Clinical neurological outcome and quality of life among patients with limited small-cell cancer treated with two different doses of prophylactic cranial irradiation in the intergroup phase III trial (PCI99-01, EORTC 22003-08004, RTOG 0212 and IFCT 99-01)

Abstract

Background: We recently published the results of the PCI99 randomised trial comparing the effect of a prophylactic cranial irradiation (PCI) at 25 or 36 Gy on the incidence of brain metastases (BM) in 720 patients with limited small-cell lung cancer (SCLC). As concerns about neurotoxicity were a major issue surrounding PCI, we report here midterm and long-term repeated evaluation of neurocognitive functions and quality of life (QoL).

Patients and methods: At predetermined intervals, the European Organisation for Research and Treatment of Cancer (EORTC) QLQ-C30 and brain module were used for self-reported patient data, whereas the EORTC–Radiation Therapy Oncology Group Late Effects Normal Tissue–Subjective, Objective, Management, Analytic scale was used for clinicians’ assessment. For each scale, the unfavourable status was analysed with a logistic model including age, grade at baseline, time and PCI dose.

Results: Over the 3 years studied, there was no significant difference between the two groups in any of the 17 selected items assessing QoL and neurological and cognitive functions. We observed in both groups a mild deterioration across time of communication deficit, weakness of legs, intellectual deficit and memory (all P < 0.005).

Conclusion: Patients should be informed of these potential adverse effects, as well as the benefit of PCI on survival and BM. PCI with a total dose of 25 Gy remains the standard of care in limited-stage SCLC.

introduction

The recommended treatment of patients presenting with limited small-cell lung cancer (SCLC) is combined thoracic radiotherapy and platinum-based chemotherapy, as well as prophylactic cranial irradiation (PCI) for good responders [1]. A previously published meta-analysis has shown that PCI reduced the lifetime risk of brain metastases (BM) (from 58.6% to 33.3% at 3 years) and improved survival (from 15.3% observed in the control group to 20.7%) [2, 3]. The reduction of BM incidence by itself is a great achievement as median survival does not exceed 5 months thereafter, and brain failure is correlated with neurocognitive decline and altered quality of life (QoL) [2, 4–7]. As the meta-analysis also suggested that the risk of BM might be reduced with higher PCI doses, our study's goal was to evaluate the effect of PCI dose escalation comparing a total dose of 25 Gy with 36 Gy. This study showed no significant reduction in the total incidence of BM after higher dose PCI [hazard ratio (HR) = 0.80 (0.57–1.11), P = 0.18] and an unexpected significant increase in mortality [HR = 1.20 (1.00–1.44), P = 0.05] [8]. Due to concerns about the neurotoxicity of PCI, the trial also focused on long-term repeated evaluation of neurocognitive functions and QoL [9, 10]. The potential impact of PCI on QoL is a relevant issue and particularly challenging in this multi-institutional trial setting as few studies have been conducted using longitudinal designs. Patients completed self-report European Organisation for Research and Treatment of Cancer (EORTC) QoL questionnaires (QLQ-C30/BN20) [11, 12]. In addition, physicians’ ratings of changes in functional and cognitive status were obtained using the EORTC–Radiation Therapy Oncology Group (RTOG) Late Effects Normal Tissue (LENT)–Subjective, Objective, Management, Analytic (SOMA) scale (LS scale) [13, 14]. Neuroimaging studies were also rated yearly to assess treatment-related radiological abnormalities. We report here the results of this analysis.

patients and methods

The trial modalities as well as patients’ characteristics were previously detailed [8]. Briefly, the trial included patients with limited-stage SCLC in complete remission after chemotherapy and thoracic radiotherapy. Cerebral imaging had to be carried out during the month preceding randomisation with no evidence of metastases either on brain computed tomography (CT) or on magnetic resonance imaging (MRI). All patients were required to have a baseline evaluation before randomisation both for QoL and for clinical neurological assessment. They were randomised to a standard (25 Gy/10 daily fractions) or higher PCI total dose (36 Gy/18 once-daily or 24 twice-daily fractions). Chemotherapy or any antitumour agent administered during PCI was not permitted because of the risk of additional neurological toxicity.

The main end point was the incidence of BM. Secondary end points were survival, QoL and late sequelae.

Patients filled the QLQ-C30/BN20 questionnaires at 6 and 12 months and then yearly [11, 12]. The LS scale was used yearly to evaluate potential late sequelae from radiation (see Table 1 for grade definitions) [13, 14]. A brain CT scan or MRI was carried out yearly or before in case of neurological symptoms. It was recommended to always use the same imaging technique for a given patient.

Table 1.

