Second-line treatment improves survival and is increasingly used in patients with advanced NSCLC [2, 5, 25]. Prospective data evaluating patients' overall health condition during second-line treatment as reflected by PS or HR-QoL are still rare, in particular for elderly patients or patients with poor PS as frequently treated in routine clinical practice. Patients older than 70 years are of special interest; even their first-line treatment is under continuous debate [25, 26] because multimorbidity and a higher incidence of toxicities are assumed to reduce their overall clinical treatment benefit.
Our study demonstrates that the majority of patients with Stage III/IV NSCLC in routine clinical practice can maintain or improve physician-rated PS (KI) and patient-rated HR-QoL (EQ-5D) during second-line treatment with single-agent pemetrexed. Our population included 36% of elderly patients. The majority of all patients (58.0%) achieved KI benefit response after the second treatment cycle (with 90% of patients still on study), and KI benefit response rates increased further during the subsequent cycles. Without second-line treatment, PS and HR-QoL would most likely have declined due to the disease progression. Even with treatment, a decline in PS and HR-QoL would be expected for patients with disease progression or major toxicities. Correspondingly, patients who had experienced no Grade 3/4 toxicities during the first 2 cycles had a significant advantage with respect to achieving KI benefit response. Patients with high baseline KI also had a significant advantage with respect to achieving KI benefit response, indicating that it is important to optimize patients' general condition before starting second-line treatment.
Several studies have underlined the importance of considering PS for clinical practice. Lilenbaum et al. found high prevalences of poor PS (Eastern Cooperative Oncology Group [ECOG] PS 2-4, corresponding to a KI < 80%) in lung cancer patients of 34% when rated by physicians and of 48% when rated by patients . Blagden et al. showed that both patient-and physician-rated PS scores reflected duration of survival and disease stage; physician-rated scores were only marginally more predictive of survival . A recent study in more than 26,000 NSCLC patients has shown that PS is an independent risk factor for patient OS .
No previous study has looked at changes in PS during second-line treatment. Only one post-hoc analysis of data from a first-line treatment study has been published by Sculier et al. . In their study, 485 patients received three cycles of triple-agent treatment with cisplatin, gemcitabine and ifosfamide. 25% of patients who had a poor baseline KI of 60% to 70% achieved clinical improvement, defined as high KI of ≥ 80%. These findings are consistent with the improvement rate of 20.8% in patients with poor baseline KI we observed with second-line pemetrexed treatment.
We additionally looked at patient-rated HR-QoL, using the EQ-5D self-estimation instrument. The EQ-5D was chosen because it is a simple, validated and commonly used questionnaire which can be completed within a short time . The disease-specific, multidimensional tools commonly used in lung cancer studies, such as the Lung Cancer Symptom Scale (LCSS) used by Hanna et al.  in a study which compared second-line treatment with pemetrexed versus docetaxel, were too complex for use in a routine clinical practice setting. The EQ-5D has been validated into 36 official languages (http://www.euroqol.org), including the German language . We found that on average, baseline EQ-5D ratings were at least maintained, consistent with the physician-based KI ratings. Mean EQ-5D index and EQ-VAS scores even showed small improvements after the second treatment cycle, i.e., as early as 6-weeks after the initiation of pemetrexed treatment. The EQ-5D instrument has rarely been used in NSCLC patients so far. Grutters et al. have recently applied the EQ-5D in patients surviving lung cancer. They found that patients with severe adverse events (dyspnoea grade ≥ 3) during treatment had statistically significantly lower EQ-5D index scores than patients without severe adverse events , indicating that the instrument is sensitive enough to detect health status changes during chemotherapy.
In our study, second-line pemetrexed treatment was well tolerated by most patients. Two of 516 patients (0.4%) died due to drug-related toxicity. There were no signs of cumulative toxicities. On the contrary, Grade 3/4 toxicity rates decreased after repeated treatment cycles. Patients in our study received a higher number of treatment cycles than in the randomized phase III trial of second-line pemetrexed treatment by Hanna et al.  which allowed for treatment up to disease progression (median 5 vs. 4 cycles). In our study, 20% of patients received at least 9 cycles and 5% continued pemetrexed treatment after the end of study. Considering in addition that most patients maintained or improved their PS and HR-QoL, our data indicate that a high number of treatment cycles do not impair patients' quality of life.
