
1.14
(0.76-1.72)
.532
Activity/vigor problems
29.6 (26.3-33.0)
0.83
(0.65-1.06)
.128
Abdominal pain
14.4 (12.1-16.7)
1.07
(0.87-1.32)
.539
Anxiety problems
19.4 (15.7-23.2)
1.20
(0.97-1.50)
.095
Nutritional issues
11.6 (8.5-14.7)
1.51
(1.09-2.11)
.015
Taste problems
34.6 (31.5-37.7)
1.30
(1.09-1.55)
.004
6.8 (4.6-8.9)
0.98
(0.84-1.15)
.834
Sore mouth/tongue
8.5 (5.8-11.3)
1.06
(0.91-1.24)
.437
Tingling in fingers
15.8 (12.5-19.1)
1.00
(0.91-1.09)
.947
5.5 (3.4-7.6)
0.00
e
e
Contact with friends
10.2 (7.4-13.1)
0.94
(0.77-1.15)
.563
Talking about feelings
31.1 (26.5-37.7)
0.94
(0.79-1.12)
.488
Sexual function
30.6 (25.1-36.1)
1.06
(0.98-1.15)
.123
Dry mouth
Jaundice
Abbreviations: CRC ¼ colorectal cancer; EORTC ¼ European Organization for Research and Treatment of Cancer; HR ¼ hazard ratio; PRO ¼ patient-reported outcome. a Higher score indicates better function. b Higher score indicates worse symptom or problem.
4
-
Clinical Colorectal Cancer Month 2015
Jonathan R. Rees et al Table 3 Adjusteda HR With 95% CI for Model 1 for 10-Point Change in Baseline PRO Scores From the EORTC QLQ-C30 and QLQ-LMC21 Questionnaires and Survival in 232 Patients Treated With Curative Surgery for CRC Liver Metastases (Model 2)
Variable
Mean PRO Score Before Surgery (95% CI)
Model 2 HR
95% CI
P
QLQ-C30 Scales Global health
76.3 (73.8-78.7)
0.45
(0.31-0.65)
<.001
Physical function
88.0 (85.9-90.0)
1.14
(0.76-1.70)
.535
Role function
80.7 (77.2-84.2)
0.96
(0.77-1.19)
.684
Emotional function
79.8 (77.2-82.3)
1.29
(0.91-1.82)
.157
Cognitive function
87.2 (85.0-89.4)
1.26
(0.91-1.74)
.159
Social function
78.5 (75.0-81.9)
1.16
(0.88-1.53)
.292
QLQ-C30 Scales and Items Fatigue Nausea and vomiting
21.8 (19.1-24.6)
1.02
(0.71-1.47)
.898
3.4 (9.6-14.8)
0.82
(0.52-1.31)
.410
Pain
12.2 (2.0-4.7)
1.16
(0.89-1.51)
.271
Dyspnoea
11.0 (8.3-13.6)
0.97
(0.77-1.23)
.823
Insomnia
.212
22.3 (18.8-25.9)
0.89
(0.74-1.07)
Appetite loss
7.6 (5.1-10.2)
0.85
(0.53-1.38)
.518
Constipation
9.7 (7.4-12.1)
0.80
(0.65-1.00)
.046
Diarrhea
12.2 (9.3-15.0)
0.79
(0.63-1.00)
.044
Financial difficulty
13.5 (10.2-16.8)
0.99
(0.81-1.21)
.910
QLQ-LMC21 Scales and Items Eating problems
7.6 (5.5-9.7)
1.31
(0.75-2.29)
.349
Activity/vigor problems
29.6 (26.3-33.0)
0.72
(0.51-1.03)
.075
Abdominal pain
14.4 (12.1-16.7)
1.38
(0.95-2.00)
.095
Anxiety problems
19.4 (15.7-23.2)
1.19
(0.84-1.70)
.329
Nutritional issues
11.6 (8.5-14.7)
1.52
(0.87-2.66)
.141
Taste problems
.020
34.6 (31.5-37.7)
1.48
(1.06-2.05)
Dry mouth
6.8 (4.6-8.9)
0.90
(0.71-1.15)
.400
Sore mouth/tongue
8.5 (5.8-11.3)
1.18
(0.85-1.63)
.323
Tingling in fingers
15.8 (12.5-19.1)
1.04
(0.88-1.22)
.634
5.5 (3.4-7.6)
0.00
e
Contact with friends
10.2 (7.4-13.1)
0.93
(0.67-1.28)
.636
Talking about feelings
31.1 (26.5-37.7)
0.99
(0.76-1.28)
.915
Sexual function
30.6 (25.1-36.1)
1.17
(1.03-1.32)
.013
Jaundice
e
Abbreviations: CRC ¼ colorectal cancer; EORTC ¼ European Organization for Research and Treatment of Cancer; HR ¼ hazard ratio; PRO ¼ patient-reported outcome. a Adjusted for predefined clinical covariates.
