- Email: [email protected]
90-Days mortality rate in patients treated within the context of a phase-I trial: How should we identify patients who should not go on trial?
EUROPEAN JOURNAL OF CANCER
4 4 ( 2 0 0 8 ) 1 5 3 6 –1 5 4 0
available at www.sciencedirect.com
journal homepage: www.ejconline.com
90-Days mortality rate in patients treated within the context of a phase-I trial: How should we identify patients who should not go on trial? 5 Hendrik-Tobias Arkenaua,c,*, David Olmosa,c,*, Joo Ern Anga, Jorge Barriusoa, Vasilios Karavasilisa, Sue Ashleyb, Johann de Bonoa, Ian Judsona, Stan Kayea a
Drug Development Unit, Royal Marsden Hospital and Institute of Cancer Research, Downs Road, SM2 5PT Sutton, United Kingdom Department of Computing and Statistics, Royal Marsden Hospital, United Kingdom
b
A R T I C L E I N F O
A B S T R A C T
Article history:
Background: The primary objectives of phase-I trials include the definition of drug toxicities
Received 7 April 2008
and the recommendation of phase-II doses. In order to safeguard the well-being of its partic-
Accepted 24 April 2008
ipants, a common inclusion criterion is that of life expectancy >3 months. However, previous
Available online 10 June 2008
studies have shown that about 20% of these patients do not survive beyond this time-point. Methods: We identified 97 patients who died within the first 90 days of treatment out of a total
Keywords:
of 654 consecutively treated phase-I patients, from June 2003 to June 2007. This cohort was
Phase-I trial
compared to a control group comprising 215 patients who lived >90 days on phase-I studies
90-Days mortality
and were treated from January 2005 to June 2006.
Survival
Results: In keeping with our recently reported phase-I survival risk score, multivariate anal-
Prognostic factors
ysis demonstrated that patients who died within the first 90 days had lower albumin (p = 0.010), greater number of metastatic sites (p = 0.00001) and higher frequency of elevated LDH (p = 0.0002). This analysis also showed that 86% of patients who died during the first 90 days had an increased risk score of 2/3 compared to 39% in the control group. Furthermore, three additional factors were identified, namely younger age (p = 0.024), higher white cell count (p = 0.028) and poorer ECOG PS (p = 0.012) but the addition of these did not improve the ability to predict 90-day mortality compared to the afore-mentioned risk score. Conclusions: There is good evidence that our easily derivable scoring system provides an objective method to identify patients with a very limited life expectancy in whom participation in phase-I trials should be carefully evaluated. 2008 Elsevier Ltd. All rights reserved.
1.
Introduction
Patients with advanced cancers commonly face the dilemma of not having any standard treatment option, and a minority of these with good performance status (PS) and adequate or5
gan function are sometimes offered treatment within the context of phase-I trials. Phase-I trials are designed primarily to evaluate the tolerability and toxicity profile of new therapies. In order to ensure safety and minimise risk, the generally accepted inclusion and exclusion criteria for these trials
Part of this work has been accepted as oral presentation for the 2008 Annual Meeting of American Society of Clinical Oncology. * Corresponding authors: Tel.: +44 (0) 208 6426011; fax: +44 (0) 208 6427979. E-mail addresses: [email protected] (H.-T. Arkenau), [email protected] (D. Olmos). c H.T.A. and D.O. contributed equally. 0959-8049/$ - see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.ejca.2008.04.017
EUROPEAN JOURNAL OF CANCER
include adequate organ function and reasonable PS. However, a common entry criterion is predicted life expectancy of greater than 3 months, and this is notoriously difficult to predict with a reliable degree of accuracy. There have been very few published studies exploring factors associated with clinical outcome, toxicity and prognosis in this context.1–4 To our knowledge, there has not been any published report specifically examining phase-I patients with very poor prognosis, e.g. death within the first 90 days of treatment. Hence, it is useful to identify characteristics of patients which might predict worse survival and therefore minimal benefit because the same attendant risk of toxicity affects all patients taking part in these toxicity-defining trials. In our previously reported analysis, we found that a significant fraction of patients died within the first 90 days of treatment (18%).4 Previous phase-I series have described a 3-month mortality rate ranging between 13% and 20%.5–7 The overwhelming majority of these deaths were attributed to the underlying malignancy and were not toxic-deaths. In fact, the toxic death rate in our recently reported analysis was 0.47% and in keeping with other reports where toxicdeath rates ranged between 0.49% and 0.54%.1–4 All these point to a strong rationale for identifying pre-study factors which could allow improved patient selection for phase-I clinical trials. We performed this retrospective analysis in all patients who were consecutively treated within phase-I trials at the Drug Development Unit, Royal Marsden Hospital NHS Foundation Trust, over a period of four years. The aim of this retrospective study was to identify factors associated with an overall survival of less than 90 days.
