Early Versus Delayed Therapy of Advanced Gastric Cancer Patients – Does it make a difference?

Abstract

Background

~50% of patients are diagnosed with advanced gastric cancer (AGC). Therapy is palliative but results in ill effects. The median overall survival (OS) of AGC patients is often <12 months. It is unclear if the early initiation of therapy in all AGC patients is beneficial.

Methods

A retrospective analysis of AGC patients in our database was carried out. The patients were divided in two groups: asymptomatic or symptomatic. We sought to assess whether the delay of systemic therapy was harmful in asymptomatic patients.

Results

135 patients were analyzed. Most patients were symptomatic (68%), males (67%), with low ECOG scores (0–1; 85%). In a univariate analysis, ECOG PS 0 (p=0.005), delayed initiation of therapy (p=0.03), and lack of symptoms (p=0.03) were associated with a longer OS. The multivariate model for OS identified only ECOG PS as an independent prognosticator of longer OS (p=0.02). Asymptomatic patients who had delayed (≥4 weeks) systemic therapy, had an OS rate of 77% at 1-year compared to 58% for patients treated within 4 weeks (p=0.47).

Conclusion

Symptomatic AGC patients had poor outcome compared to asymptomatic AGC patients. Asymptomatic patients with treatment delay had no detrimental effect on OS suggesting that timing of therapy can be based on patient selection.

Introduction

Gastric cancer (GC) represents a serious health problem on a global scale and is the second leading cause of cancer-related death worldwide.[1] In the United States, GC is less common, with 21,600 new cases and 10,990 cancer deaths occurring in 2013.[2] Between 2002 and 2008 the 5-year relative survival rate was only 27% according to the SEER database. [3] Nearly half of the patients diagnosed with GC have advanced, unresectable, and incurable cancer (AGC). Therapy for these patients is palliative and associated with considerable acute and chronic ill effects. The median overall survival (OS) of patients with AGC is often <12 months. Since their survival is short, the goal of therapy should be to relieve cancer-related symptoms and preserve the quality of life while we strive to prolong their OS. Thus, one can consider one of two approaches: (1) once the AGC is documented, start systemic therapy until cancer progression or intolerance or (2) treat based on tumor burden/symptoms and select the timing for treatment (meaning, delay therapy when AGC is not measurable) and symptoms are sparse. Such patients can be counseled/carefully monitored and treated later. In our institution multiple oncologists treat AGC patients and one of the two mentioned approaches are employed. The literature does not provide any guidance on this issue, therefore, we hypothesized that asymptomatic patients with low-volume AGC, will fare well with initial observation and therapy later.

A hypothetical representation of the typical course of AGC with and without therapy is presented. In the early phase AGC can be not measurable and patients have no or minimal symptoms. In the second phase, the bulk of AGC increases (measurable/evaluable by imaging) and patients have symptoms. This may be the best time to initiate therapy. In the third phase (without standard options), the cancer burden is high, with many symptoms and short OS. In the 3^rd phase, supportive care is best.

Methods

Patient selection

A retrospective review from a prospectively maintained database in the Department of Gastrointestinal Medical Oncology at The University of Texas M.D. Anderson Cancer Center (UTMDACC) was carried out. The UTMDACC Institutional Review Board approved this analysis. AGC patients managed between 2000 and 2013 were reviewed and had to meet the following criteria: confirmed histology, stage IV GC, in the first line therapy setting. Patients for whom no date of treatment initiation existed were excluded. No other selection criteria were implemented. Patients with no symptoms or non-specific symptoms were designated as “asymptomatic” independent of ECOG status. Two authors determined whether the patients were correctly designated, when it was not clear. The ECOG PS was extracted directly from the dictated medical oncology note.

Study design

The primary objective of this analysis was to assess whether the delay of therapy was harmful to asymptomatic AGC patients.

Follow-up and survival

Patients on observation were seen with imaging studies carried out every 2–3 months. Patients on therapy were seen frequently.

Statistical methods

Differences in age and BMI were compared by the Kruskal-Wallis test. Differences between other characteristics were tested using chi-square tests or Fisher’s exact tests as needed. OS was calculated from therapy initiation until last follow-up or death by Kaplan-Meier[4] methods. Univariate and multivariate Cox proportional hazards regression models[5] were used to assess the association between patient characteristics and OS. Characteristics significant in the univariate model at the 0.10 level were included in a multivariate model (full model), followed by backward elimination (reduced model). Recursive partitioning analysis[6] (RPA) was performed to identify combinations of characteristics that classify patients into OS risk groups, with at least 20 patients required in any terminal subgroup. Terminal subgroups with a pairwise log-rank p-value greater than 0.05 were combined. The time of delay in therapy was defined as the weeks between diagnosis and initial systemic therapy. RPA was performed in R [The R Foundation for Statistical Computing], remaining analyses used SAS 9.3 [The SAS Institute, Cary, NC] and figures were created in Stata 13.1 [Stata Corp, College Station, TX].

