Validation and comparison of several published prognostic systems for patients with small cell lung cancer

Download figure
Open in new tab
Download powerpoint
Figure 1–
Survival curves according to the European Lung Cancer Working Party prognostic classification.

Table 1– Patient characteristics

Characteristic	Trial
Characteristic	ELCWP 1922 [9]	ELCWP 1923 [8]	ELCWP 1994	All ELCWP trials	Paesmans et al. [4]	Sculier et al. [2]
Subjects	204	233	256	693	763	6609
Sex
Male	163 (80)	196 (84)	211 (82)	570 (82)	689 (90)	4368 (66)
Female	41 (20)	37 (16)	45 (18)	123 (18)	74 (10)	2241 (34)
Age yrs	59 (32–75)	61 (34–74)	61 (40–83)	60 (32–84)	NR	NR
Karnofsky performance status
60–70	14 (7)	57 (24)	64 (25)	135 (19)	250 (33)	1409 (21)
80–90	190 (93)	176 (76)	192 (75)	558 (81)	513 (67)	5200 (79)
100	88 (43)	41 (18)	31 (12)	160 (23)	NR	2039 (31)
Disease extent
Limited	204 (100)	18 (8)	16 (6)	238 (34)	365 (48)	2870 (43)
Very limited	204	1	0	205	NR	NR
Extensive		215 (92)	240 (94)	455 (66)	398 (52)	3739 (57)
>2 metastatic sites		30	53	83	NR	NR
Pleural effusion
Not documented	204 (100)	200 (86)	219 (86)	623 (90)	NR	NR
Yes		33 (14)	37 (14)	70 (10)
White blood cell count
<10 IU	152 (75)	155 (67)	153 (60)	460 (66)	516	NR
≥10 IU	52 (25)	78 (33)	100 (39)	230 (33)	241
Not reported			3 (1)	3 (<1)	6
Relative neutrophil count
≤75%	166 (81)	154 (66)	152 (59)	472 (68)	528 (69)	NR
>75%	35 (17)	74 (32)	68 (27)	177 (26)	212 (28)
NR	3 (1)	5 (2)	36 (14)	44 (6)	23 (3)
Alkaline phosphatases
Normal	101 (50)	69 (30)	107 (42)	277 (40)	297	NR
Abnormal	77 (38)	158 (68)	144 (56)	379 (55)	431
NR	26 (13)	6 (3)	5 (2)	37 (5)	35
Creatinine
Normal	199 (98)	225 (97)	249 (97)	673 (97)	NR	NR
Abnormal	3 (1)	6 (3)	4 (2)	13 (2)
NR	2 (1)	2 (1)	3 (1)	7 (1)
LDH
Normal	144 (71)	123 (53)	127 (50)	394 (57)	NR	NR
Abnormal	30 (15)	88 (38)	98 (38)	216 (31)
NR	30 (15)	22 (9)	31 (12)	83 (12)

Data are presented as n, n (%) or median (range). ELCWP: European Lung Cancer Working Party; NR: not reported; LDH: lactate dehydrogenase.

Table 2– Distributions of the classifications

Classification	Group I	Group II	Group III	Group IV	Group V	Missing
All patients^#
Albain et al. [5]	145 (21)	60 (9)	227 (33)	178 (26)		83
Sagman et al. [6]	212 (31)	37 (5)	364 (53)	73 (11)		7
Paesmans et al. [4]	113 (17)	352 (53)	103 (15)	98 (15)		27
Sculier et al. [2]	195 (28)	53 (8)	279 (40)	166 (24)
Patients with extensive disease^¶
Albain et al. [5]			227 (56)	178 (44)		50
Sagman et al. [6]		23 (5)	356 (79)	73 (16)		3
Paesmans et al. [4]		227 (53)	103 (24)	98 (23)		27
Sculier et al. [2]		10 (2)	279 (61)	166 (37)
Foster et al. [7]	232 (51)	34 (7)	131 (29)	4 (1)	52 (12)	2

Table 3– Prognostic value of the four prognostic independent covariates from the study by Paesmans et al. [4]

Characteristic	Derivation series^#		Validation series^¶
	HR (95% CI)	p-value	HR (95% CI)	p-value
Karnofsky performance status ≥80	0.73 (0.60–0.86)	<0.001	0.56 (0.46–0.70)	<0.0001
Female sex	0.71 (0.54–0.94)	0.02	0.83 (0.66–1.02)	0.08
Limited disease	0.66 (0.57–0.78)	<0.001	0.45 (0.38–0.51)	<0.0001
Neutrophils ≤75%	0.80 (0.67–0.96)	0.05	0.68 (0.57–0.82)	<0.0001

HR: hazard ratio. ^#: n = 763, 45 censored observations; ^¶: n = 693, 47 censored observations.