2.
Patients and methods
We identified 97 patients who died within the first 90 days of treatment out of a total of 654 consecutively treated phase-I patients, from June 2003 to June 2007. These patients received their treatments at the Drug Development Unit, Royal Marsden Hospital NHS Foundation Trust, London and Surrey, United Kingdom. All patients had objective evidence of disease progression prior to trial entry. Using a case–control study design, we compared our ‘90-day mortality group’, with a control group consisting of 215 pts out of the remaining 557 patients surviving >90 days. Patients in the control group were treated consecutively between January 2005 and June 2006, and complete data for analysis were available. The control group was comparable to the entire study population in terms of baseline characteristics such as age, sex or tumour groups. A diagram of the study design is represented in Fig. 1. We used several clinical parameters to compare both groups, including age, sex, ECOG performance status (PS), full blood count (haemoglobin, white blood cell count (WBC) and platelets), biochemistry (lactate dehydrogenase (LDH) and albumin), sites and number of metastasis and number of previous systemic treatment for cancer. Additionally, we applied our previously reported and prospectively validated prognostic score for patients treated in phase-I trials, based on parameters associated with poor OS (albumin (<35 g/L), LDH > UNL and >2 sites of metastasis), to assess its ability
4 4 ( 2 0 0 8 ) 1 5 3 6 –1 5 4 0
1537
Fig. 1 – Patients flowchart.
to distinguish between patients who die during the first 90 days of treatment and those who do not.8 The cut-off date for the present analysis was the 30th June 2007.
2.1.
Statistical considerations
The SPSS Programme (Version 15.0, Chicago, USA) was used for statistical analysis. Associations between baseline characteristics and the 90-day mortality group were established using Chi-square and Mann–Whitney U tests, and Spearman’s rank where appropriate. A logistic regression model was applied to define the baseline characteristics associated with 90-day mortality in the multivariate analysis. Additionally, we explored our previously defined prognostic score to predict 90-day mortality and assess whether other factors could improve the ability of this score to predict 90-day mortality. Therefore, dichotomous variables based on independent and significant factors from the multivariate analysis were added to our score and their abilities to predict 90-day mortality were evaluated using receiver operator (ROC) curves, and their areas under the curve (AUCs) were compared using the Chi-square test. All p-values presented were two-sided.
3.
Results
3.1.
Baseline patient characteristics and trials
The male/female ratio of the 90-day mortality group was 2.2:1 (67 male sand 30 females) and was similar to our control group (male/female ratio 2:1). In both cohorts, the median number of treatment lines prior to phase-I trial was 3 (range 0–9). Additionally, both groups comprised a heterogeneous group of cancers (90-day mortality group versus control group): urological (18% versus 37%), breast and gynaecological (21% versus 14%), gastrointestinal (27% versus 10%), thoracic tumours (7% versus 13%), sarcoma (9% versus 15%), melanoma (7% versus 5%) and others (13% versus 4%). The imbalances of tumour types between both groups were not relevant.
1538
EUROPEAN JOURNAL OF CANCER
In addition, 74% patients in 90-day mortality group and 65% in control group were treated within first time in human drug programmes. Finally, the toxic-death rate in the 90-day mortality group was 1% (1/97). The baseline characteristics are summarised in Table 1.
3.2.
Baseline prognosticators of 90-day mortality
The univariate analysis revealed that patients who died during the first 90-day were characterised by parameters including younger age, poorer PS, low haemoglobin and serum albumin levels, higher platelet and white cell counts, higher percentage of elevated serum LDH levels, liver and lung metastases and lower frequency of bone metastasis (Table 1). The multivariate analysis revealed that factors associated with shorter survival were lower serum albumin (median 30 g/L versus 34 g/L), elevated LDH (>UNL versus normal) and number of metastatic sites (median 3 versus 2). Additionally, three other independent prognostic factors were identi-
4 4 ( 2 0 0 8 ) 1 5 3 6 –1 5 4 0
fied including younger age (median 55 versus 59 years), poorer PS (PS 2 versus PS 1 versus PS 0) and increased WBC (median 8.5 versus 6.8 · 103/mm3) (Table 2).
3.3.