Selected LENT-SOMA scales, grade definition

Subjective signs: somnolence

Normal

Grade 1 = occasional

Grade 2 = intermittent

Grade 3 = persistent

Grade 4 = refractory, prevents daily activity, coma

Subjective signs: intellectual deficit

Normal

Grade 1 = minor loss

Grade 2 = moderate loss

Grade 3 = major loss

Grade 4 = complete loss of reasoning

Subjective signs: functional competence

Normal

Grade 1 = perform complex tasks

Grade 2 = cannot perform complex tasks

Grade 3 = cannot perform simple tasks

Grade 4 = incapable of self-care

Subjective signs: memory

Normal

Grade 1 = decreased short term, difficulty with learning

Grade 2 = decreased long term, loss of short-term memory

Grade 3 = loss of short- and long-term memory

Grade 4 = complete disorientation

Objective signs: cognitive functions

Normal

Grade 1 = minor loss of memory, reason and/or judgement

Grade 2 = moderate loss of memory, reason and/or judgement

Grade 3 = major intellectual impairment

Grade 4 = complete memory loss/incapable of rational thought

Objective signs: mood and personality changes

Normal

Grade 1 = occasional

Grade 2 = intermittent

Grade 3 = persistent

Grade 4 = total disintegration

MRI and/or CT

Normal

Grade 1

MRI: focal white matter changes; dystrophic cerebral calcification

CT: limited swelling, oedema, atrophy

Grade 2

MRI: white matter changes affecting £ 1 cerebral lobe

CT: limited perilesional necrosis

LENT-SOMA scale, Late Effects Normal Tissue–Subjective, Objective, Management, Analytic scale; MRI, magnetic resonance imaging; CT, computed tomography.

outcomes

The QLQ-C30 includes 30 items combined to derive 15 scales according to the EORTC guidelines. In order to study neurocognitive functions evolution and to limit the number of end points, we focused on the five functional scales (physical, role, emotional, cognitive and social functioning) and the global health status/QoL. The EORTC BN20 includes 20 items combined to derive 11 scales. We focused on four symptom scales: visual disorder, motor dysfunction, communication deficit and weakness of legs. These 26 QLQ-C30/BN20 scales range from 0 to 100, with high scores corresponding to a healthy/high status for the global and functional scales but to a high level of symptoms for the symptom scales.

The LS scale was used by physicians to assess neuropsychological status. It has 14 items graded from normal (0) to worst (4). It was a priori decided to focus on seven items: subjective signs (somnolence, intellectual deficit, functional competence and memory), objective signs (cognitive functions, mood and personality changes) and imaging (MRI and/or CT).

statistical methods

For each patient, a time series of QoL/LS forms was constituted excluding forms filled out after the occurrence of BM. Patients with no initial QoL/LS form were excluded as well as patients with only an initial QoL/LS form.

Since exact compliance to the QoL/LS schedules was rare, it was decided to allow a time window around each prespecified time: until the start of PCI for the baseline evaluation, within ±2 months for the 6-month evaluation (QoL) and within ±3 months for all subsequent evaluations (QoL/LS). In order to evaluate the compliance rate of the QoL/LS forms, the number of forms in each window was divided by the potential number of patients at risk in the same window. This number was obtained from the number of patients with sufficient follow-up and not known to be dead or to have a BM at the end of the window.

With either QoL or LS scales, most patients have a normal value (0 or 100). This important floor or ceiling effect makes an analysis with normal assumptions inappropriate. An alternative model is to consider the outcome variable in classes using for QoL the scales in four classes ([0–25], [25–50], [50–75] and [75–100]) and for LS the five raw values (0–4). These class variables can be modelled as a multinomial distribution or, more simply, as a binomial distribution with the event being all unfavourable classes ([0–75] for QoL functional scales, [25–100] for QoL symptoms scales and ≥1 for LS).

We adopted a unified graphical presentation of the QoL/LS scales based on the above dichotomisation, plotting the percentages of patients with an unfavourable status (e.g. cognitive functioning <75, motor dysfunction ≥25 or intellectual deficit ≥1).

To evaluate the effect of increasing PCI dose on QoL/LS, each patient's repeated observations were regressed on treatment using a cumulative logit model adjusted on the baseline observation, age at randomisation and time. Interaction between treatment and time was also explored to study a possible delayed effect of higher PCI dose. Both trend and heterogeneity tests were used. Graphics and models were limited to the first 3 years because of insufficient data beyond.

After the main analysis, several sensitivity analyses were carried out. One sensitivity analysis was restricted to patients with forms in each window, i.e. initial plus 4 for QoL or initial plus 3 for LS forms (‘restricted sample’). Another sensitivity analysis was carried out including all forms available whatever their delay, including also those beyond 3 years (‘enlarged sample’).

