Unexpected long-term survival after low-dose palliative radiotherapy for nonsmall cell lung cancer

Abstract

BACKGROUND

Many experienced oncologists have encountered patients with proven nonsmall cell lung cancer (NCLC) who received modest doses of palliative radiotherapy (RT) and who unexpectedly survived for > 5 years; some were apparently cured. We used a very large prospective database to estimate the frequency of this phenomenon and to look for correlative prognostic factors.

METHODS

Patients with histologically or cytologically proven NSCLC, treated with palliative RT to a dose of ≤ 36 Gy, were identified from a prospective database containing details of 3035 new patients registered from 1984–1990.

RESULTS

An estimated 1.1% (95% confidence interval, 0.7–1.6%) of 2337 palliative RT patients survived for 5 or more years after commencement of RT, including 18 patients who survived progression-free for 5 years. Estimated median survival was 4.6 months. Five-year survivors had significantly better Eastern Cooperative Oncology Group performance status at presentation than non-5-year survivors (P = 0.024) and were less likely to have distant metastases (P = 0.020). RT dose did not appear to be a significant prognostic factor. Patients who survived 5 years without progression had an estimated 78% probability of remaining free from progression in the next 5years.

CONCLUSIONS

Approximately 1% of patients with proven NSCLC survived for > 5 years after palliative RT, and many of these patients appeared to have been cured by a treatment usually considered to be without curative potential. Because of the potential for long-term survival, doses to late-reacting normal tissues should be kept within tolerance when prescribing palliative RT in NSCLC. Cancer 2006. © 2006 American Cancer Society.

With the exception of the significant subgroup of patients with good performance status, minimal or absent weight loss, and disease suitable for inclusion within a high-dose radiation therapy (radiotherapy [RT]) target volume,1 patients with unresectable nonsmall cell lung cancer (NSCLC) are usually considered to have incurable disease.2 Patients with advanced NSCLC are candidates for a variety of management approaches, including initial chemotherapy3, 4 and/or RT,5, 6 or supportive care.7 Despite the well known dismal prognosis for such advanced disease,8 experienced radiation oncologists will have encountered patients treated with relatively low doses of RT, with purely palliative intent, who return to the clinic year after year without evidence of disease progression. A proportion of these patients appear to have been cured by treatment that would not normally be considered to have any curative potential whatsoever. There is very little information on this phenomenon in the literature, and it is, therefore, difficult to say with any certainty how frequently it occurs.

In this study, we have used a very large prospective database to identify patients with NSCLC who were treated with RT with palliative intent to a dose ≤ 36 Gy. Our aims were to determine what proportion of palliatively treated patients experienced prolonged survival and, by investigating their survival beyond 5 years, to study the possibility that some of these patients had been cured. We analyzed the well known prognostic factors for survival in NSCLC and sought features that may predict unexpectedly long survival with low-dose RT. Characteristics of survivors were compared with those of nonsurvivors. We believe that these patients are important because an understanding of the mechanisms underlying their prolonged survival could be invaluable in the development of novel treatment strategies.

MATERIALS AND METHODS

Lung Database

A lung cancer database was established at the Peter MacCallum Cancer Center (Peter Mac) in 1984. Peter Mac is a large tertiary referral center, serving the state of Victoria. Information collected on all new patients with lung cancer included demographic data, details of symptoms at presentation, paraneoplastic syndromes, initial extent of disease, prognostic factors including weight loss and performance status, and a range of radiation treatment factors including radiation dose, fractionation, and overall treatment time. Between January 1984 and March 1990, data were collected on 3035 new patients with NSCLC.

Because the international staging system was undergoing revision at the time the database was established and during the study period, detailed clinical TNM staging information was not initially collected on all patients, but disease was classified as locoregional or metastatic.