poorer survival of 16% to 47%. There was no evidence to suggest that the model violated the proportional hazards assumption (c2 ¼ 36.32; P ¼ .407).
vigor; HR, 1.31; 95% CI, 1.15-1.45; P ¼ .001), which were associated with 30% to 38% greater mortality. There was no evidence to suggest that the model violated the proportional hazards assumption (c2 ¼ 10.17; P ¼ .86).
Final Adjusted and Simplified Model for Associations Between PROs and Survival (Model 3)
Model Validation
Hazard ratios from the stepwise backwards Cox proportional hazards regression are shown in Table 4. A 10-point higher global quality of life was associated with a 54% better survival (HR, 0.46; 95% CI, 0.33-0.63; P < .001) and a 10-point higher reported weight loss was associated with 75% greater mortality (HR, 1.75; 95% CI, 1.20-2.55; P ¼ .004). Smaller effects were seen for 10-point greater abdominal pain (HR, 1.38; 95% CI, 1.08-1.77; P ¼ .01), taste problems (HR, 1.30; 95% CI, 1.08-1.57; P ¼ .005), and fatigue (activity and/or
The c index value for the model incorporating solely PROs was 0.71; when all clinical and patient-reported covariates were incorporated the index increased to 0.79 and was reduced to 0.78 for the simplified model obtained using backward stepwise procedures (Supplemental Table 2 in the online version). Bootstrap resampling showed that global health (89.5% of samples), constipation (86.2% of samples), and weight loss (64.8% of samples) were reliably (> 60% of bootstrap samples) included in the internal validation (Table 4).
Clinical Colorectal Cancer Month 2015
-5
Prognostic Role of PROs After Liver Surgery Table 4 Simplified Model Incorporating Potentially Relevant Prognostic PROs With Adjusteda HRs Describing the Association of a 10Point (Clinically Relevant) Change in Baseline PRO Score and Survival in 232 Patients Treated With Curative Surgery for CRC Liver Metastases Proportion of Bootstrap Resamples With Effect at a Significance Level of £.05
Model 3 Variable
HR
95% CI
P
%
N
QLQ-C30 Scales Global health
0.46
(0.33-0.63)
<.001
89.5
895
Emotional function
1.20
(0.99-1.47)
.068
34.2
342
Constipation
0.75
(0.62-0.89)
.001
86.2
862
Diarrhea
0.88
(0.75-1.03)
.123
17.4
174
Insomnia
0.88
(0.75-1.03)
.110
18.0
180
QLQ-C30 Items
QLQ-LMC21 Scales and Items Weight loss
1.75
(1.20-2.55)
.004
64.8
648
Abdominal pain
1.38
(1.08-1.77)
.010
44.1
441
Taste
1.30
(1.08-1.57)
.005
56.3
563
Activity/vigor problems
1.31
(1.15-1.45)
.001
56.0
560
Sex function
1.13
(1.04-1.23)
.006
26.4
264
The proportion of bootstrap resamples (n ¼ 1000) for which prognostic PRO variables identified by the final model (model 3) were selected. Abbreviations: CRC ¼ colorectal cancer; HR ¼ hazard ratio; PRO ¼ patient-reported outcome. a Adjusted for predefined clinical covariates.