Prognostic score for 90-day death
We applied our previously reported and prospectively validated survival score for phase-I patients and we could demonstrate that patients who died within the first 90 days had a significantly higher risk score compared to our control group (score 0: 1% versus 20%; 1: 13% versus 41%; 2: 40% versus 30%; 3: 45% versus 9%), p < 0.1 · 10)9 (Table 3). In addition, the AUC in predicting 90-day mortality in this series was 79.2% (confidence interval (CI) – 95% 74.0–84.4). The ability to predict the 90-day mortality rate did not improve significantly by the addition of the significant factors borne out of the multivariate analysis, namely age <55 years (AUC 78.9%, CI – 95% 73.7–84.1), PS P 1 (AUC 81.2%, CI – 95% 76.2–86.1) and white cell count >8.5 · 103/mL (AUC 77.8%, CI
Table 1 – Baseline characteristics: 90-days mortality group and control group Baseline characteristics
Mortality group 90-days
Control group
p-Value
Age Median (range)
55 years
(18–85)
59 years
(22–74)
0.005
Sex Male Female
67 30
69.1% 30.9%
140 75
65.1% 34.9%
NS
Previous systemic treatment Median (range)
3
(0–8)
3
(0–9)
NS
Performance status ECOG 0 ECOG 1 ECOG 2
6 74 16
6.3% 77.1% 16.7%
69 134 12
32.1% 62.3% 5.6%
<0.001
Haemoglobin (HGB) Median (range)
11.4 g/dL
(8.4–15.5)
12.1 g/dL
(8.8–16.0)
<0.001
Albumin Median (range)
30.0 g/L
(18.0–42.0)
34.0 g/L
(18.0–44.0)
<0.001
White blood cells count Median (range)
8.5 · 103/mm3
(3.5–46.9)
6.8 · 103/mm3
(2.9–21.2)
<0.001
Platelets count Median (range)
317 · 103/mm3
(105–797)
317 · 103/mm3
(90–784)
NS
Lactate dehydrogenase (LDH) Normal range Elevated value
23 74
23.7% 76.3%
113 102
52.6% 47.4%
<0.001
Number of metastatic sites Median (range)
3
(0–6)
2
(0–5)
<0.001
Lung metastasis Absent Present
37 60
38.1% 61.9%
125 90
58.1% 41.9%
0.001
Liver metastasis Absent Present
54 43
55.7% 44.3%
166 49
77.2% 22.8%
<0.001
Bone metastasis Absent Present
77 20
79.4% 20.6%
135 80
62.8% 37.2%
0.004
p-Values for univariate analysis.
EUROPEAN JOURNAL OF CANCER
Table 2 – Independent significant baseline characteristics associated with death during the first 90-days within a phase-I trial Risk factors for 90-days mortality in MVA
p-Value
Factor (median or %)
90-Days
Control
Age, years (median) Albumin, g/L (median) Number of metastatic sites (median) ECOG PS (%) PS 2 PS 1 PS 0 WBC, 103/mm3 (median) Elevated LDH (%)
55 30 3
59 34 2
0.024 0.012 0.010
17 77 6 8.5 76
6 62 32 6.8 47
0.028
0.00001 0.0002
MVA, multivariate-analysis.
Table 3 – Patients by group and risk-score, Chi-square test p-value <0.001 Group
Score 0
90-Days mortality Control
Score 1
Score 2
Score 3
1 (1%)
13 (13.4%)
39 (40.2%)
44 (45.4%)
42 (19.5%)
89 (41.4%)
64 (29.8%)
20 (9.3%)
4,8
Risk score : albumin >35 g/L (0) versus albumin <35 g/L (+1); LDH normal (0) versus LDH >UNL (+1); number of metastatic sites <2 (0) versus >2 (+1).
– 95% 72.5–84.1); there were not statistical differences between AUCs for these four different ROC curves.
4.