Results

123 patients with AGC were analyzed. Table 1 shows the patient characteristics. Patients were similar across symptom status except for ECOG score (p<0.001). Asymptomatic patients were more likely to have an ECOG score of 0 (44% vs. 7%).

Table 1.

Patient Characteristics by Symptom Status

Characteristics	All Patients N (%)	Asymptomatic Patients N (%)	Symptomatic Patients N (%)	P-value
All Patients	123(100%)	39(100%)	84(100%)
At Diagnosis
Age - median (min,max) N=123	58 (33, 83)	60 (33, 83)	58 (33, 79)	0.37
BMI - median (min,max) N=123	25.7 (17.1, 41.1)	26.8 (18.1, 41.1)	25.5 (17.1, 38.5)	0.71
Gender				0.68
Female	41(33.3%)	14(35.9%)	27(32.1%)
Male	82(66.7%)	25(64.1%)	57(67.9%)
Smoking History				0.55
Nonsmoker	62(50.4%)	17(43.6%)	45(53.6%)
Nonsmoker Quit	46(37.4%)	16(41%)	30(35.7%)
Smoker	15(12.2%)	6(15.4%)	9(10.7%)
Alcohol History				0.89
Never	41(33.3%)	12(30.8%)	29(34.5%)
Rarely	39(31.7%)	12(30.8%)	27(32.1%)
Frequent	29(23.6%)	11(28.2%)	18(21.4%)
Past	14(11.4%)	4(10.3%)	10(11.9%)
ECOG PS				<0.001*
0	23(18.7%)	17(43.6%)	6(7.1%)
1	81(65.9%)	22(56.4%)	59(70.2%)
2	10(8.1%)	0(0%)	10(11.9%)
Missing	9(7.3%)	0(0%)	9(10.7%)
Site of Mets				0.49
1-VisceralMets	34(27.6%)	10(25.6%)	24(28.6%)
2-Non-VisceraMets	76(61.8%)	23(59%)	53(63.1%)
3-Both	13(10.6%)	6(15.4%)	7(8.3%)
Type of Non-Visceral Mets				0.36
Pos Peritoneal Cyto	21(27.6%)	8(34.8%)	13(24.5%)
Other	55(72.4%)	15(65.2%)	40(75.5%)
Treatment
Treatment Delay (Weeks)				0.77
1–4	38(28.1%)	11(25.6%)	27(29.3%)
5–8	49(36.3%)	14(32.6%)	35(38%)
9–12	20(14.8%)	8(18.6%)	12(13%)
13–16	9(6.7%)	4(9.3%)	5(5.4%)
≥17	7(5.2%)	2(4.7%)	5(5.4%)
Type of Chemotherapy				0.53
P+A	33(26.8%)	12(30.8%)	21(25%)
P+A+T	72(58.5%)	20(51.3%)	52(61.9%)
Other	18(14.6%)	7(17.9%)	11(13.1%)
Surgery				0.46
Yes**	9(7.3%)	4(10.3%)	5(6.0%)
No	114(92.7%)	35(89.7%)	79(94.0%)

P=Platinum Analog; A=Antimetabolite; T=Taxane.

^*P-value for Fisher’s exact test, excluding patients with missing values

^**Among the 9 patients who received surgery, 8 patients had R0 resection (3 are asymptomatic and 5 are symptomatic) and 1 asymptomatic patient had R1 resection.

Treatment

Of the 123 patients studied, only 9 patients had surgery, 8 patients had R0 resection (3 asymptomatic and 5 symptomatic) and 1 asymptomatic patient had R1 resection. The majority of patients were treated with a platinum analog and fluoropyrimidine ± taxane (85%) while the remaining patients were treated in clinical trials. There was no difference in terms of therapy received by symptomatic versus asymptomatic patients (p=0.53). There was also no significant difference in terms of the type of therapy received by asymptomatic patients with therapy delayed by more than 4 weeks versus those treated within 4 weeks (p=0.07; Table 3).

Table 3.

Overall Survival Estimates among Asymptomatic Patients.