Discussion Results of this exploratory study suggest that differences in baseline PROs in selected domains including symptoms might be prognostic in patients who undergo surgery with curative intent for CRC liver metastases. Improved quality of life, less constipation, and less diarrhea were associated with better survival, and worse problems with weight loss, abdominal pain, and taste; fatigue (activity and vigor) and worse problems with sex life were associated with poorer survival. Internal validation using bootstrap resamples highlighted that global health (89.5% of samples); constipation (86.2% of samples), and weight loss (64.8% of samples) were the most stable predictors of outcome. These findings suggested that in addition to patient and tumor characteristics, self-reported measures of quality of life and symptoms have prognostic value. Therefore, clinicians should be alerted to patients who at baseline have poorer global quality of life scores or particular issues with constipation and weight loss because they might be at increased risk of disease recurrence and/or death and these are findings that should be considered when initial treatment and later follow-up are planned. The prognostic value of PROs in patients who undergo curative liver surgery for CRC has not previously been described, although results of a single study in patients with hepatitis B and unresectable or metastatic hepatocellular carcinoma10 have suggested that appetite loss, and physical and role function might be prognostic. However, these findings should be viewed with caution because they might not be directly relevant to patients with resectable CRC liver metastases. Global quality of life and weight loss have, however, been suggested to be prognostic in a number of other tumor types. Studies in localized esophagogastric cancer,32 lung cancer and lymphoma,6 but not localized breast cancer12 suggest a prognostic role for global quality of life, and in patients with metastatic disease, global quality of
6
-
Clinical Colorectal Cancer Month 2015
life might be prognostic in breast carcinoma,33 CRC,34 lung cancer,35 and prostate malignancy.9 In addition, weight loss has been associated with poor prognosis across a range of cancer patients who receive chemotherapy36 and results of specific work in metastatic lung cancer also suggest that this symptom has prognostic value.37 The reasons that PROs have prognostic significance are less clear. It is possible they might be more sensitive indicators of disease progression responding to subtle physiological changes perhaps driven by micrometastatic disease or early tumor recurrence. These changes, reflecting subclinical disease and hypothetically mediated by proinflammatory agents such as interleukin (IL) 2, IL-6, and tumor necrosis factor a might influence PROs before clinically detectable disseminated disease becomes apparent.38-40 Further possible explanations for the predictive value of PROs include the suggestion that patients who report better PROs might have more positive lifestyle behaviors. These individuals putatively have greater physiological reserve and therefore experience longer survival. In addition, it is suggested that patients who describe better PROs are recognized to have greater adherence to treatment regimens. The reasons are unclear but might reflect that these individuals have fewer comorbidities and fewer reactions to previous therapy, meaning that their exposure to potentially curative therapy is maximized and their survival potentially improved.1 Considering weight loss in cancer patients, possible etiologies are multiple; tumors might directly impede gut function or suppress appetite, and many therapies interfere with appetite, taste, and gut function, and are often emetogenic.37 Furthermore, systemic inflammation caused by tumors might also drive weight loss. Weight loss has also been associated with poorer survival but the mechanisms for this effect are also unclear although increases in proinflammatory cytokines and growth factors that drive metabolic
Jonathan R. Rees et al disturbance and lead to cachexia and tumor growth are thought to be important.40 In this study we described a large (n ¼ 232) well characterized cohort and used validated generic and disease PRO instruments in a prospective design. Missing PRO data were rare but participants in this study were from a single center, and those with unresectable liver disease did have a lower proportion of stage III and IV primary colorectal tumors although the absolute numbers were very small; therefore care must be exercised when considering the generalizability of these findings. The analysis approach used did result in multiple testing, however, this limitation might be offset by the application of a stringent significance level at P .01. Furthermore, model simplification methods resulted in a small decrease in the Harrell c statistic when the model incorporating all covariates considered clinically relevant (model 2) was compared with the final model (model 3) generated by the backward stepwise selection approach. This phenomenon reflected the necessary compromise struck between model simplicity (by reducing the number of variables) and prognostic ability. However the final model (model 3) did have sufficient events per covariate for unbiased estimation and accurate CI coverage probabilities, potentially offsetting the consequences of model simplification.41 Confounding might also arise from the more frequent future use of chemotherapy in this setting.