Discussion
The difficulty in weighing up the balance between risks and benefits has pervaded ethical considerations surrounding phase-I trials. Whilst the clinical benefit derived from participation in these trials is usually small, toxicities from these investigational agents and the considerable time commitments required of participants are significant.9,10 In particular, patients with very poor prognosis who take part in these studies have their risk/benefit profiles tipped in favour of the former and having more objective and evidence-based means of patient selection would represent a major step forward in trying to reduce risks to patients. Currently, predictions are based on subjective clinical judgements which comprise the natural history and extent of the disease, patient symptoms and/or the previous treatment history. Evidence points to the lack of objective and evidence-based criteria being a major cause of significant mortality rate by 3 months; the toxic-death rates reported in these studies are very low, implying that the vast majority of deaths were attributable to cancer-related complications. In our analysis, patients who died within the first 90 days of treatment had significantly higher LDH, white cell counts and number of metastatic sites, poorer PS and lower albumin levels compared to the control group. Although similar results
4 4 ( 2 0 0 8 ) 1 5 3 6 –1 5 4 0
1539
were demonstrated in various studies, major criticisms include: firstly, these included small patient numbers, and secondly, data were obtained retrospectively over long periods of time, sometimes >10 years. For example, one phase-I study which included 70 patients revealed that all patients who presented with both of the following two risk factors, albumin< 38 g/L and lymphocyte count <0.7 · 109/L, died within 90 days.11 Another study involving 154 patients over an eightyear period identified two independent risk factors, namely LDH >600 U/I and PS > 1, which were correlated with a shorter OS for patients with both factors.5 Furthermore, a study which included 420 patients over a period of 10 years found following parameters: PS > 1, LDH > UNL, white cell count > UNL, metastatic sites > 2 and haemoglobin < 12 g/dl.6 The authors recommended that patients with these risk factors should be excluded from phase-I trial participation. Prognostic scores and nomograms that are otherwise widely available for specific tumour types at the point of diagnosis or prior to receiving standard treatments are lacking in the context of phase-I trials. In addition, there is a uniqueness common amongst phase-I patients which is poorly defined. Hence, although the heterogeneity of tumour types in this study may be a confounder factor in our statistical analysis, the prognostic markers derived from this study are relevant and independent of tumour origin or histology. In this way, we have demonstrated that nearly 90% of patients who died during the first 90 days had an increased risk score of 2 or 3 compared to 39% in the control group. Additional factors such as younger age, high WBC count and poor PS had a negative impact on survival in our multivariate analysis. However, the inclusion of these factors to our recently validated prognostic score did not significantly improve the ability to predict 90-day survival. Nonetheless, these additional variables might further help in the selection of appropriate candidates for phase-I trial entry, and the role of these factors needs to be addressed in prospective studies. In summary, this easily derivable scoring system could aid in the complex decision making process surrounding phase-I patient selection by the provision of clear, transparent and objective measures through the early identification of a stratum of poor prognostic patient group and the avoidance of doing more harm than good to them, in accordance with the classical aphorism: primum non-nocere.
Conflict of interest statement None declared.
R E F E R E N C E S
1. Roberts TG, Goulart BH, Squitieri L, et al. Trends in the risk and benefits to patients with cancer participating in phase-I clinical trials. JAMA 2004;292:2130–40. 2. Horstmann E, McCabe MS, Grochow L, et al. Risks and benefits of phase-1 oncology trials, 1991 through 2002. New Engl J Med 2005;352:895–904. 3. Sekine I, Yamamoto N, Kunitoh H, et al. Relationship between objective responses in phase-I trials and potential efficacy of
1540
EUROPEAN JOURNAL OF CANCER
non-specific cytotoxic investigational new drugs. Ann Oncol 2002;13:1300–6. 4. Arkenau HT, Olmos D, Ang JE, de Bono J, Judson I, Kaye S. Clinical outcome and prognostic factors for patients treated within the context of a phase-I study: The Royal Marsden Hospital experience. Br J Cancer 2008;98:1029–33. 5. Bachelot T, Ray-Coquard I, Catimel G, et al. Multivariable analysis of prognostic factors for toxicity and survival for patients enrolled in phase-I clinical trials. Ann Oncol 2000;11:151–6. 6. Han C, Braybrooke JP, Deplanque G, et al. Comparison of prognostic factors in patients in phase I trials of cytotoxic drugs vs new noncytotoxic agents. Br J Cancer 2003;89:1166– 71.
4 4 ( 2 0 0 8 ) 1 5 3 6 –1 5 4 0
7. Yamamoto N, Tamura T, Fukuoka M, Saijo N. Survival and prognostic factors in lung cancer patients treated in phase-I trials: Japanese experience. Int J Oncol 1999;15:737–74. 8. Arkenau HT, Barriuso J, Olmos D et al. Prospective validation of a prognostic score to improve patient selection for phase-I trials. In: Proc. ASCO; 2008 [abstr 2508]. 9. Seidenfeld J, Horstmann E, Emanuel EJ, Grady C. Participants in phase-I oncology research trials: are they vulnerable? Arch Intern Med 2008;168:16–20. 10. Connors T. Anticancer drug development: the way forward. Oncologist 1996;1:180–1. 11. Penel N, Vanseymortier M, Bonneterre ME, et al. Prognostic factors among cancer patients with good performance status screened for phase I trials. Invest New Drugs 2008;26(1):53–8.
