鉴于此，来自日本横滨的 Katsuya 教授及其同事开展了一项回顾性研究，该研究主要分析非小细胞肺癌术后的复发模式及其复发间隔时间，并探讨肺癌的复发模式，其结果发表在 EJCTS 杂志上。
该研究纳入 2005～2007 年间在横滨财团隶属的 9 所附属医院胸外科接受肺癌根治术的 829 例非小细胞肺癌患者的资料。 这部份患者术后均接受规则的随访，统计分析患者复发的最大风险时间，及其与患者临床资料的相关性。
该研究发现，非小细胞肺癌患者术后 6～8 月其风险率曲线出现第一个高峰，术后第二年出现第二个高峰。矫正后显示男性患者风险率曲线出现最大峰值为术后 6～8 个月。女性患者风险率曲线出现最大峰值为术后 22～24 月，相比男性患者峰值延迟 16 个月。风险曲线的峰值时间不受肺癌组织学类型，病理分期和年龄的影响。
Lung cancer is the leading cause of cancer-related death in Japan and many countries around the world, and non-small-cell lung cancer (NSCLC) accounts for 75–85% of all cases [ 1 ]. Surgery is the mainstay of treatment for early-stage NSCLC [ 2 ]. Unfortunately, local or distant recurrence (or both) often develops even in patients with early disease who undergo complete resection. Although adjuvant platinum-based chemotherapy improves survival, the benefits are modest. Several recent studies have evaluated combinations of chemotherapy and biological targeted therapies [ 3 ].
Alternative statistical methods have been used to analyse the risks of recurrence in relation to the time after surgery in a population of patients. Cumulative incidence curves, which indicate the cumulative risk of incurring an event over time, are used most frequently. Another variable used to express risk is the median interval from surgery to recurrence. In NSCLC, the median time from surgery to local failure is 13.9 months and that to distant failure is 12.5 months in patients who have recurrence [ 4 ]. However, such methods do not provide direct information about changes in event probabilities over the course of time (i.e. event dynamics), which can be estimated by calculating event-specific hazard rates over the follow-up time interval [ 5 ].
At present, evidence-based methods for postoperative follow-up remain to be established, and guidelines recommended by major organizations in western countries differ considerably. Our study was designed to visually represent recurrence patterns after surgery for lung cancer with the use of event dynamics and to clarify postoperative follow-up methods based on the times of recurrence.
PATIENTS AND METHODS
A total of 829 patients (538 men, 291 women) with NSCLC who underwent complete pulmonary resection from January 2005 through December 2007 in 9 hospitals affiliated with the Yokohama Consortium of Thoracic Surgeons (Yokohama City University Hospital and affiliated hospitals) were studied. Preoperative staging investigations were routinely performed using computed tomographic (CT) scans of the chest and abdomen. Magnetic resonance imaging (MRI) scans or CT scans of the brain and nuclear medicine scans of bone were done to exclude possible distant metastases. During the study period, positron emission tomography (PET) was generally not performed for disease staging. However, PET was performed along with the standard examinations in selected patients. A single primary tumour was diagnosed in all patients, and no patient had a history of lung cancer (excluding those with multicentric cancers).
Patients who died in the immediate postoperative period (within 30 days after surgery or during the initial hospitalization) were excluded. Postsurgical pathological tumour–node–metastasis (TNM) staging was performed according to the guidelines of the American Joint Cancer Committee (AJCC), 6th edition.
The postoperative follow-up schedule consisted of a clinic visit every 3–6 months through the fifth year, and annually thereafter. Follow-up evaluations included physical examination, serum tumour markers, chest radiography and CT scanning of the chest and abdomen. In general, chest radiography was done every 3–6 months for the first 2–3 years and then at 6-month intervals or annually thereafter. CT scans were obtained every 6 months in the first 2–3 years after resection and annually thereafter during follow-up. In patients who had signs or symptoms of recurrence, CT scans of the chest and abdomen, brain MRI, bone scintigraphy and PET scanning were performed as required.
Recurrence was diagnosed on the basis of the results of physical examinations and diagnostic imaging and was confirmed by pathological examination of biopsy specimens if necessary. The date of recurrence was defined as the date of confirmation of recurrence based on clinical and radiological findings. Local recurrence was defined as tumour recurrence in the ipsilateral lung or lymph nodes, and distant metastasis was defined as tumour recurrence in the contralateral lung or lymph nodes or in a distant organ such as the brain, liver or bone (or both). Second primary lung cancers (diagnosed when a new lung tumour with different histological features was detected on standard histological and immunological studies or when the clinical scenario was more consistent with a new primary tumour than a local recurrence) were excluded. Time to treatment failure was defined as the interval between the date of surgery and the date of disease recurrence (local recurrence or distant metastasis). Only first events were considered.