Conclusion Results of this study provide the first evidence to suggest that patients who report worse baseline global quality of life, more constipation, and greater problems with weight loss before potentially curative liver resection for CRC liver metastases have significantly poorer survival. External validation of these findings is necessary and must demonstrate that use of PRO data with current clinical predictors of outcome significantly increase the ability of the proposed model to predict outcome. If these criteria are achieved then the findings might be used in clinical practice to inform patients, influence treatment planning including surgery, neoadjuvant and/or adjuvant therapy, and supportive care, and also help determine follow-up strategies.
Clinical Practice Points The prognostic value of PROs in patients who undergo liver
resection for CRC metastases is unknown. Patients who report poorer global quality of life and greater
weight loss at baseline have worse long-term survival. Clinicians should counsel patients regarding the potential effect
that poorer global quality of life and greater weight loss have on survival. They should further ensure that psychological, physical, and optimal nutritional support is available to individuals who undergo liver resection for CRC metastases.
Disclosure The authors have stated that they have no conflicts of interest.
Supplemental Data Supplemental tables accompanying this article can be found in the online version at http://dx.doi.org/10.1016/j.clcc.2015.07.003.
References 1. Gotay CC, Kawamoto CT, Bottomley A, Efficace F. The prognostic significance of patient-reported outcomes in cancer clinical trials. J Clin Oncol 2008; 26:1355-63. 2. Blazeby JM, Brookes ST, Alderson D. Prognostic value of quality of life scores in patients with oesophageal cancer. Br J Surg 2000; 87:362-73. 3. Blazeby JM, Metcalfe C, Nicklin J, Barham CP, Donovan J, Alderson D. Association between quality of life scores and short-term outcome after surgery for cancer of the oesophagus or gastric cardia. Br J Surg 2005; 92:1502-7. 4. Barney BJ, Wang XS, Lu C, et al. Prognostic value of patient-reported symptom interference in patients with late-stage lung cancer. Qual Life Res 2013; 22: 2143-50. 5. Efficace F, Lo-Coco F. Using patient-reported health status to improve prognostic assessment in patients with acute myeloid leukemia: current challenges and future applications. Haematologica 2013; 98:7-9. 6. Jerkeman M, Kaasa S, Hjermstad M, Kvaloy S, Cavallin-Stahl E. Health-related quality of life and its potential prognostic implications in patients with aggressive lymphoma: a Nordic Lymphoma Group Trial. Med Oncol 2001; 18:85-94. 7. Dubois D, Dhawan R, van de Velde H, et al. Descriptive and prognostic value of patient-reported outcomes: the bortezomib experience in relapsed and refractory multiple myeloma. J Clin Oncol 2006; 24:976-82. 8. Efficace F, Bottomley A, Coens C, et al. Does a patient’s self-reported healthrelated quality of life predict survival beyond key biomedical data in advanced colorectal cancer? Eur J Cancer 2006; 42:42-9. 9. Sullivan PW, Nelson JB, Mulani PM, Sleep D. Quality of life as a potential predictor for morbidity and mortality in patients with metastatic hormonerefractory prostate cancer. Qual Life Res 2006; 15:1297-306. 10. Yeo W, Mo FK, Koh J, et al. Quality of life is predictive of survival in patients with unresectable hepatocellular carcinoma. Ann Oncol 2006; 17:1083-9. 