In all analyses, treatment refers to the allocated treatment. HR expresses the risk of 36 versus 25 Gy. All reported P-values are two sided, P-values <1% were considered significant in order to take into account the multiplicity of tests.

The EORTC SAS program was used to score the QLQ-C30/BN20 items. Analyses were carried out with SAS® version 9.1.

This study was registered with the ClinicalTrials.gov number NCT00005062.

results

From September 1999 to December 2005, 720 patients, 360 in each arm, were enrolled. The median (range) follow-up was 39 months (0–89 months).

The availability of QoL/LS forms is given in Table 2. At start, the compliance rate was good since >92% patients had an initial QoL/LS form. The compliance rate among survivors without BM decreased for the subsequent evaluations reaching ∼50% at 3 years. As the completion rate was <40% thereafter, it was decided to limit the analysis to the first 3 years.

Table 2.

Compliance of quality of life (QoL) and LENT-SOMA (LS) scales according to time and treatment group (in each cell the number of forms available is divided by the expected number of forms, see ‘statistical methods’ section)

	QoL		LS
	25 Gy	36 Gy	25 Gy	36 Gy
	Number of forms available/expected number of forms, percentage of forms available
Initial: before start of PCI (or before randomisation if no PCI)	330/360, 92%	337/360, 94%	336/360, 93%	339/360, 94%
6 months: form in [4–8] months	197/263, 75%	172/256, 67%
12 months: form in [9–15] months	123/178, 69%	119/175, 68%	129/180, 72%	122/174, 70%
24 months: form in [21–27] months	72/125, 58%	68/111, 61%	81/126, 64%	63/111, 57%
36 months: form in [33–39] months	51/101, 51%	41/83, 49%	59/103, 57%	41/83, 49%
48 months: form in [45–51] months	24/66, 33%	19/55, 35%	26/67, 39%	20/55, 36%

LENT-SOMA scale, Late Effects Normal Tissue–Subjective, Objective, Management, Analytic scale; PCI, prophylactic cranial irradiation.

As patients with no baseline QoL/LS form were excluded, as well as those with only an initial QoL/LS form, there remained 448 patients with a baseline QoL and at least one follow-up QoL and 322 patients with an initial LS and at least one follow-up LS. The sample size at each time is mentioned in the legend of Figure 1 for QoL and Figure 2 for LS; the rate of intermediate missing forms (i.e. missing form followed by an available form) was around 15%–20%.

Figure 1a.

Quality of life (QoL) evaluation according to treatment (continuous line: 25-Gy group, dotted line: 36-Gy group). Total sample size: 448, 387, 283, 143 and 92 at 0, 6, 12, 24 and 36 months, respectively. Upper curve: proportion of patients (95% confidence interval) in an unfavourable status by treatment and time (see ‘statistical methods’ section for definition). Lower histogram: proportion of patients in the three classes of unfavourable status, from better (left) to worst (right) in each arm. The P-value corresponds to the treatment effect estimated in the model of analysis. There are two scales (communication deficit and weakness of legs) that decline with time, and two scales (physical functioning and motor dysfunction) where a decline according to age at baseline, has been observed.

Figure 2a.

Evaluation of clinical and radiological sequelae using Late Effects Normal Tissue (LENT)-Subjective, Objective, Management, Analytic (SOMA) scale according to treatment (continuous line: 25-Gy group, dotted line: 36-Gy group). Total sample size: 322, 288, 146 and 101 at 0, 12, 24 and 36 months, respectively. Upper curve: proportion of patients (95% confidence interval) in an unfavourable status by treatment and time (see ‘statistical methods’ section for definition). Lower histogram: proportion of patients in the four classes of unfavourable status [grade 1 (left), grade 2, grade 3 and grade 4 (right)] in each arm. The P-value corresponds to the treatment effect estimated in the model of analysis. Two scales (intellectual deficit and memory) decline with time, and two scales (memory and brain magnetic resonance imaging and/or computed tomography abnormalities) where a decline according to age at baseline has been observed.

The variation of the proportion of patients in an unfavourable status, as defined in the ‘statistical methods’ section, is represented in Figures 1 and 2. For each selected scale, the evolution is presented by treatment and time. For instance, the proportions of patients with abnormal QoL-cognitive functioning (scale <75) at 0, 6, 12, 24 and 36 months are, respectively, 23%, 35%, 38%, 41% and 35% in the 25-Gy group and 25%, 34%, 41%, 46% and 47% in the 36-Gy group. For LS-intellectual deficit, the proportions of patients with abnormal status at baseline, 12, 24 and 36 months are, respectively, 10%, 12%, 20% and 27% in the 25-Gy group and 9%, 20%, 28% and 34% in the 36-Gy group. As can be seen on the histogram, only one case of complete loss of reasoning was reported (24 months 36 Gy). Overall, the majority of patients have minor or moderate complaints (<14% with the worst QoL status and <6% with grade 3 or 4 LS).