Character		Deaths /N	1-Year OS (SE)	P-value
Asymptomatic Patients		28 / 39	71.8 %(7.6 %)
Age				0.16*
	<60	11 / 16	72.2 %(11.9 %)
	≥ 60	17 / 23	71.6 %(9.9 %)
BMI				0.88
	Underweight(<18.5)	1 / 1	100 %(0 %)
	Normal(18.5–24.9)	11 / 15	79 %(10.8 %)
	Overweight(25–29.9)	10 / 14	63.5 %(13.1 %)
	Obese(>=30)	6 / 9	68.6 %(18.6 %)
Gender				0.25
	Female	9 / 14	55.2 %(15.4 %)
	Male	19 / 25	79.6 %(8.1 %)
Smoking History				0.50
	nonsmoker	10 / 17	66.3 %(12.4 %)
	nonsmoker quit	13 / 16	74 %(11.2 %)
	smoker	5 / 6	80 %(17.9 %)
Alcohol History				0.41
	Frequent/ Past	11 / 15	78.6 %(11 %)
	Rarely or Never	17 / 24	67.5 %(10.2 %)
ECOG PS				0.38
	0	12 / 17	68.8 %(11.6 %)
	≥1	16 / 22	75.8 %(9.5 %)
Treatment Delayed				0.47
	Within 4 weeks	7 / 11	58.4 %(16.3 %)
	Over 4 weeks	21 / 28	76.6 %(8.4 %)
Site Mets				0.26
	1-Visceral Mets	5 / 10	90 %(9.5 %)
	2-Non-Viscera Mets	18 / 23	67 %(10.2 %)
	3-Both	5 / 6	62.5 %(21.3 %)
Type of Non-Visceral Mets				0.22
	Pos Peritoneal Cyto	5 / 8	71.4% (17.1%)
	Other	13 / 15	64.6% (12.8%)
T Stage				0.83
	T2	0 / 0	N/A
	T3	20 / 26	71.2 %(9.2 %)
	T4	3 / 5	80 %(17.9 %)
	TX	5 / 8	70 %(18.2 %)
N Stage				0.97
	N0	5 / 6	83.3 %(15.2 %)
	N1	16 / 20	72.6 %(10.5 %)
	N2	2 / 4	50 %(25 %)
	N3	0 / 2	N/A
	NX	5 / 7	68.6 %(18.6 %)
Type of Chemotherapy				0.17
	P+A	8 / 12	63.5%(14.8%)
	P+A+T	15 / 20	79.2%(9.3%)
	Other	5 / 7	62.5%(21.4%)

P=Platinum Analog; A=Antimetabolite; T=Taxane.

^*P-value calculated based on continuous values

Overall Survival

Ninety-nine (80%) patients have died at last follow-up. The median follow-up time was 10.8 (range, 0.5 to 130.6) months. The median OS (95% CI) was 12.9 (10.4, 15.2) months. The 1- year OS rate was 53% (SE=5%).

Table 2 shows OS by patient categories. RPA for therapy delay partitioned the patients into two groups at 4 weeks and this grouping was used for further analyses. More asymptomatic vs. symptomatic patients were alive at 1 year (72% vs. 45%; p=0.03; Fig 1a). ECOG PS 0 (p=0.005) and delayed therapy (>4 weeks; p=0.03) were also associated with a prolonged OS. The site of metastases (visceral versus non-visceral versus both) did not affect OS. Patients with positive peritoneal cytology as their only metastatic site tended to have a longer OS (p=0.06). In the multivariate model, ECOG PS was an independent prognosticator of longer OS (p=0.04). After backward elimination, only ECOG PS remained.

Table 2.

Overall Survival Outcomes by Patient Characteristics.