11. Brown PD, Maurer MJ, Rummans TA, et al. A prospective study of quality of life in adults with newly diagnosed high-grade gliomas: the impact of the extent of resection on quality of life and survival. Neurosurgery 2005; 57:495-504, discussion 495e504. 12. Efficace F, Therasse P, Piccart MJ, et al. Health-related quality of life parameters as prognostic factors in a nonmetastatic breast cancer population: an international multicenter study. J Clin Oncol 2004; 22:3381-8. 13. Efficace F, Bottomley A, Smit EF, et al. Is a patient’s self-reported health-related quality of life a prognostic factor for survival in nonesmall-cell lung cancer patients? A multivariate analysis of prognostic factors of EORTC study 08975. Ann Oncol 2006; 17:1698-704. 14. Coates AS, Hurny C, Peterson HF, et al. Quality-of-life scores predict outcome in metastatic but not early breast cancer. International Breast Cancer Study Group. J Clin Oncol 2000; 18:3768-74. 15. Tross S, Herndon J 2nd, Korzun A, et al. Psychological symptoms and disease-free and overall survival in women with stage II breast cancer. Cancer and Leukemia Group B. J Natl Cancer Inst 1996; 88:661-7. 16. Nordlinger B, Sorbye H, Glimelius B, et al. Perioperative FOLFOX4 chemotherapy and surgery versus surgery alone for resectable liver metastases from colorectal cancer (EORTC 40983): long-term results of a randomised, controlled, phase 3 trial. Lancet Oncol 2013; 14:1208-15. 17. Rees J, Blazeby J, Fayers P, et al. Patient-reported outcomes after hepatic resection of colorectal cancer metastases. J Clin Oncol 2012; 30:1364-70. 18. Strasberg SM, Belghiti J, Clavien PA. Terminology of liver anatomy and resections. HPB (Oxford) 2000; 2:333-9. 19. Blazeby JM, Fayers P, Conroy T, et al. Validation of the European Organization for Research and Treatment of Cancer QLQ-LMC21 questionnaire for assessment of patient-reported outcomes during treatment of colorectal liver metastases. Br J Surg 2009; 96:291-8. 20. Kavadas V, Blazeby JM, Conroy T, et al. Development of an EORTC diseasespecific quality of life questionnaire for use in patients with liver metastases from colorectal cancer. Eur J Cancer 2003; 39:1259-63. 21. Fayers P, Aaronson N, Bjordal K, Groenvold M, Curran D, Bottomley A. EORTC QLQ-C30 Scoring Manual. 2nd ed. Brussels: EORTC Publications; 2002. 22. Quinten C, Martinelli F, Coens C, et al. A global analysis of multitrial data investigating quality of life and symptoms as prognostic factors for survival in different tumor sites. Cancer 2014; 120:302-11. 23. Kirkwood BR, Sterne JA. Essential Medical Statistics. 2nd ed. Malden, MA: Blackwell Science; 2003. 24. Rees M, Tekkis PP, Welsh FK, O’Rourke T, John TG. Evaluation of long-term survival after hepatic resection for metastatic colorectal cancer: a multifactorial model of 929 patients. Ann Surg 2008; 247:125-35. 25. Kaszycki P, Fedorovych D, Ksheminska H, Babyak L, Wojcik D, Koloczek H. Chromium accumulation by living yeast at various environmental conditions. Microbiol Res 2004; 159:11-7. 26. Babyak MA. What you see may not be what you get: a brief, nontechnical introduction to overfitting in regression-type models. Psychosom Med 2004; 66: 411-21. 27. Grambsch PM. TTM: proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994; 81:515-26. 28. Harrell FE Jr, Lee KL, Califf RM, Pryor DB, Rosati RA. Regression modelling strategies for improved prognostic prediction. Stat Med 1984; 3:143-52. 29. Somers RH. A new asymmetric measure of association for ordinal variables. Am Sociol Rev 1962; 27:799-811.