Using the binomial logit model, there was no significant difference between PCI dose groups in any of the 17 QoL/LS selected scales (all P-values >0.02). Interactions between treatment and time were also all not significant. The QoL/LS results did not change using a multinomial model. The P-value for intellectual deficit just reached P = 0.01 with an HR = 1.9.

Using the ‘restricted sample’ of 60 patients for whom all five QoL are available, the multinomial analysis gives the same results concerning treatment effect. With the ‘restricted sample’ of 68, for whom all four LS are available, the multinomial analysis gives the same results regarding treatment effect except for memory that shows heterogeneity with time, i.e. more memory deficit in the 36-Gy arm (heterogeneity P = 0.008 and trend P = 0.05 with a treatment effect restricted to 2 years).

Using the ‘enlarged sample’ including all forms whatever the time, the multinomial analysis gives the same results concerning treatment effect except for emotional functioning that shows an interaction of treatment with time (trend P = 0.004, heterogeneity P = 0.0009) and for weakness of legs (trend P = 0.01, heterogeneity P = 0.12).

Four QoL/LS scales worsen with time after randomisation (communication deficit: HR = 1.40, P = 0.005; weakness of legs: HR = 1.31, P = 0.004; intellectual deficit: HR = 1.53, P = 0.003; LS-memory: HR = 1.43, P = 0.001), and four scales worsen with age considering age at baseline (physical functioning: HR = 1.04, P = 0.0007; motor dysfunction: HR = 1.03, P = 0.01; LS-memory: HR = 1.04, P = 0.005; MRI and/or CT: HR = 1.06, P = 0.002). Social functioning becomes better with time: HR = 0.78, P = 0.009. Concerning LS-memory deficit, there is a trend for memory deterioration over time as represented in Figure 1, but only two patients developed a grade 4 memory deficit (at 24 months). There was no significant interaction between time and age, i.e. the effect of time did not vary between younger (≤60 years) and older patients (all P-values >0.08).

As seen in Table 3, there were more patients followed up with yearly brain CT scan (∼76% at baseline to 54% in year 3) than with MRI. The percentage of abnormalities is higher in MRI as it is a more sensitive exam. Most abnormalities observed concerned grade 1 changes; there were few grade 2 changes and no grade 3 or 4 changes. There is some relationship between radiological and neurological abnormalities: among the six grade 2 radiological abnormalities, we could observe 50%, 33% and 33% grade 2 or more intellectual, memory and cognitive deficits as compared with 5%, 6% and 5% among grade 0 or 1 radiological imaging (all P-values <0.04).

Table 3.

Results of brain CT and/or brain MRI

Radiological changes	Baseline	12 months	24 months	36 months
Brain MRI (total number)a	75	62	42	33
% Grade 0: normal	84	74	57	61
% Grade 1: focal white matter changes; dystrophic cerebral calcification	16	24	41	34
% Grade 2: white matter changes affecting ≤1 cerebral lobe	0	2	3	6
Brain CT (total number)a	256	135	70	48
% Grade 0: normal	97	85	92	85
% Grade 1: limited swelling, oedema, atrophy	3	14	8	13
% Grade 2: limited perilesional necrosis	0	1	0	2
MRI and/or CT (total number)a	315	184	98	72
% Grade 0	94	80	77	74
% Grade 1	6	19	22	22
% Grade 2	0	1	1	5

^aNumbers do not add up because some patients had both brain CT and MRI.

CT, computed tomography; MRI, magnetic resonance imaging.

discussion

As our trial addressed PCI dose escalation from 25 to 36 Gy, in patients with limited SCLC who were potentially long-term survivors, the purpose of the QoL and neurological follow-up was to detect any deterioration that could be linked to radiation dose. Thus, we prospectively collected this information from both the patients’ and the clinicians’ standpoint. Our trial shows that there was no significant difference between the two PCI dose groups in any of the QoL and neurological and cognitive functions QoL/LS selected scales. Since in the 36-Gy group, compared with the 25-Gy group, there was no significant reduction in the total BM incidence but a significant unexpected increase in overall mortality, the discussion will not focus on the PCI dose effect but rather on the different factors that may influence QoL/LS: PCI per se, time, baseline and age.