4 4 ( 2 0 0 8 ) 1 5 3 6 –1 5 4 0
available at www.sciencedirect.com
journal homepage: www.ejconline.com
90-Days mortality rate in patients treated within the context of a phase-I trial: How should we identify patients who should not go on trial? 5 Hendrik-Tobias Arkenaua,c,*, David Olmosa,c,*, Joo Ern Anga, Jorge Barriusoa, Vasilios Karavasilisa, Sue Ashleyb, Johann de Bonoa, Ian Judsona, Stan Kayea a
Drug Development Unit, Royal Marsden Hospital and Institute of Cancer Research, Downs Road, SM2 5PT Sutton, United Kingdom Department of Computing and Statistics, Royal Marsden Hospital, United Kingdom
b
A R T I C L E I N F O
A B S T R A C T
Article history:
Background: The primary objectives of phase-I trials include the definition of drug toxicities
Received 7 April 2008
and the recommendation of phase-II doses. In order to safeguard the well-being of its partic-
Accepted 24 April 2008
ipants, a common inclusion criterion is that of life expectancy >3 months. However, previous
Available online 10 June 2008
studies have shown that about 20% of these patients do not survive beyond this time-point. Methods: We identified 97 patients who died within the first 90 days of treatment out of a total
Keywords:
of 654 consecutively treated phase-I patients, from June 2003 to June 2007. This cohort was
Phase-I trial
compared to a control group comprising 215 patients who lived >90 days on phase-I studies
90-Days mortality
and were treated from January 2005 to June 2006.
Survival
Results: In keeping with our recently reported phase-I survival risk score, multivariate anal-
Prognostic factors
ysis demonstrated that patients who died within the first 90 days had lower albumin (p = 0.010), greater number of metastatic sites (p = 0.00001) and higher frequency of elevated LDH (p = 0.0002). This analysis also showed that 86% of patients who died during the first 90 days had an increased risk score of 2/3 compared to 39% in the control group. Furthermore, three additional factors were identified, namely younger age (p = 0.024), higher white cell count (p = 0.028) and poorer ECOG PS (p = 0.012) but the addition of these did not improve the ability to predict 90-day mortality compared to the afore-mentioned risk score. Conclusions: There is good evidence that our easily derivable scoring system provides an objective method to identify patients with a very limited life expectancy in whom participation in phase-I trials should be carefully evaluated. 2008 Elsevier Ltd. All rights reserved.
1.
Introduction
Patients with advanced cancers commonly face the dilemma of not having any standard treatment option, and a minority of these with good performance status (PS) and adequate or5
gan function are sometimes offered treatment within the context of phase-I trials. Phase-I trials are designed primarily to evaluate the tolerability and toxicity profile of new therapies. In order to ensure safety and minimise risk, the generally accepted inclusion and exclusion criteria for these trials
Part of this work has been accepted as oral presentation for the 2008 Annual Meeting of American Society of Clinical Oncology. * Corresponding authors: Tel.: +44 (0) 208 6426011; fax: +44 (0) 208 6427979. E-mail addresses: [email protected] (H.-T. Arkenau), [email protected] (D. Olmos). c H.T.A. and D.O. contributed equally. 0959-8049/$ - see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.ejca.2008.04.017
EUROPEAN JOURNAL OF CANCER
include adequate organ function and reasonable PS. However, a common entry criterion is predicted life expectancy of greater than 3 months, and this is notoriously difficult to predict with a reliable degree of accuracy. There have been very few published studies exploring factors associated with clinical outcome, toxicity and prognosis in this context.1–4 To our knowledge, there has not been any published report specifically examining phase-I patients with very poor prognosis, e.g. death within the first 90 days of treatment. Hence, it is useful to identify characteristics of patients which might predict worse survival and therefore minimal benefit because the same attendant risk of toxicity affects all patients taking part in these toxicity-defining trials. In our previously reported analysis, we found that a significant fraction of patients died within the first 90 days of treatment (18%).4 Previous phase-I series have described a 3-month mortality rate ranging between 13% and 20%.5–7 The overwhelming majority of these deaths were attributed to the underlying malignancy and were not toxic-deaths. In fact, the toxic death rate in our recently reported analysis was 0.47% and in keeping with other reports where toxicdeath rates ranged between 0.49% and 0.54%.1–4 All these point to a strong rationale for identifying pre-study factors which could allow improved patient selection for phase-I clinical trials. We performed this retrospective analysis in all patients who were consecutively treated within phase-I trials at the Drug Development Unit, Royal Marsden Hospital NHS Foundation Trust, over a period of four years. The aim of this retrospective study was to identify factors associated with an overall survival of less than 90 days.