Clinical Colorectal Cancer Month 2015
-7
Prognostic Role of PROs After Liver Surgery 30. Bacon SL, Watkins LL, Babyak M, et al. Effects of daily stress on autonomic cardiac control in patients with coronary artery disease. Am J Cardiol 2004; 93:1292-4. 31. Royston P, Sauerbrei W. Multivariable Model-Building: A Pragmatic Approach to Regression Analysis Based on Fractional Polynomials for Modelling Continuous Variables. Chichester, England, Hoboken, NJ: John Wiley; 2008. 32. Chau I, Norman AR, Cunningham D, Waters JS, Oates J, Ross PJ. Multivariate prognostic factor analysis in locally advanced and metastatic esophago-gastric cancerepooled analysis from three multicenter, randomized, controlled trials using individual patient data. J Clin Oncol 2004; 22:2395-403. 33. Winer EP, Berry DA, Woolf S, et al. Failure of higher-dose paclitaxel to improve outcome in patients with metastatic breast cancer: cancer and leukemia group B trial 9342. J Clin Oncol 2004; 22:2061-8. 34. Maisey NR, Norman A, Watson M, Allen MJ, Hill ME, Cunningham D. Baseline quality of life predicts survival in patients with advanced colorectal cancer. Eur J Cancer 2002; 38:1351-7. 35. Maione P, Perrone F, Gallo C, et al. Pretreatment quality of life and functional status assessment significantly predict survival of elderly patients with advanced nonesmall-cell lung cancer receiving chemotherapy: a prognostic analysis of the
8
-
Clinical Colorectal Cancer Month 2015
36. 37.
38. 39. 40. 41.
multicenter Italian lung cancer in the elderly study. J Clin Oncol 2005; 23: 6865-72. Dewys WD, Begg C, Lavin PT, et al. Prognostic effect of weight loss prior to chemotherapy in cancer patients. Eastern Cooperative Oncology Group. Am J Med 1980; 69:491-7. Sánchez-Lara K, Turcott JG, Juárez E, et al. Association of nutrition parameters including bioelectrical impedance and systemic inflammatory response with quality of life and prognosis in patients with advanced non-small-cell lung cancer: a prospective study. Nutr Cancer 2012; 64:526-34. Blazeby JM, Brookes ST, Alderson D. The prognostic value of quality of life scores during treatment for oesophageal cancer. Gut 2001; 49:227-30. Djarv T, Metcalfe C, Avery KN, Lagergren P, Blazeby JM. Prognostic value of changes in health-related quality of life scores during curative treatment for esophagogastric cancer. J Clin Oncol 2010; 28:1666-70. Abramovitch R, Marikovsky M, Meir G, Neeman M. Stimulation of tumour growth by wound-derived growth factors. Br J Cancer 1999; 79:1392-8. Vittinghoff E, McCulloch CE. Relaxing the rule of ten events per variable in logistic and Cox regression. Am J Epidemiol 2007; 165:710-8.
Jonathan R. Rees et al Supplemental Table 1 Cox Regression Models for the Association of a 10-Point (Clinically Relevant) Change in Baseline PRO Score and Survival in 232 Patients Treated With Curative Surgery for CRC Liver Metastases Variable
Difference in Coefficients of HR (95% CI)
PROs and Clinical Covariates in Model 2 Compared With Prognostically Relevant PROs and Clinical Covariates in Model 3 PROs and Clinical Covariates in Model 2 Compared With PROs in Model 1 Prognostically Relevant PROs and Clinical Covariates in Model 3 Compared With PROs in Model 1
P
0.06 (0.06 to 0.18)
.354
0.04 (0.04 to 0.12) 0.01 (0.11 to 0.08)
.312 .760
Comparison of coefficient of HR using linear combinations of coefficients after estimation in a randomly generated test set comparing unadjusted (PROs: model 1), adjusted (PROs and clinical covariates: model 2) and simplified (prognostically relevant PROs and clinical covariates: model 3), suggesting no loss in model predictive utility. Abbreviations: CRC ¼ colorectal cancer; HR ¼ hazard ratio; PRO ¼ patient-reported outcome.
Supplemental Table 2 Somers D and Harrell C Statistics for Unadjusted (Model 1), Adjusted (Model 2), and Simplified (Model 3) Cox Regression Models for the Association of a 10-Point (Clinically Relevant) Change in Baseline PRO Score and Survival in 232 Patients Treated With Curative Surgery for CRC Liver Metastases Concordance Metric Variable PROs: Model 1 PROs and Clinical Covariates: Model 2 Prognostically Relevant PROs and Clinical Covariates: Model 3
Somers D
Harrell C
0.42 0.58 0.57
0.71 0.79 0.78
Abbreviations: CRC ¼ colorectal cancer; PRO ¼ patient-reported outcome.
Clinical Colorectal Cancer Month 2015
- 8.e1