Event dynamics were studied by using life-table methods to estimate the discrete hazard rate for the considered event, i.e. the conditional probability of the event occurring within a defined time interval, given that the patient did not previously have the event at the beginning of the interval [ 6 ]. A discretization of the time axis in 2-month units was applied, and all hazard rate levels were measured as ‘events/patients at risk per 2-month interval’. Because the hazard rate estimates showed some instability owing to random variation, a kernel-like smoothing procedure was used, and the smoothed curve was graphically represented to facilitate understanding of the underlying pattern [ 7 ].
In addition to the kernel smoothing approach with discrete hazards, a flexible piecewise exponential regression model was used to obtain smoothed hazard estimates [ 8 ]. Potential covariates were sex, histological type, pathological stage and age. Natural cubic splines were used to model the time dependence of each variable with internal knots placed equidistantly within the month range (0–72 months). The number of knots, which corresponded to the number of basic functions between 4 and 10, was chosen according to the Akaike Information Criteria (AIC).
Patient, tumour and treatment characteristics are given in Table 1 . At a median follow-up of 65.6 months (range, 1–98.7 months), disease recurrence developed in 274 patients (128 with only local recurrence and 146 patients with only distant metastasis or local recurrence plus distant metastasis).
Patient, treatment and tumour characteristics ( n = 829)
|Characteristics||No. of patients (%)|
|Tumour size (mm)|
|Right upper lobe||275 (33.2)|
|Right middle lobe||64 (7.7)|
|Right lower lobe||181 (21.8)|
|Right lung, NOS||1 (0.1)|
|Left upper lobe||202 (24.4)|
|Left lower lobe||105 (12.7)|
|Left lung, NOS||1 (0.1)|
|Hybrid VATS||150 (18.1)|
|Visceral pleural invasion|
|Squamous cell||208 (25.1)|
|Large cell||41 (4.9)|
|NSCLC, NOS||62 (7.5)|
VATS: video-assisted thoracoscopic surgery; NSCLC: non-small-cell lung cancer; NOS: not otherwise specified; NS: not stated.View Large
We first analysed the hazard rate for treatment failure in all 829 patients. The resulting curve (Fig. 1 ) displayed an initial surge in the hazard rate, which peaked about 6–8 months after surgery. Another distinct peak was noted at the end of the second year of follow-up. A small peak was found even 5 years after surgery.Figure 1:
Smoothed hazard rate estimates for first event. LR: local recurrence; DM: distant metastasis.
On non-parametric kernel smoothing, the hazard rate curve displayed an initial sharp, high peak 6–8 months after surgery for men (Fig. 2 ). In women, several small peaks were noted during the first year, and the highest peak occurred 22–24 months after surgery, which was about 16 months later than the peak in men.Figure 2:
Smoothed hazard rate estimates for first event (LR and DM) in 538 men and 291 women. LR: local recurrence; DM: distant metastasis.
As for histological type, squamous cell carcinomas had a higher risk of recurrence than adenocarcinomas during follow-up (Fig. 3 A). Squamous cell carcinomas showed a sharp peak in the hazard rate in the first year, followed by four or five small peaks. In contrast, the hazard rate for adenocarcinomas showed a different pattern. After a gradual increase during the first 6–14 months, the hazard rate moderately decreased subsequently.Figure 3:
( A ) Smoothed hazard rate estimates for first event (LR and DM) according to histological type. ( B ) Smoothed hazard rate estimates for first event (LR and DM) according to histological type and sex. LR: local recurrence; DM: distant metastasis.
Although the hazard rate for the first event after surgery differed according to histological type at 1 year, there was only a slight difference in the timing of the first peak and the second peak of recurrence. When the hazard rate curves were compared between men with squamous cell carcinoma and those with adenocarcinoma, the first maximum peak of recurrence was found 6–8 months after surgery in both groups. When women with squamous cell carcinoma were compared with women with adenocarcinoma, the first maximum peak was noted at 22–24 months in both groups (Fig. 3 B).
As for pathological stage, as expected, the absolute magnitude of the recurrence peaks was higher in patients with advanced disease in both sexes, but the maximum peak of recurrence was found 6–8 months after surgery in both groups (Fig. 4 ). With respect to age, the hazard rates and timing of recurrence did not distinctly differ between patients younger than 70 years and those 70 years or older.Figure 4:
Smoothed hazard rate estimates for first event (LR and DM) according to pathological stage. LR: local recurrence; DM: distant metastasis.