2.
Patients and methods
We identified 97 patients who died within the first 90 days of treatment out of a total of 654 consecutively treated phase-I patients, from June 2003 to June 2007. These patients received their treatments at the Drug Development Unit, Royal Marsden Hospital NHS Foundation Trust, London and Surrey, United Kingdom. All patients had objective evidence of disease progression prior to trial entry. Using a case–control study design, we compared our ‘90-day mortality group’, with a control group consisting of 215 pts out of the remaining 557 patients surviving >90 days. Patients in the control group were treated consecutively between January 2005 and June 2006, and complete data for analysis were available. The control group was comparable to the entire study population in terms of baseline characteristics such as age, sex or tumour groups. A diagram of the study design is represented in Fig. 1. We used several clinical parameters to compare both groups, including age, sex, ECOG performance status (PS), full blood count (haemoglobin, white blood cell count (WBC) and platelets), biochemistry (lactate dehydrogenase (LDH) and albumin), sites and number of metastasis and number of previous systemic treatment for cancer. Additionally, we applied our previously reported and prospectively validated prognostic score for patients treated in phase-I trials, based on parameters associated with poor OS (albumin (<35 g/L), LDH > UNL and >2 sites of metastasis), to assess its ability
4 4 ( 2 0 0 8 ) 1 5 3 6 –1 5 4 0
1537
Fig. 1 – Patients flowchart.
to distinguish between patients who die during the first 90 days of treatment and those who do not.8 The cut-off date for the present analysis was the 30th June 2007.
2.1.
Statistical considerations
The SPSS Programme (Version 15.0, Chicago, USA) was used for statistical analysis. Associations between baseline characteristics and the 90-day mortality group were established using Chi-square and Mann–Whitney U tests, and Spearman’s rank where appropriate. A logistic regression model was applied to define the baseline characteristics associated with 90-day mortality in the multivariate analysis. Additionally, we explored our previously defined prognostic score to predict 90-day mortality and assess whether other factors could improve the ability of this score to predict 90-day mortality. Therefore, dichotomous variables based on independent and significant factors from the multivariate analysis were added to our score and their abilities to predict 90-day mortality were evaluated using receiver operator (ROC) curves, and their areas under the curve (AUCs) were compared using the Chi-square test. All p-values presented were two-sided.
3.
Results
3.1.
Baseline patient characteristics and trials
The male/female ratio of the 90-day mortality group was 2.2:1 (67 male sand 30 females) and was similar to our control group (male/female ratio 2:1). In both cohorts, the median number of treatment lines prior to phase-I trial was 3 (range 0–9). Additionally, both groups comprised a heterogeneous group of cancers (90-day mortality group versus control group): urological (18% versus 37%), breast and gynaecological (21% versus 14%), gastrointestinal (27% versus 10%), thoracic tumours (7% versus 13%), sarcoma (9% versus 15%), melanoma (7% versus 5%) and others (13% versus 4%). The imbalances of tumour types between both groups were not relevant.
1538
EUROPEAN JOURNAL OF CANCER
In addition, 74% patients in 90-day mortality group and 65% in control group were treated within first time in human drug programmes. Finally, the toxic-death rate in the 90-day mortality group was 1% (1/97). The baseline characteristics are summarised in Table 1.
3.2.
Baseline prognosticators of 90-day mortality
The univariate analysis revealed that patients who died during the first 90-day were characterised by parameters including younger age, poorer PS, low haemoglobin and serum albumin levels, higher platelet and white cell counts, higher percentage of elevated serum LDH levels, liver and lung metastases and lower frequency of bone metastasis (Table 1). The multivariate analysis revealed that factors associated with shorter survival were lower serum albumin (median 30 g/L versus 34 g/L), elevated LDH (>UNL versus normal) and number of metastatic sites (median 3 versus 2). Additionally, three other independent prognostic factors were identi-
4 4 ( 2 0 0 8 ) 1 5 3 6 –1 5 4 0
fied including younger age (median 55 versus 59 years), poorer PS (PS 2 versus PS 1 versus PS 0) and increased WBC (median 8.5 versus 6.8 · 103/mm3) (Table 2).
3.3.