Treatment Policy

During the study period, almost all patients who had symptomatic unresectable NSCLC, and were unsuitable for radical high-dose RT, were offered treatment with palliative RT. Radical RT (50–60 Gy in 2 Gy fractions) was offered to patients with Stage I-IIIB disease with medically or surgically inoperable NSCLC who had loss of < 10% of body mass, Eastern Cooperative Oncology Group (ECOG) performance status < 2, and adequate lung function for high-dose RT. Patients treated palliatively were given lower-dose palliative RT (< 30 Gy) or higher-dose palliative RT (30–36 Gy) depending on an estimate of prognosis made by the treating oncologist.9 There was, however, an increasing trend over time to use lower doses because of greater convenience and lower toxicity. The most commonly used fractionation schedules were 20 Gy in 5 fractions in 1 week (biologically effective dose [BED] for tumor, assuming α/β of 10, = 28 Gy), 30 Gy in 10 fractions in 2 weeks (BED = 39 Gy), and 36 Gy in 12 fractions in 2.5 weeks (BED = 46.8 Gy), with treatment given 5 days each week.10 In the case of patients without known distant metastases who received thoracic RT, all known disease was generally included within the radiation field, provided this could be accomplished without a high risk of significant toxicity. For patients with distant metastasis, no attempt was usually made to include asymptomatic disease within the RT field. During the study period, palliative chemotherapy was not routinely used for NSCLC.

Eligibility

Patients were eligible for this study if they presented to Peter Mac between January 1984 and March 1990, had a diagnosis of NSCLC, and were treated with RT with palliative intent to a dose ≤ 36 Gy. The intent of the study was to exclude patients who were to receive > 36 Gy; however, planned doses were available for only a small (nonrandom) subset of patients (including the patients who were known to have survived 5 yrs). Instead, patients were considered eligible if they were recorded in the database as having completed RT with ≤ 36 Gy, or if they had not completed RT but had received ≤ 36 Gy with fraction sizes > 2.5 Gy. Patients who had been prescribed a radical RT dose (≥ 40 Gy) but who had failed to complete RT were included in our analysis of results of radical RT, which have been described in a separate report.11

Data Collection and Analysis

During the study period, detailed information on each new NSCLC patient seen at Peter Mac was collected prospectively, entered into the database, and updated frequently during a patient's clinical course. After RT, patients were closely monitored in the clinic with regular physical examinations and chest radiographs. Inquiries were made to the family physicians of patients who had not been seen in clinic for more than 12 months to collect information on relapse and mortality. Further information was subsequently obtained from medical case notes of patients identified as 5-year survivors from the database, using a standardized data collection form. Results of staging investigations and pathology reports were used to retrospectively assign patients to clinical or pathologic stages according to the 1997 update of the international TNM classification.12 The RT treatment record was consulted to record acute toxicities of RT.

The data were analyzed with a close-out (study censor) date of November 29, 2003. Only 4 (0.2%) of the 2297 NSCLC patients were lost before this date, 3 patients in the first 6 months after RT and 1 patient 9.2 years after commencing RT. The estimated median duration of potential follow-up from the date of commencement of RT to the close-out date was 19.3 years.13 Survival has been measured from the date of commencement of RT until the date of death, regardless of cause. Patients who were lost to follow-up before the close-out date had their survival data censored at the date of last contact. Patients who were alive at the close-out date had their survival data censored at that date. Time to progression has been measured from the date of commencement of RT until the date of progression. Patients who died without progression before the close-out date had their time to progression censored at their date of death. Patients who were lost to follow-up before the close-out date had their time to progression censored at the date of last contact if they were progression-free at the time. Patients who were alive and progression-free at the close-out date had their time to progression censored at that date. Two patients for whom the progression status was unknown at the time of death had their times to progression censored backward to the last date at which the patient was known to be progression-free. One patient with a missing history was excluded from the study of time to progression. Estimates of overall survival and freedom from progression were made using the Kaplan–Meier method.14

To compare characteristics of 5-year survivors and nonsurvivors, each patient who was lost to follow-up before 5 years was classified as a 5-year survivor if, given their observed survival time, their conditional probability of surviving 5 years was greater than 0.5, that is if the patient was more likely to have been a 5-year survivor than not. Characteristics of 5-year survivors were compared with those of nonsurvivors using exact statistical tests: the Fisher exact test for binomial data (gender, weight loss, etc.), the Cochran–Armitage test for trend for ordinal data (performance status), the chi-square test for contingency tables for categorical data (histologic groups), and the Pitman permutation test for continuous data (age, RT dose).

Given the absence of data on planned RT dose, it was not possible to look at the prognostic significance of this factor. There is a potential bias in studying the prognostic significance of the actual RT dose because patients who died before completing their RT could possibly have received smaller doses than those who survived long enough to complete their RT. All patients who completed RT did so within 68 days of commencing RT. Thus, to study the prognostic significance of the actual RT dose given, a “landmark” analysis was performed, restricting attention to the 1671 patients who survived at least 68 days after commencing RT.