In the present study, we examined recurrence dynamics in patients with NSCLC and found several recurrence peaks during postoperative follow-up. We confirmed that the hazard of postoperative recurrence peaked at certain times after surgery and was not always constant. A structured, multiple-peak pattern of recurrence risk is not a new finding and has been reported in patients with cancer arising in organs other than the lung, such as breast cancer [ 9 ] and head and neck cancer [ 10 ]. Demicheli et al . [ 6 ] and Kelsey et al . [ 11 ] reported that patients with NSCLC showed a bimodal recurrence pattern similar to that in patients with breast cancer. Despite differences in various patient characteristics, including race, sex and histological types of cancer, it is extremely interesting that we obtained results similar to those of their study. The fact that the presence of cancer in multiple organs including the lung is apparently associated with a certain pattern of postoperative recurrence casts doubt on the conventional concept that tumour cells continue to proliferate in a disordered manner, leading to disease progression. To date, many concepts have been proposed to convincingly explain the clinical behaviour of breast cancer, such as tumour homeostasis, tumour dormancy and surgery-related enhancement of metastatic growth [ 12 ]. The fact that the first peak of recurrence occurs within 1 year after surgery suggests that surgical invasion disrupts homeostasis, accelerating the proliferation of dormant tumour cells. The second and subsequent peaks of recurrence found in our study can be explained by the hypothesis that residual tumour cells proliferated and micrometastases developed after entering a transient state of dormancy. Our findings suggested that tumour cells gradually proliferate after passing through a relatively long dormancy period in certain types of lung cancer. At present, however, the detailed mechanisms underlying the hypothesis of tumour dormancy remain to be clarified.
Another interesting finding in our study was that the hazard rate and the peak times of recurrence differed considerably between men and women. In men, the first peak in recurrence appeared 6–8 months after surgery, and the hazard rate then showed a downward sloping tendency. In women, however, there was only a small peak within the first year after surgery. The hazard rate then gradually increased to reach its peak value 22–24 months after surgery. Our results also suggested that recurrence of cancer peaks during the first year in male patients, whereas female patients lack such a large peak during the first year and instead have two small peaks during the second year after surgery. Initially, we assumed that the timing of recurrence was somewhat affected by histological type because the peak timing of recurrence was later for adenocarcinoma than for squamous cell carcinoma. However, the hazard rate curve did not differ markedly between men with squamous cell carcinoma and those with adenocarcinoma. In addition, the highest peak of recurrence in women was noted 22–24 months after surgery for both histological types. We therefore speculated that the delayed time of peak recurrence of adenocarcinoma might be attributed to the high rate of adenocarcinoma in women (∼85%). The timing of recurrence thus appeared to be more strongly influenced by sex than by histological type. Moreover, pathological stage and age did not affect the peak timing of the hazard curve in either sex. These findings suggested that the recurrence dynamics of lung cancer show a bimodal characteristic pattern and that the timing of recurrence after surgery is probably sex-dependent. However, definitive evidence supporting a sex-dependent difference in recurrence patterns has yet to be obtained. Further prospective studies in larger numbers of patients are needed.
As for postoperative follow-up of lung cancer, many previous studies have examined methods potentially contributing to overall survival [ 13 ]. Westeel et al . [ 14 ] reported that asymptomatic patients in whom recurrence was detected on intensive imaging studies after surgery for NSCLC had better survival than symptomatic patients with recurrence. On the other hand, Virgo et al . [ 15 ] reported no significant difference in the detection of recurrence or in outcomes between patients who were ‘intensively’ followed up and those who were ‘non-intensively’ followed up. Similarly, Younes et al . [ 16 ] reported that disease-free survival and median survival did not differ significantly between patients who were followed up according to a routine follow-up protocol and those who were followed up based on symptoms. They therefore concluded that intensive screening of asymptomatic patients was unwarranted from the viewpoint of cost-effectiveness. At present, there is thus no clear-cut basis to recommend aggressive routine screening, and it remains unclear whether the early detection of recurrence contributes to improved outcomes. Therefore, the optimal follow-up protocol for postoperative patients with lung cancer remains to be established [ 17 ]. However, history taking and physical examinations should be regularly performed to facilitate the detection of postoperative complications on an outpatient basis, an understanding of the patient’s condition and the provision of mental support. Although guidelines proposed by major organizations in Europe and North America have consistently recommended intensive follow-up during the first 2 years after surgery, followed by annual follow-up examinations from postoperative year 3 or year 5, major differences exist among current guidelines, including who should perform follow-up and when and what follow-up examinations should be performed [ 18–22 ] (Table 2 ).