Prognostic score for 90-day death
We applied our previously reported and prospectively validated survival score for phase-I patients and we could demonstrate that patients who died within the first 90 days had a significantly higher risk score compared to our control group (score 0: 1% versus 20%; 1: 13% versus 41%; 2: 40% versus 30%; 3: 45% versus 9%), p < 0.1 · 10)9 (Table 3). In addition, the AUC in predicting 90-day mortality in this series was 79.2% (confidence interval (CI) – 95% 74.0–84.4). The ability to predict the 90-day mortality rate did not improve significantly by the addition of the significant factors borne out of the multivariate analysis, namely age <55 years (AUC 78.9%, CI – 95% 73.7–84.1), PS P 1 (AUC 81.2%, CI – 95% 76.2–86.1) and white cell count >8.5 · 103/mL (AUC 77.8%, CI
Table 1 – Baseline characteristics: 90-days mortality group and control group Baseline characteristics
Mortality group 90-days
Control group
p-Value
Age Median (range)
55 years
(18–85)
59 years
(22–74)
0.005
Sex Male Female
67 30
69.1% 30.9%
140 75
65.1% 34.9%
NS
Previous systemic treatment Median (range)
3
(0–8)
3
(0–9)
NS
Performance status ECOG 0 ECOG 1 ECOG 2
6 74 16
6.3% 77.1% 16.7%
69 134 12
32.1% 62.3% 5.6%
<0.001
Haemoglobin (HGB) Median (range)
11.4 g/dL
(8.4–15.5)
12.1 g/dL
(8.8–16.0)
<0.001
Albumin Median (range)
30.0 g/L
(18.0–42.0)
34.0 g/L
(18.0–44.0)
<0.001
White blood cells count Median (range)
8.5 · 103/mm3
(3.5–46.9)
6.8 · 103/mm3
(2.9–21.2)
<0.001
Platelets count Median (range)
317 · 103/mm3
(105–797)
317 · 103/mm3
(90–784)
NS
Lactate dehydrogenase (LDH) Normal range Elevated value
23 74
23.7% 76.3%
113 102
52.6% 47.4%
<0.001
Number of metastatic sites Median (range)
3
(0–6)
2
(0–5)
<0.001
Lung metastasis Absent Present
37 60
38.1% 61.9%
125 90
58.1% 41.9%
0.001
Liver metastasis Absent Present
54 43
55.7% 44.3%
166 49
77.2% 22.8%
<0.001
Bone metastasis Absent Present
77 20
79.4% 20.6%
135 80
62.8% 37.2%
0.004
p-Values for univariate analysis.
EUROPEAN JOURNAL OF CANCER
Table 2 – Independent significant baseline characteristics associated with death during the first 90-days within a phase-I trial Risk factors for 90-days mortality in MVA
p-Value
Factor (median or %)
90-Days
Control
Age, years (median) Albumin, g/L (median) Number of metastatic sites (median) ECOG PS (%) PS 2 PS 1 PS 0 WBC, 103/mm3 (median) Elevated LDH (%)
55 30 3
59 34 2
0.024 0.012 0.010
17 77 6 8.5 76
6 62 32 6.8 47
0.028
0.00001 0.0002
MVA, multivariate-analysis.
Table 3 – Patients by group and risk-score, Chi-square test p-value <0.001 Group
Score 0
90-Days mortality Control
Score 1
Score 2
Score 3
1 (1%)
13 (13.4%)
39 (40.2%)
44 (45.4%)
42 (19.5%)
89 (41.4%)
64 (29.8%)
20 (9.3%)
4,8
Risk score : albumin >35 g/L (0) versus albumin <35 g/L (+1); LDH normal (0) versus LDH >UNL (+1); number of metastatic sites <2 (0) versus >2 (+1).
– 95% 72.5–84.1); there were not statistical differences between AUCs for these four different ROC curves.
4.