Table 4– Estimates of median survival times in each category and overall comparison

Classification	Group I	Group II	Group III	Group IV	Group V	Overall	p-value
All patients
Albain et al. [5]	17.0 (14.4–20.8)	10.6 (7.5–12.5)	10.5 (9.4–11.5)	7.3 (6.4–11.5)		1.58 (1.45–1.72)	<0.0001
Sagman et al. [6]	15.7 (13.4–19.1)	13.3 (10.6–16.7)	9.5 (8.8–10.0)	6.4 (4.7–7.1)		1.60 (1.47–1.75)	<0.0001
Paesmans et al. [4]	19.4 (13.9–23.9)	11.1 (10.6–11.8)	7.1 (6.6–8.7)	6.4 (4.0–7.3)		1.68 (1.53–1.84)	<0.0001
Sculier et al. [2]	15.9 (13.5–19.9)	10.4 (7.6–13.0)	10.5 (9.6–11.2)	6.8 (6.1–7.8)		1.57 (1.45–1.71)	<0.0001
Patients with extensive disease
Albain et al. [5]			10.5 (9.4–11.5)	7.3 (6.4–8.3)		1.87 (1.51–2.31)	<0.001
Sagman et al. [6]		15.9 (11.0–21.3)	9.6 (9.0–10.3)	6.4 (4.7–7.1)		1.70 (1.35–2.13)	<0.001
Paesmans et al. [4]		10.9 (10.0–11.6)	7.1 (6.6–8.7)	6.3 (4.0–7.3)		1.44 (1.27–1.64)	<0.001
Sculier et al. [2]			10.5 (9.6–11.2)	6.8 (6.1–7.8)		1.72 (1.41–2.10)	<0.001
Foster et al. [7]	9.7 (8.8–10.7)	9.5 (7.0–11.1)	9.4 (8.6–10.4)		6.4 (3.3–7.8)	1.23^# (1.12–1.35)	<0.001

Data are presented as hazard ratios (95% CI), unless otherwise stated. For the calculation of the overall hazard ratio, only patients with all classifications available were considered (n = 588 for all patients; n = 383 for patients with extensive disease). The calculation of the hazard ratio was made using the classification as a continuous covariate. ^#: groups IV and V pooled.

Table 5– Paired comparisons

Classification	Group I versus II	Group II versus III	Group III versus IV	Group IV versus V
All patients
Albain et al. [5]	1.93 (1.39–2.64)	1.21 (0.90–1.66)	1.88 (1.53–2.31)
Sagman et al. [6]	1.47 (1.00–2.11)	1.73 (1.23–2.51)	1.67 (1.29–2.15)
Paesmans et al. [4]	2.24 (1.77–2.87)	1.74 (1.39–2.18)	1.33 (1.00–1.77)
Sculier et al. [2]	2.17 (1.57–2.96)	1.13 (0.84–1.53)	1.77 (1.45–2.15)
Patients with extensive disease
Albain et al. [5]			1.88 (1.53–2.31)
Sagman et al. [6]			1.70 (1.30–2.18)
Paesmans et al. [4]		1.52 (1.20–1.92)	1.33 (1.00–1.77)
Sculier et al. [2]			1.77 (1.45–2.15)
Foster et al. [7]	1.18 (0.79–1.69)	1.04 (0.71–1.57)		1.59 (1.15–2.17)

Table 6– Predictive value of the previously published Cox models

Classification	HR (95% CI)	p-value
All patients
Paesmans et al. [4]	8.23 (5.88–11.69)	<0.001
Sculier et al. [2]	9.46 (6.67–13.50)	<0.001
Patients with extensive disease
Paesmans et al. [4]	5.60 (3.13–9.93)	<0.001
Sculier et al. [2]	12.49 (5.57–28.01)	<0.001
Foster et al. [7]	8.83 (4.66–16.74)	<0.001

Hazard ratios (HRs) are for a continuous covariate calculated as a linear combination of the independent prognostic covariates weighted by their regression coefficients.

Table 7– Concordance probability estimates (CPEs)

	CPE (95% CI)
RECPAM classifications
All patients
Albain et al. [5]	0.63 (0.61–0.65)
Sagman et al. [6]	0.62 (0.61–0.65)
Paesmans et al. [4]	0.62 (0.60–0.64)
Sculier et al. [2]	0.63 (0.61–0.65)
Patients with extensive disease
Albain et al. [5]	0.58 (0.55–0.60)
Sagman et al. [6]	0.55 (0.53–0.57)
Paesmans et al. [4]	0.58 (0.55–0.60)
Sculier et al. [2]	0.57 (0.54–0.59)
Foster et al. [7]	0.56 (0.54–0.59)
Cox models
All patients
Paesmans et al. [4]	0.64 (0.62–0.66)
Sculier et al. [2]	0.65 (0.63–0.67)
Patients with extensive disease
Paesmans et al. [4]	0.58 (0.55–0.60)
Sculier et al. [2]	0.59 (0.56–0.62)
Foster et al. [7]	0.60 (0.57–0.63)

The CPE represents the proportion of patients in whom the predicted and observed survival times are ordered the same way. RECPAM: recursive partitioning and amalgamation.