Current recommendations for surveillance after curative-intent therapy of NSCLC
|Organization||Summary of recommendations||Classification of recommendations|
|ACCP [ 18 ]||Surveillance by clinical examination and chest radiography or CT should be performed every 6 months for 2 years and then yearly for patients with good performance status and pulmonary function||Grade 1C|
|ESMO [ 19 ]||A follow-up visit every 3–6 months is recommended during 2–3 years, less often—e.g. annually—thereafter||III, B|
|For follow-up, history and physical examination, chest CT and, to a lesser extent, chest X-ray are appropriate tools||III, B|
|NCCN [ 20 ]||History and physical examination with contrast-enhanced CT scan every 4–6 months for 2 years||2B|
|Then history and physical examination and non-contrast-enhanced CT scan annually||2B|
|ASCO ||For patients treated with curative intent, in the absence of symptoms, a history and physical examination should be performed every 3 months during the first 2 years; every 6 months thereafter through year 5; and yearly thereafter||None|
|For patients treated with curative intent, there is no clear role for routine studies in asymptomatic patients and patients in whom no interventions are planned|
|NICE [ 22 ]||Offer all patients an initial specialist follow-up appointment within 6 weeks of completing treatment to discuss ongoing care. Offer regular appointments thereafter, rather than relying on patients requesting appointments when they experience symptoms||None|
|Offer protocol-driven follow-up led by a lung cancer clinical nurse specialist as an option for patients with a life expectancy of more than 3 months|
|Ensure that patients know how to contact the lung cancer clinical nurse specialist involved in their care between their scheduled hospital visits|
ACCP: American College of Chest Physicians; ESMO: European Society of Medical Oncology; NCCN: National Comprehensive Cancer Network; ASCO: American Society of Clinical Oncology; NICE: National Institute for Health and Clinical Excellence.View Large
Recent studies have reported that the accuracy of imaging studies has improved and that CT is useful for follow-up, contributing to longer survival [ 23 ]. Progress in drug therapy, the development of new anticancer agents and the advent of molecular targeted agents have prolonged survival [ 24 ] and improved the quality of life (QOL) [ 25 ] of patients with advanced, recurrent NSCLC. In such patients, CT-based imaging studies should be aggressively performed at times of high risk of recurrence to facilitate the early detection and early treatment of recurrence and thereby most likely contribute to improved QOL and outcomes. As the methodology for personalized treatment of lung cancer gradually becomes more widely accepted, individually designed follow-up programmes based on the biological characteristics of tumours and risk factors for recurrence, rather than conventional follow-up in which standardized imaging studies are performed at predefined intervals in all patients, will most likely be required.
Based on the current situation and our results, hospital visitation programmes should be designed to focus on 6–8 and 22–24 months after surgery (i.e. the times of peak hazard rates of recurrence), and appropriate CT-based imaging studies should be performed at these times. In addition, because the peak times of recurrence differed between men and women, imaging studies should be planned according to sex to most intensively cover the period from 6–8 months during the first year after surgery in men and 22–24 months during the second year after surgery in women (Fig. 5 ). If CT examinations are performed every 6 months during the first 2 years after surgery, followed by once per year from 3 years onwards in accordance with the guidelines issued by the American College of Chest Physicians (ACCP) and the European Society for Medical Oncology (ESMO), CT would be performed four times during the first 2 years and seven times during the first 5 years. Potential cost-benefits, patient satisfaction and postoperative treatment strategies should be taken into account, and the numbers of hospital visits and CT examinations should be designed individually. In this respect, the use of recurrence dynamics allows the times of peak recurrence to be visualized, potentially allowing more efficient follow-up surveillance.Figure 5:
Postoperative follow-up schedules for men (left) and women (right). CT: computed tomography.
An important limitation of our study is that it was retrospective and therefore subject to the effects of lead-time and length-time bias. All hazard rate levels were measured at 2-month intervals in our study. However, the timing of the first event largely depends on the timing of imaging studies or hospital visits. Obviously, follow-up and analysis at shorter assessment intervals would be needed to more precisely estimate the risk of recurrence (Fig. 6 ). There is no doubt that a randomized, prospective study should be performed to evaluate whether follow-up surveillance based on recurrence dynamics is more useful than conventional protocols for postoperative follow-up in terms of the early detection of recurrence, survival outcomes, health-related QOL, cost-effectiveness factors and other factors. At present, postoperative follow-up strategies should be based on currently recommended follow-up programmes, give adequate consideration to cost-effectiveness and be modified as required to meet the needs of individual patients.Figure 6:
Time to first event and numbers of patients still at risk.
The timing of recurrence after surgery for lung cancer was characterized by a bimodal pattern, and the times with the highest risk of recurrence were suggested to differ between men and women. Postoperative follow-up strategies should be based on currently recommended follow-up programmes, take into account the recurrence patterns of lung cancer, and be modified as required to meet the needs of individual patients.