Ninety-five percent confidence intervals (95% CI) have been given for the main results. Statistical analyses were carried out using S-PLUS 2000 (MathSoft, Seattle, WA), StatXact 6.0 (CYTEL Software Corporation, Cambridge, MA), and SPSS 11.0.1 (SPSS Inc., Chicago, IL) software.

RESULTS

Overall, 3035 (74%) of 4123 patients in the database were coded as NSCLC patients. Of these, 2337 (77% of 3035) received palliative RT after presentation at Peter Mac. The dose received was ≤ 36 Gy for 2308 (99%) of these 2337 patients. Of 2308 patients who received ≤ 36 Gy, 2172 (94% of 2308) were recorded as completing RT and were, thus, eligible for the study. Of 2308 patients who received < 36 Gy, 2172 were recorded as completing RT and were, thus, eligible for the study. Of the remaining 136 patients who were recorded as not completing RT, 10 had fraction sizes < 2.5 and were, thus, ineligible for analysis on the basis of an intention to treat because the original treatment intent was radical, and their therapy was changed to palliative during the course of therapy, leaving 126 of these 136 patients eligible for the study. A further patient was excluded, as, on review of her records, she was presumed to have metastatic breast cancer and not NSCLC. Thus, the number of eligible patients was 2172 + 126 − 1 = 2297.

The estimated median survival after RT was 4.6 months (95% CI, 4.3–4.9 mos) for 2297 eligible patients. An estimated 1.1% (0.7–1.6%) survived 5 years (Fig. 1). Only 3 patients were lost to follow-up in the first 5 years after RT. These were all presumed to be non-5-year survivors as the estimated conditional probability of surviving 5 years given the observed survival times (1.6, 3.3, and 5.5 mos) was < 0.5 for each patient. There were 24 patients who were known to be 5-year survivors. These patients had significantly better ECOG performance status than non-5-year survivors (P = 0.024), and they were less likely to have distant metastases (P = 0.020) (Table 1). These 2 factors appeared to be independently associated with the probability of surviving 5 years (P = 0.032, Cochran–Armitage test for trend for testing significance of performance status stratified by presence or absence of distant metastases, and P = 0.032, testing common odds ratio in stratified 2×2 contingency tables for testing significance of distant metastases stratified by ECOG performance status).

Figure 1. Median survival after radiotherapy.

Table 1. Patient Characteristics and Percentage of 5-Year Survivors

	No. of patients	% 5-year survivors	95% CI	P
SVCO: superior vena caval obstruction; NSCLC: nonsmall cell lung cancer. a Excludes 14 patients with NSCLC of other or mixed or uncertain cell type. b Excludes 76 patients with unknown weight loss status; 98 patients who were known to have lost weight, but the amount of weight loss was unknown, so they have been included in the group not known to have > 10% weight loss. c Excludes 328 patients with unknown performance status. d Excludes 1 patient with missing SVCO status.
All patients	2297	1.1	0.7–1.5
Histologya
Squamous	1134	1.1	0.6–1.9
Adenocarcinoma	627	0.8	0.3–1.9
Alveolar cell	5	0.0	0.0–50
Large cell undifferentiated	517	1.4	0.7–2.8	0.63
Weight loss known to be > 10%b
No	1672	1.1	0.7–1.8
Yes	549	0.9	0.4–2.2	0.81
Performance statusc
0	75	5.3	2.0–13
1	905	1.2	0.7–2.2
2	600	0.5	0.2–1.5
3	299	1.0	0.3–3.1
4	90	0.0	0.0–3.9	0.024
Age, yrs
< 50	125	2.4	0.8–7.2
50–59	442	0.9	0.3–2.4
60–69	921	0.9	0.4–1.7
70–79	710	1.1	0.6–2.2
≥ 80	99	1.0	0.1–6.8	0.43
Gender
Male	1792	0.9	0.5–1.5
Female	505	1.6	0.8–3.1	0.21
Disease extent
Local/locoregional	1381	1.4	0.9–2.2
Metastatic	916	0.4	0.2–1.2	0.020
SVCOd
No	2179	1.1	0.7–1.6
Yes	117	0.0	0.0–3.0	0.63

Given the absence of data on planned RT dose, it was not possible to look at the prognostic significance of this factor. The prognostic significance of the RT dose given was studied in a landmark analysis, restricting attention to 1671 patients who survived at least 68 days after commencing RT. Five-year survivors received significantly higher doses than non-5-year survivors (P = 0.046). However, differences were not significant after adjusting for either performance status (P = 0.11, Pitman permutation test stratified by performance status) or for the presence of distant metastases (P = 0.11, Pitman permutation test stratified by the presence or absence of distant metastases).