In this trial, the effect of PCI per se cannot be evaluated since all patients were allocated to PCI. The potential neurotoxicity of PCI is well known; however, there are few studies that have addressed it prospectively [2, 15, 16]. Confounding factors such as fatigue and depression, as well as divergent patient experiences, have made studying the neurocognitive sequelae of cancer treatment problematic. As there are difficulties in conducting sophisticated neuropsychological evaluations in the context of a large international clinical trial, we decided to assess cognitive functions and QoL in the whole population by patients’ self-report with the QoL questionnaires and physicians’ ratings with the EORTC-RTOG LS scale. Selection bias can be introduced by specific external neuropsychological assessments that are considered more objective [17]. However, it should be outlined that all North American patients included in our trial within the RTOG 0212 also underwent objective neuropsychological test batteries, soon to be published [18]. As described in retrospective studies, the patterns of neurocognitive impairments may include decline or loss of memory, attention, learning and executive functions, sometimes severe enough to meet the definition of dementia [19]. Seldom motor manifestations may be observed such as gait impairment or apraxia. In our trial, of 720 patients, one case of dementia and no case of apraxia were reported. Alzheimer's disease, the most common cause of dementia affects 4%–6% of Western populations aged >60 years [20]. Most studies that have reported neurological and intellectual impairment or abnormalities on brain CT scan potentially related to PCI were small and retrospective [21–23]. They have suggested that the concomitant use of cytotoxic drugs and the use of fractions ≥3 Gy should be avoided in order to lessen the risk of toxicity [16, 24, 25].

In our trial, many patients experienced a mild cognitive decline over time, but only few developed severe deterioration of QoL or neurological and cognitive functions, in the 3 years following PCI. This decline over time was significant for four items including intellectual deficit and memory. At 3 years, if the majority of patients had no intellectual deficit, 25% of patients had a grade 1 deficit and 5% had a grade 2 deficit or more. LS memory decline at 3 years was more frequent with grade 1 deficit observed in 44% and grade 2 deficit or more in 8% of patients.

When documented, baseline evaluations of neurological and/or neurocognitive functions are impaired before PCI in most SCLC patients [10, 24–26], and this has been confirmed in two prospective studies [2, 15]. In our study, the proportion of abnormal baseline LS evaluations varied from 0% to 27%, depending on the item, but only very few patients had grade ≥2. Clinicians found memory and cognitive deficits (grade ≥1) in 17% and 12% of patients, respectively.

Concerning baseline imaging, as shown on Table 3, the CT scans or MRI were considered normal in 94% of patients with a higher rate of abnormalities detected with MRI as expected since MRI is superior to CT scan for diagnosis of BM and white matter modifications [27–29]. As for QoL, the proportion of patients with baseline unfavourable QoL status varied according to the studied item from 6% (communication deficit) to 62% (global health status).

Baseline evaluations unavailable in most studies are important since many factors aside PCI may impact neurocognitive and neuropsychological functions, such as clinical characteristics of the patient (effects of chronic tobacco use, diabetes, age, vascular risk factors, depression or anxiety), the disease process itself (paraneoplastic syndromes and undiagnosed micrometastases) and other treatment effects (neurotoxicity related to chemotherapy) [24, 29, 30]. Delivering PCI to patients with epilepsy needing oral treatment should be discussed as one patient in our trial died of generalised seizures after PCI [8].

Our study confirms the importance of age as a cofactor of neurocognitive decline. Unsurprisingly, items such as physical functioning, motor dysfunction, LS memory and MRI and/or CT abnormalities worsen with age, in older compared with younger patients. Age >60 to 65 years is a risk factor for neurocognitive impairment and poorer neurobehavioral outcome for elderly patients after cerebral irradiation has been observed [19, 28, 29]. Therefore, age has to be considered in the decision to deliver PCI.

Concerning the availability of baseline and follow-up brain MRI and CT ratings of radiological abnormalities, which is one of the strengths of our study, we found some relationship between radiological and neurological abnormalities. Whether radiation-induced white matter injury correlates with any change in neurocognitive symptoms varies across studies probably because of the variety of imaging modalities used [28, 29].

As BM can cause considerable deterioration in QoL and neurocognitive function [31], we decided to limit the analysis to QoL and neurological evaluation before BM occurrence and censored patients who developed BM from further analysis. Our study differs from the EORTC study on PCI in metastatic SCLC patients where all QoL data were included whatever the patient's evolution [32].

As is commonly the case in all studies evaluating QoL, our study is also hampered by the decreasing patient compliance throughout time [33]. Missing data may be due to various reasons: worsening condition from progressive extracranial disease but also quite frequently logistical problems to fill up questionnaires as few centres have research nurses dedicated to QoL. Some fit patients may refuse to fill up the same questionnaire every year! The missing rate of QoL forms among potential compliers increased from 7% at baseline to 32% at 1 year and to 50% at 3 years. Although important, this is similar or slightly better than in other studies [15, 32]. With LS, missing rates varied from 6% at baseline to 29% at 1 year and to 46% at 3 years. In order to investigate the possible impact of missing data on the analysis, we introduced patterns of censoring depending on the timing of the last available data. We did not find any variation neither for QoL nor for LS.