Discussion
The difficulty in weighing up the balance between risks and benefits has pervaded ethical considerations surrounding phase-I trials. Whilst the clinical benefit derived from participation in these trials is usually small, toxicities from these investigational agents and the considerable time commitments required of participants are significant.9,10 In particular, patients with very poor prognosis who take part in these studies have their risk/benefit profiles tipped in favour of the former and having more objective and evidence-based means of patient selection would represent a major step forward in trying to reduce risks to patients. Currently, predictions are based on subjective clinical judgements which comprise the natural history and extent of the disease, patient symptoms and/or the previous treatment history. Evidence points to the lack of objective and evidence-based criteria being a major cause of significant mortality rate by 3 months; the toxic-death rates reported in these studies are very low, implying that the vast majority of deaths were attributable to cancer-related complications. In our analysis, patients who died within the first 90 days of treatment had significantly higher LDH, white cell counts and number of metastatic sites, poorer PS and lower albumin levels compared to the control group. Although similar results
4 4 ( 2 0 0 8 ) 1 5 3 6 –1 5 4 0
1539
were demonstrated in various studies, major criticisms include: firstly, these included small patient numbers, and secondly, data were obtained retrospectively over long periods of time, sometimes >10 years. For example, one phase-I study which included 70 patients revealed that all patients who presented with both of the following two risk factors, albumin< 38 g/L and lymphocyte count <0.7 · 109/L, died within 90 days.11 Another study involving 154 patients over an eightyear period identified two independent risk factors, namely LDH >600 U/I and PS > 1, which were correlated with a shorter OS for patients with both factors.5 Furthermore, a study which included 420 patients over a period of 10 years found following parameters: PS > 1, LDH > UNL, white cell count > UNL, metastatic sites > 2 and haemoglobin < 12 g/dl.6 The authors recommended that patients with these risk factors should be excluded from phase-I trial participation. Prognostic scores and nomograms that are otherwise widely available for specific tumour types at the point of diagnosis or prior to receiving standard treatments are lacking in the context of phase-I trials. In addition, there is a uniqueness common amongst phase-I patients which is poorly defined. Hence, although the heterogeneity of tumour types in this study may be a confounder factor in our statistical analysis, the prognostic markers derived from this study are relevant and independent of tumour origin or histology. In this way, we have demonstrated that nearly 90% of patients who died during the first 90 days had an increased risk score of 2 or 3 compared to 39% in the control group. Additional factors such as younger age, high WBC count and poor PS had a negative impact on survival in our multivariate analysis. However, the inclusion of these factors to our recently validated prognostic score did not significantly improve the ability to predict 90-day survival. Nonetheless, these additional variables might further help in the selection of appropriate candidates for phase-I trial entry, and the role of these factors needs to be addressed in prospective studies. In summary, this easily derivable scoring system could aid in the complex decision making process surrounding phase-I patient selection by the provision of clear, transparent and objective measures through the early identification of a stratum of poor prognostic patient group and the avoidance of doing more harm than good to them, in accordance with the classical aphorism: primum non-nocere.
Conflict of interest statement None declared.
R E F E R E N C E S
1. Roberts TG, Goulart BH, Squitieri L, et al. Trends in the risk and benefits to patients with cancer participating in phase-I clinical trials. JAMA 2004;292:2130–40. 2. Horstmann E, McCabe MS, Grochow L, et al. Risks and benefits of phase-1 oncology trials, 1991 through 2002. New Engl J Med 2005;352:895–904. 3. Sekine I, Yamamoto N, Kunitoh H, et al. Relationship between objective responses in phase-I trials and potential efficacy of
1540
EUROPEAN JOURNAL OF CANCER
non-specific cytotoxic investigational new drugs. Ann Oncol 2002;13:1300–6. 4. Arkenau HT, Olmos D, Ang JE, de Bono J, Judson I, Kaye S. Clinical outcome and prognostic factors for patients treated within the context of a phase-I study: The Royal Marsden Hospital experience. Br J Cancer 2008;98:1029–33. 5. Bachelot T, Ray-Coquard I, Catimel G, et al. Multivariable analysis of prognostic factors for toxicity and survival for patients enrolled in phase-I clinical trials. Ann Oncol 2000;11:151–6. 6. Han C, Braybrooke JP, Deplanque G, et al. Comparison of prognostic factors in patients in phase I trials of cytotoxic drugs vs new noncytotoxic agents. Br J Cancer 2003;89:1166– 71.
4 4 ( 2 0 0 8 ) 1 5 3 6 –1 5 4 0
7. Yamamoto N, Tamura T, Fukuoka M, Saijo N. Survival and prognostic factors in lung cancer patients treated in phase-I trials: Japanese experience. Int J Oncol 1999;15:737–74. 8. Arkenau HT, Barriuso J, Olmos D et al. Prospective validation of a prognostic score to improve patient selection for phase-I trials. In: Proc. ASCO; 2008 [abstr 2508]. 9. Seidenfeld J, Horstmann E, Emanuel EJ, Grady C. Participants in phase-I oncology research trials: are they vulnerable? Arch Intern Med 2008;168:16–20. 10. Connors T. Anticancer drug development: the way forward. Oncologist 1996;1:180–1. 11. Penel N, Vanseymortier M, Bonneterre ME, et al. Prognostic factors among cancer patients with good performance status screened for phase I trials. Invest New Drugs 2008;26(1):53–8.