The Grade 3/4 toxicity profile, recorded by solicited questioning of specific non-hematologic toxicities, was largely comparable to the unsolicited toxicity rates during pemetrexed treatment published by Hanna et al. (e.g. febrile neutropenia 1.9% vs. 1.9%, nausea 3.1% vs. 2.6%, vomiting 1.5% vs. 1.0%, rash 1.4% vs. 0.8%). However, we found higher rates of Grade 3/4 asthenia/fatigue (15.9% vs. 5.3%) and Grade 3/4 neutropenia (8.7% vs. 5.3%).
Further, we found no evidence that Grade 3/4 toxicity rates were higher in elderly than in younger patients. These findings are well in line with a subgroup analysis of toxicities in the Hanna study, performed by Weiss et al., which also did not find any significant difference in toxicities between older and younger patients ( ≥ 70 vs. < 70 years) . A previous metaanalysis of 3 studies on pemetrexed first-line treatment in 764 NSCLC patients also concluded that pemetrexed or pemetrexed-based combinations produced similar treatment effects in older and younger patients ( ≥ 65 and < 65 years) and appeared to be well tolerated in the elderly population . Thus, it cannot be stated in general that treatment toxicity is higher in elderly patients; this may depend on the type of treatment used. A recent retrospective study by Chrischilles et al. evaluated chemotherapy use and adverse events during treatment of advanced NSCLC in routine clinical practice and concluded that toxicity was increased in elderly patients, but his study mainly looked at first-line treatment with platinum-based combinations . One additional result of this US study was that physicians used carboplatin-based combinations more frequently than cisplatin-based combinations (65.3% vs. 10.4%). Our German and Austrian data were fully consistent with these findings: Physicians used carboplatin-based combinations as first-line treatment in approximately two thirds of patients (carboplatin 62.8%, cisplatin 25.9%). Administration of carboplatin is more convenient and less emetogenic than cisplatin , and carboplatin has been associated with less toxicity . This may explain why the latter is preferred in routine clinical practice.
Our study has limitations. First, the non-interventional design to observe patients in routine clinical practice prohibits any definite conclusion on the efficacy of second-line pemetrexed treatment. In particular, tumor response data are of limited value and cannot be compared to clinical trial data because physicians may not have performed standard radiologic assessments or classified response in accordance with the standard response evaluation criteria in solid tumors (RECIST). Second, our toxicity data cannot be directly compared with toxicity data from clinical trials, because the collection of toxicities differed substantially: only a few solicited toxicities were specifically asked for via tick-boxes; no complete lists of Common Terminology Criteria for Adverse Events (CTC-AE) were handed out to the investigators. Patients may also have reported e.g. fatigue more frequently due to the solicited questioning for specific toxicities. The validity of these data may therefore be questioned, although the majority of toxicity results were in line with previous clinical studies of pemetrexed . Further, we did not differentiate between in-and outpatient treatment although it can be assumed that the majority of patients received the 10-minute infusion either in outpatient practice or in the ambulance of the hospital. Finally, only half of the patients returned completed EQ-5D questionnaires. The internal validity of our QoL data may therefore be compromised because the missing data may be informative and not at random. The major strength of our study is that we looked at second-line treatment in a large sample of patients as routinely treated in clinical practice where there is a particular lack of data . Clinical trials in NSCLC patients are often criticized for rarely enrolling elderly patients or patients with poor PS. In the randomized phase III trial by Hanna et al., 15.1% of patients in the pemetrexed arm were ≥ 70 years old, and 10.6% of patients had a poor baseline ECOG-PS of 2 [6, 30]. In our study, 36.1% of patients were ≥ 70 years of age, and 37.8% of patients had a poor baseline KI of < 80%. Despite the higher proportions of elderly and poor-performance patients in our study, overall disease control rates with pemetrexed were comparable to those found by Hanna et al. (60.3% vs. 54.9%).