There was no evidence of excessive acute toxicity in the 5-year survivors; the only acute toxicities of > Grade 1 were 1 case of Radiation Therapy Oncology Group (RTOG) Grade 3 and 3 cases of RTOG Grade 2 esophageal toxicity (Table 2). 2

Table 2. Characteristics of 5-Year Survivors

Patient no.	Site treated	Dosage	Stage	Histology	Basis
scf: supraclavicular fossa; fx: fractions; SCC: squamous carcinoma; adeno: adenocarcinoma; rp: retroperitoneal.
1	Upper lobe	20 Gy 5 fx	I	SCC	Bronchoscopy
2	Upper Lobe	20 Gy 5 fx	I	SCC	Bronchoscopy
3	Lung	20 Gy 5 fx	I	Large cell	Bronchoscopy
4	Upper lobe	20 Gy 5 fx	I	SCC	Sputum cytology
5	Upper lobe	20 Gy 5 fx	IIA	SCC	Bronchoscopy
6	Lower lobe	20 Gy 5 fx	IIIA	SCC	Bronchscopy
7	Upper lobe scf	20 Gy 5 fx	IIIB	Large cell	Needle biopsy
8	Lower lobe	20 Gy 5 fx	IIIB	SCC	Open biopsy
9	Lower lobe	20 Gy 5 fx	IIIB	Adeno	Open biopsy
10	Lumbar spine	20 Gy 5 fx	IV	Large cell	Rib excision + sputum
11	Brain	24 Gy 6 fx	IV	Adeno	Open biopsy
12	Brain	24 Gy 6 fx	IV	Large cell	Previous surgery
13	Middle lobe	28 Gy 10 fx	IIIA	SCC	Bonchoscopy
14	Middle lobe	30 Gy 10 fx	I	Large cell	Sputum cytology
15	Lower lobe	30 Gy 10 fx	II	Adeno	Washings & sputum
16	Abdomen rp mass	30 Gy 10 fx	IV	Large cell	Laparotomy biopsy
17	Upper lobe	36 Gy 12 fx	IIIA	SCC	Sputum cytology
18	Upper lobe	36 Gy 12 fx	pIIIA	Adeno	Mediastinal biopsy
19	Upper lobe	36 Gy 12 fx	IIIA	SCC	Needle biopsy
20	Upper lobe	36 Gy 12 fx	IIIA	SCC	Bronchoscopy
21	Upper lobe	36 Gy 12 fx	IIIA	SCC	Needle biopsy
22	Lower lobe	36 Gy 12 fx	IIIB	SCC	Bronchial washings
23	Upper lobe	36 Gy 12 fx	IIIA	Adeno	Open biopsy

Long-Term Survival and Freedom from Progression

Of 24 known 5-year survivors, an estimated 32% (95% CI, 17–53%) survived another 5 years. One had no information available with respect to time to progression; 2 had experienced locoregional progression; and 3 had experienced systemic disease progression within 5 years of commencing RT. The 2 with locoregional progression died 1.3 and 5.8 years after progressing, and the 3 with systemic disease progression died 1.3, 4.8, and 6.9 years after progressing. Of the remaining 18 patients who were free from progression (FFP) at 5 years, only 3 patients were known to have experienced disease progression within the next 5 years, in all 3 cases because of locoregional disease. These patients all died within 1 year of progression. No further patients were known to have progressed before the study close-out date. For patients who had survived for 5 years without progression, the estimated probability of surviving for another 5 years was 43% (95% CI, 23–66%), and the estimated probability of remaining FFP for the next 5 years was 78% (95% CI, 48–93%) (Fig. 2).

Figure 2. Long-term survival and freedom from disease progression.

The only significant late toxicities recorded were 3 cases of Grade 2 pulmonary toxicity; 1 case of mitral regurgitation; 1 case with aortic stenosis, congestive cardiac failure, left anterior descending coronary artery stenosis, and aortic regurgitation; and 1 case with an abscess in the tumor site. All but 1 of these cases were treated to 30–36 Gy.