In conclusion, we did not observe any significant difference in terms of QoL and LS evaluation between patients allocated to 25-Gy or 36-Gy PCI. Few patients had severe deterioration of neuropsychological and cognitive functions over the first 3 years. However, as the median survival rate in our population study was 19 months, mild deterioration of some items such as memory, intellectual deficit and cognitive functions possibly related to PCI should be balanced by considering the beneficial effects of PCI on survival and incidence of BM. Taking into account the overall results of this trial, PCI with a total dose of 25 Gy in 10 fractions remains the standard of care in limited-stage SCLC.

funding

Funded by unrestricted grants from Institut Gustave-Roussy; Association pour la Recherche sur le Cancer (2001); Programme Hospitalier de Recherche Clinique (2007). The EORTC contribution to this trial was supported by National Cancer Institute (Bethesda, MD) (5U10 CA11488-30 through 5U10 CA011488-38).

disclosure

The authors declare no conflict of interest.

Contributors—CLP, AD, AL and RA conceived and designed the trial. CLP, AD, SS, AW, EQ, RA, RJ and AL supervised the trial. CLP, SS, AW, EQ, CFF, TC, RK, MH RA, RJ, RW and DL contributed to patient enrolment and data collection. AD and AL did the statistical analysis. CLP, AD, SS, AW, RA and AL interpreted the results. CLP, AD and AL wrote the draft with critical revision from all other authors. CLP, AD and AL obtained funding. All authors have seen and approved the final version.

PCI Collaborative Group—The members of the PCI committees were as follows:

Secretariat (IGR)—C. Le Pechoux, A. Dunant, A. Laplanche, M. Tarayre and N. Bouvet-Forteau.

Steering Committee—R. Arriagada, T. Ciuleanu, A. Gregor, R. Jones, R. Komaki, A. Wolfson, C. Le Pechoux, E. Quoix and S. Senan.

Independent Data-Monitoring Committee—P. Postmus, M. Schemper and P. Van Houtte.

Data centre—IGR Biostatistics and Epidemiology Unit: A. Dunant, A. Laplanche, M. Tarayre and N. Bouvet-Forteau.

Associated data centres—EORTC Headquarters: L. Collette and E. Musat; RTOG: J. Moughan and R. Paulus.

The investigators of the PCI Collaborative Group are listed below by group in decreasing order of patients enrolled (shown in square parentheses).