Univariate
Characteristics		Deaths /N	1-Year OS (SE)	P-value
All Patients		99 / 123	53.3 %(4.7 %)
Age				0.82*
	<60	56 / 67	50.6 %(6.3 %)
	≥ 60	43 / 56	56.7 %(7 %)
BMI				0.92*
	Underweight(<18.5)	3 / 3	66.7 %(27.2 %)
	Normal(18.5–24.9)	40 / 50	58.5 %(7.1 %)
	Overweight(25–29.9)	33 / 40	42.7 %(8.3 %)
	Obese(>=30)	23 / 30	56.9 %(9.5 %)
Gender				0.53
	Female	33 / 41	55.1 %(8.2 %)
	Male	66 / 82	52.4 %(5.7 %)
Smoking History				0.90
	nonsmoker	48 / 62	52.1 %(6.6 %)
	nonsmoker quit	38 / 46	53.3 %(7.7 %)
	smoker	13 / 15	58.2 %(13.1 %)
Alcohol History				0.20
	Frequent/ Past	33 / 43	59.5 %(7.9 %)
	Rarely or Never	66 / 80	50.1 %(5.8 %)
ECOG PS				0.005
	0	15 / 23	72.7 %(9.5 %)
	≥1	76 / 91	48 %(5.5 %)
Treatment Delayed				0.03
	Within 4 weeks	31 / 38	34.8 %(8.2 %)
	Over 4 weeks	68 / 85	61.3 %(5.5 %)
Site Mets				0.82
	1-Visceral Mets	26 / 34	45.1 %(8.8 %)
	2-Non-Viscera Mets	62 / 76	56.6 %(5.9 %)
	3-Both	11 / 13	57.1 %(14.6 %)
Type of Non-Visceral Mets				0.06
	Pos Peritoneal Cyto	15 / 21	63.8% (11.0%)
	Other	47 / 55	53.8% (7.0%)
T Stage				0.78
	T2	2 / 2	50 %(35.4 %)
	T3	66 / 77	54.8 %(5.9 %)
	T4	13 / 17	50.7 %(12.5 %)
	TX	16 / 24	53 %(11.1 %)
N Stage				0.76
	N0	19 / 22	61.4 %(10.8 %)
	N1	50 / 58	55.9 %(6.8 %)
	N2	12 / 16	33.3 %(12.2 %)
	N3	0 / 3	100 %(0 %)
	NX	16 / 21	50.8 %(11.2 %)
Symptoms				0.03
	No	28 / 39	71.8 %(7.6 %)
	Yes	71 / 84	44.6 %(5.6 %)
Type of Chemotherapy				0.41
	P+A	23 / 33	45.3%(9.4%)
	P+A+T	62 / 72	56%(5.9%)
	Other	14 / 18	54.5%(13.1%)
Multivariate
Characteristics		HR	95% CI	P-value
ECOG PS				0.04
	≥1 vs. 0	1.86	(1.03, 3.38)
Treatment Delayed				0.08
	Over 4 weeks vs. Within 4 weeks	0.67	(0.43, 1.05)
Symptoms				0.20
	Yes vs. No	1.37	(0.85, 2.21)

^*P-value calculated based on continuous values

HR, Hazard Ratio; CI, Confidence Interval; P=Platinum Analog; A=Antimetabolite; T=Taxane.

Figure 1.

a. Overall Survival by Symptom Status.

b. Overall Survival by Treatment Delay among Asymptomatic Patients.

No Delay refers to treatment within 4 weeks of diagnosis while Delay refers to treatment more than 4 weeks after diagnosis. The selection of 4 weeks was determined by univariate recursive partitioning analysis (RPA).

Patients with a >4-week delay in treatment were more likely to be alive at 1-year (61%) compared to those who started treatment <4 weeks of diagnosis (35%; p=0.03). Asymptomatic patients who received delayed therapy had 1-year OS rate of 77% compared to 58% for patients treated in <4 weeks (p=0.47; Fig 1b). Among asymptomatic patients, the median time to therapy was 7 weeks (95% CI [5, 10]).

Discussion

In our practices, asymptomatic patients with low tumor burden are often observed till the onset of modest symptoms (interfering with patients’ routine) and especially when AGC is progressing on imaging studies. The issue is partially addressed in colorectal cancer patients.[7] In one study, 183 patients were randomized to chemotherapy or best supportive care until symptoms. There was a 5-month OS advantage for patients who were treated early. This was offset by 2 major factors. Firstly, patients with “delayed” therapy received suboptimal therapy, 40% received no therapy, and less than 20% received the same chemo as those in the comparison group. Secondly, this study contradicts published literature. [8, 9] We find that one can select patients to delay therapy based on three variables: symptom status, performance status, and tumor volume. Symptom status and tumor volume are easy to discern, however, PS is a soft call.[10, 11] We find that patients with low tumor volume and good PS do well even if their treatment is delayed. We did not include the treatment delay period in OS calculations to be stringent. This timeframe between diagnosis and treatment is an opportunity to avoid treatment-related toxicity. We may have preserved the quality of life in some patients, however this needs to be investigated further.

In other diseases, ECOG PS is an important OS prognosticator.[12, 13] Our report is consistent as patients with an ECOG score of 0 had the 1-year survival rate of 73% compared to those with an ECOG score of ≥1 (48%; p=0.005).

Our study being retrospective has shortcomings. We acknowledge that some patients naturally had favorable clinical course. This is a single high-volume center experience, and not necessarily generalizable. An ideal study would be a multi-center and prospective comparison of two approaches. Our data have strengths that can contribute to the management of patients with AGC: (1) we studied a large cohort of patients and (2) we studied a uniform group of untreated patients.

In conclusion, AGC patients with low tumor volume/good PS do at least as well with treatment delay. Patient selection may be important to provide patients with a treatment-free interval. Therapy is best given when it can AGC is progressive (especially associated with worsening symptoms) rather than given when it will only produce side-effects and no other material benefit. Although our patients are often anxious to start therapy immediately after diagnosis, our data in all patients suggests that a treatment delay of ≥4 weeks is safe.