DISCUSSION

Our data indicate that a chance for prolonged survival and possibly even cure exists for approximately 1% of patients with NSCLC who receive palliative RT. This is a very small proportion, but lung cancer is a very common malignancy. It is important that the frequency of this phenomenon should be appreciated, so that claims of apparent cure by novel treatment strategies or even by unconventional medicine or “faith healing” can be seen in an appropriate context. All patients in this study had histologic or cytologic diagnoses of NSCLC in an appropriate clinical context. It is possible that errors could have been made in diagnosis in a proportion of cases, but it is very unlikely that all of the cases were misdiagnoses. In many of these patients, biopsy specimens were generous, including some surgical cases. It is well known that conventional cytologic or histopathologic tumor morphology is, by itself, a poor predictor of treatment response in NSCLC. The phenomenon reported here is potentially an important one, in that a subset of patients with NSCLC appears to have disease that is curable with minimal therapy and that prospective identification of such patients could potentially profoundly influence treatment.

There is one other large published series showing prolonged survival after palliative RT. Quddus and colleagues from Edinburgh15 reported on a series of more than 4000 patients, 70% of whom had histologic or cytologic evidence of NSCLC. They reported that 5-year survival was 1.3% and suggested that 1% of patients might be cured. These results are strikingly similar to our own. When Satoh et al. reported that 3.5% of 195 patients with Stage IIIA-IV NSCLC treated with platinum-containing chemotherapy survived for at least 3 years and that 3 (1.5%) patients survived for at least 5 years, they suggested that a small proportion of such patients might be cured by “appropriate” therapy.16 It is noteworthy that 10 of the 14 patients in that series who survived for at least 3 years had received “additional” local therapy with RT or surgery. These results are very similar to our own results with low-dose RT alone. The “appropriate therapy” for attaining this small proportion of 5-year survivors with the least toxicity may be palliative RT. In another study of long-term survival after platinum-based chemotherapy for NSCLC, Sculier et al. described the outcome for 1052 patients with advanced unresectable NSCLC treated with platinum-based chemotherapy in 7 clinical trials.17 In that series, 1.8% survived for 5 years and, interestingly, all of the survivors had received “complementary chest irradiation and/or surgery.” It is possible that such long-term survival may require administration of an effective local treatment, as in the case of solitary brain metastasis18 and may even be independent of chemotherapy.

We have been unable to identify any common features among survivors in our patients that could account for their unusually long survival. Spontaneous regression is exceedingly rare in lung cancer.19, 20 We have no information on psychological status of these patients, but limited evidence in the literature suggests that pretreatment optimism or pessimism is unlikely to influence survival.21 There is no evidence that our patients had abnormally high intrinsic radiosensitivity of normal tissues, which could be associated with enhanced tumor response.22 Several patients experienced disease progression in the second 5 years after treatment and might have had slowly proliferating tumors. For such prolonged survival to occur in our patients, distant metastases either must not have been present, or they were oligometastases (extrathoracic metastases) that were contained within treatment fields. It is well known that a proportion of patients with NSCLC who have solitary metastasis to the brain23 or adrenal glands24, 25 may be cured by resecting both the metastasis and primary tumor, and it is possible that this phenomenon could also occur with RT to both primary tumors and metastases if the disease is unusually radiosensitive.

In this study, prolonged survival was associated with a treatment that was given to relieve symptoms, and we do not mean to suggest that lower doses of RT should be used when cure is the intention of treatment. There is randomized evidence of a benefit from higher RT doses in terms of local control.26 Randomized studies show, however, that radical RT is associated with improved survival when enhanced by platinum-based chemotherapy27 or given as continuous hyperfractionated RT (CHART).28 Both the randomized British Medical Research Council study5 and our own nonrandomized study9 suggest that there may be a survival benefit associated with use of a higher rather than a lower dose of RT for palliatively treated patients.

In conclusion, our data show that close to 1% of patients with NSCLC have prolonged survival with doses of palliative RT that would not normally be considered sufficient for long-term disease control. When prescribing hypofractionated palliative RT for NSCLC, radiation oncologists should be aware of the potential for long-term survival and keep doses to late-reacting normal tissues within tolerance. We do not have an explanation for this phenomenon but believe that these patients have tumors that are both unusually radiosensitive and have a low tendency to produce blood-borne metastasis. Further studies of the biology of these tumors may provide valuable information.