PCI99-01 and IFCT (351 patients): T. Ciuleanu (Institutul Oncologic I. Chiricuta, Cluj-Napoca, Romania [43]); D. Lerouge (Centre Francois Baclesse, Caen, France [24]); C. Le Pechoux (Institut Gustave-Roussy, Villejuif, France [22]); C. Faivre-Finn (Christie Hospital, Manchester, UK [14]); R. Jones (Beatson Oncology Centre, Glasgow, UK [14]); F. Guichard (Polyclinique Bordeaux Nord, Bordeaux, France [14]); J. Salinas (Hospital Virgen de la Arrixaca, Murcie, Spain [11]); H. Horova (Masaryk Memorial Cancer Institute, Brno, Czech Republic [10]); P. Mulvenna (Northern Centre for Cancer Treatment, Newcastle, UK [10]); B. Dubray (Centre Henri-Becquerel, Rouen, France [9]); G. Frezza (Ospedale Bellaria, Bologna, Italy [9]); A. Price (Western General Hospital, Edinburgh, UK [9]); C. Tuchais (Centre Paul Papin, Angers, France [9]); P. Moisson (Hopital Foch, Suresnes, France [8]); E. Bardet (Centre Rene Gauducheau, St Herblain, France [7]); B. Lebeau (Hopital St Antoine, Paris, France [7]); F. Mascarenhas (University Hospital Santa-Maria, Lisbonne, Portugal [7]); F. Mornex (Centre Hospitalier Lyon-Sud, Pierre Benite, France [7]); I Monnet (Centre Intercommunal de Creteil, Creteil, France [6]); F. Ozanne (Centre Hospitalier de Beauvais, Beauvais, France [6]); U. Ricardi (University of Turin, Turin, Italy [6]); P. Bondiau (Centre Antoine Lacassagne, Nice, France [5]); J. Huet (Centre Hospitalier Gilles de Corbeil, Corbeil, France [5]); P. Verrelle (Centre Jean Perrin, Clermont-Ferrand, France [5]); K. Park (Samsung Medical Center, Seoul, Korea [5]); F. Rothe-Thomas (Centre Hospitalier Victor Dupouy, Argenteuil, France [5]); R Baeza (Instituto de Radiomedicina, Santiago, Chile [4]); P. Clavere (Hopital Dupuytren, Limoges, France [4]); P. Giraud (Institut Curie, Paris, France [4]); R. Leloup (Centre Hospitalier Regional La Source, Orleans, France [4]); X. Sun (Centre Hospitalier de Montbeliard, Montbeliard, France [4]); R. Abratt (Groote Schuur Hospital, Cape, South Africa [3]); E. Chirat (Centre de Radiologie et de Traitement des Tumeurs, Meudon la Foret, France [3]); F. Honnadel (Centre Pierre Curie, Bevry, France [3]); P. Leclerc (Cabinet de Pneumologie et de Cancerologie, Le Vesinet, France [3]); M. Ogawara (National Kinki Central Hospital, Osaka, Japan [3]); E. Quoix, Hopital Lyautey, Strasbourg, France [3]); D. Whillis (Raigmore Hospital, Inverness, UK [3]); M. Benchalal (Centre Eugene Marquis, Rennes, France [2]); M. Ciupa (Centre de Radiotherapie d'Oncologie Medicale, Compiegne, France [2]); D. J. Fairlamb (New Cross Hospital, Wolverhampton, UK [2]); P. Gomez (Centre Frederic Joliot, Rouen, France [2]); A. Kanoui (Centre Physique Rouget, Sarcelles, France [2]); L. Martin (Centre Guillaume Le Conquerant, Le Havre, France [2]); M. R. Pfeffer (Chaim Sheba Medical Center, Tel-Hashomer, Israel [2]); J. F. Rosier (Centre Hospitalier Jolimont-Lobbes, Haine Saint-Paul, Belgium [2]); C. Schumacher (Centre Paul Strauss, Strasbourg, France [2]); H. Vidal (Hospital Virgen de las Nieves, Grenade, Spain [2]); R. Anghel (Institutul Oncologic Bucuresti, Bucarest, Romania [1]); I. Barillot (Centre Georges Francois Leclerc, Dijon, France [1]); H. Berard (Hopital Inter-Armees Saint-Anne, Toulon, France [1]); F. Blanchon (Centre Hospitalier de Meaux, Meaux, France [1]); J. Bretel (Centre de Radiotherapie de Charlebourg, La Garenne Colombes, France [1]); T. Ceschia (Azienda Ospedaliera Santa Maria Della Misericordia, Udine, Italy [1]); M. Chasen (Sandton Oncology Center, Arcadia, South Africa [1]); J. Cretin (Clinique Rochebelle, Ales, France [1]); J. Godoy (Hospital Militar, Bogota, Colombia [1]); P. Gonzalez-Mella (Clinica Reñaca, Viña Del Mar, Chile [1]); D. Kardamakis (University of Patras Medical School, Rion, Greece [1]); L. Petruzelka (Czech Lung Cancer, Prague, Czech Republic [1]); A. Serre (Centre Regional de Lutte contre le Cancer Val d'Aurelle Paul-Lamarque, Montpellier, France [1]).

EORTC (223 patients): C. Faivre-Finn (Christie Hospital, Manchester, UK [40]); R. Wanders (Radiotherapeutisch Instituut Limburg–Maastricht, Maastricht, Netherlands [30]); G. W. P. Kramer^†, R. Keus (Arnhem's Radiotherapeutisch Instituut, Arnhem, The Netherlands [24]); M. Hatton (Weston Park Hospital, Sheffield, UK [19]); A. Kobierska (Medical University of Gdansk, Gdansk, Poland [17]); J. Bussink (University Medical Center Nijmegen, Nijmegen, The Netherlands [15]); M. U. Abacioglu, Marmara University Hospital, Istanbul, Turkey [14]); F. Macbeth (Velindre Hospital, Cardiff, UK [12]); R. Ramlau (University School of Medical Sciences Poznan, Poland [10]); E. Salamon (Clinique Sainte Elisabeth, Namur, Belgium [7]); C. Pottgen (Universitaetsklinikum, Essen, Germany [5]); J. Van Meerbeeck (Universiteit Gent, Gent, Belgium [5]); V. Klein (Centre Saint-Yves, Vannes, France [4]); M. J. M. van Mierlo (AZ Rotterdam—Daniel Den Hoed Kliniek, Rotterdam, Netherlands [4]); P. Fourneret (Centre Hospitalier Universitaire, Grenoble, France [3]); B. Slotman (Academisch Ziekenhuis der Vrije Universiteit, Amsterdam, The Netherlands [3]); D. Papamichael (Bank of Cyprus Oncology Centre, Strovolos–Nicosia, Cyprus [2]); J. F. Rosier (Centre Hospitalier Jolimont-Lobbes, Haine Saint-Paul, Belgium [2]); K. Stellamans (Cazk Groeninghe—Campus Maria's Voorzienigheid, Kortrijk, Belgium [2]); A. Brewster (Royal Gwent Hospital, Cardiff, UK [1]); C. Goor (Algemeen Ziekenhuis Middleheim, Antwerpen, Belgium [1]); C. Focan (Centre Hospitalier Saint Joseph, Liege, Belgium [1]); J. Lester (Nevill Hall Hospital, Abergavenny, UK [1]); S. Morgan (Nottingham City Hospital, Nottingham, UK [1]).

RTOG, Eastern Cooperative Oncology Group, Southwest Oncology Group and Cancer and Leukaemia Group B (146 patients): G. Videtic (Cleveland Clinic Foundation, Cleveland, OH, USA [12]); H. Barthold (South Suburban Oncology Center, Quincy, MA, USA [5]); W. Curran (Thomas Jefferson University Hospital, Philadelphia, PA, USA [5]); A. Raben (Christiana Care Health Services, Newark, DE, USA [5]); D. Brachman (Arizona Oncology Services Foundation, Phoenix, AZ, USA [4]); A. Cmelak (Vanderbilt University Medical Center, Nashville, TN, USA [4]); E. Gore (Medical College of Wisconsin, Milwaukee, WI, USA [4]); R. Komaki (University of Texas MD Anderson Cancer Center, Houston, TX, USA [4]); R. McGarry (Indiana University Medical Center, Indianapolis, IN, USA [4]); J. Atkins (Southeast Cancer Control Consortium, Winston-Salem, NC, USA [3]); W. Clapper (Waukesha Memorial Hospital, Waukesha, WI, USA [3]); K. Collins (Trinity Cancer Care Center, Minot, ND, USA [3]); W. Demas (Summa Health System, Akron, OH, USA [3]); N. Kaufman (Riverview Medical Center, Red Bank, NJ, USA [3]); M. Kuettel (Roswell Park Cancer Institute, Buffalo, NY, USA [3]); M. Suntharalingam (University of Maryland Medical Systems, Baltimore, MD, USA [3]); G. Trivette (St Mary's Hospital, Richmond, VA, USA [3]); A. Fortin (L'Hotel-Dieu de Quebec, Quebec, Canada [2]); M. Greenberg (Pocono Cancer Center, East Stroudsburg, PA, USA [2]); J. Holland (Oregon Health and Science University, Portland, OR, USA [2]); D. Hornback (Northern Indiana Cancer Research Consortium, South Bend, IN, USA [2]); H. Kim (Wayne State University, Detroit, MI, USA [2]); A. Konski (Fox Chase Cancer Center, Philadelphia, PA, USA [2]); K. Lanier, Columbia River Oncology, Portland, OR, USA [2]); A. Leylek (Cancer Care Manitoba Foundation, Winnipeg, Manitoba, Canada [2]); A. Markoe (University of Miami, Miami, FL, USA [2]); C. Miyamoto (Temple University School of Medicine, Philadelphia, PA, USA [2]); S. Rivkin (Swedish Cancer Institute, Seattle, WA, USA [2]); D. Rock (St Mary's Hospital and Medical Center, Grand Junction, CO, USA [2]); H. Sandler (University of Michigan Medical Center, Ann Arbor, MI, USA [2]); P. Schaefer (Toledo Community Hospital, Toledo, OH, USA [2]); L. Souhami (McGill University, Montreal, Quebec, Canada [2]); J. Taylor (All Saints Cancer Center, Racine, WI, USA [2]); R. Timmerman (University of Texas Southwestern Medical School, Dallas, TX, USA [2]); H. Wagner (Milton S Hershey Medical Center, Hershey, PA, USA [2]); K. Sarma (Carle Clinic CCOP, Urbana, IL, USA [1]); R. Allison (East Carolina Medical School, Greenville, NC, USA [1]); J. Bahary (Notre Dame Hospital, Montreal, Quebec, Canada [1]); J. Bonner (Ocean Medical Center, Boston, MA, USA [1]); J. Brooks (Madigan Army Medical Center, Tacoma, WA, USA [1]); L. Cho (Fairview-University Medical Center, Minneapolis, MN, USA [1]); H. Choy (Moncrief Cancer Center, Fort Worth, TX, USA [1]); A. Christie (Geisinger Medical Center, Danville, PA, USA [1]); G. Cooley (St Vincent Regional Cancer Center CCOP, Green Bay, WI, USA [1]).