Fax: (416) 946-4568
Evidence-based guidelines for following stage 1 seminoma
Article first published online: 16 APR 2007
Copyright © 2007 American Cancer Society
Volume 109, Issue 11, pages 2248–2256, 1 June 2007
How to Cite
Martin, J. M., Panzarella, T., Zwahlen, D. R., Chung, P. and Warde, P. (2007), Evidence-based guidelines for following stage 1 seminoma. Cancer, 109: 2248–2256. doi: 10.1002/cncr.22674
- Issue published online: 18 MAY 2007
- Article first published online: 16 APR 2007
- Manuscript Accepted: 12 FEB 2007
- Manuscript Revised: 8 FEB 2007
- Manuscript Received: 9 JAN 2007
- evidenced-based guidelines;
- stage 1;
- testicular germ cell tumor
The authors developed evidence-based guidelines for a follow-up schedule after orchiectomy for stage 1 seminoma. Required investigations, frequency of assessment, overall duration of follow-up, and management strategies were identified.
A systematic review of the literature was performed of prospective studies in stage 1 seminoma. Studies published after 1980 were considered eligible for inclusion. Data extracted included relapse-free rates, number of patients at risk, and relapse locations. Five strategies were identified: Surveillance, Extended-Field Radiotherapy, Para-aortic Radiotherapy, and either 1 or 2 cycles of Carboplatin Chemotherapy. For each strategy, Kaplan-Meier relapse-free estimates were used to calculate weighted-mean cumulative hazards of relapse over time. These were used to calculate semiannual weighted-mean relapse hazards.
Seventeen prospective studies with a total of 5561 patients were identified. Actuarial data on relapse was available in 5013 (90.1%) patients, and 92.9% of all relapses had location data reported. Annual hazard rates for relapse were determined.
Evidence-based recommendations for follow-up frequency based on risk of relapse were formulated. The authors suggested 3 times per year when the risk is >5%, 2 times per year when the risk is 1% to 5%, and annually until the risk is <0.3%. Investigations should reflect location(s) at risk of relapse and include computed tomography of the abdomen and pelvis for surveillance and adjuvant carboplatin, whereas for para-aortic radiotherapy, pelvic computed tomography alone is required. These recommendations offer the possibility of maximal patient convenience and optimal healthcare resource allocation without compromising disease control. Cancer 2007. © 2007 American Cancer Society.
Testicular germ cell tumors (GCTs) are uncommon, but they are the most common solid malignancies in males who are between 20 and 35 years of age. The incidence of GCTs has doubled in the past 30 years, for reasons which are, as yet, unclear.1
Approximately 40% to 60 percent of GCTs are pure seminoma, and the majority of patients (70% to 80%) present with clinical stage 1 disease.2 With modern management that includes effective salvage treatment, 5-year overall survival exceeds 99%.
Before the 1980s, standard adjuvant management after radical inguinal orchidectomy was extended-field radiotherapy (EFRT), including the para-aortic and ipsilateral pelvic lymph nodes. However, because of long-term toxicity concerns, alternative management strategies, including surveillance (S), para-aortic radiotherapy (PART), and single-agent carboplatin chemotherapy have been explored. Although long-term follow-up data are not available for all of these strategies, disease-specific survival does not appear to have been compromised.
Follow-up after definitive tumor management is necessary not only for early detection and treatment of relapses but also for management of associated conditions and treatment-related toxicities. With many different management approaches available to the clinician, an unresolved issue is how best to monitor patients with stage 1 seminoma for disease relapse. Prolonged follow-up with overly frequent imaging is clearly undesirable, as this management strategy exposes patients to potential hazards associated with diagnostic radiologic procedures. As extent of disease at relapse is directly related to modality of treatment and its associated toxicity, an optimal follow-up strategy may allow early detection of relapse with minimal disease burden. The purposes of this study were to formulate evidence-based guidelines for frequency and duration of follow-up in this group of patients and imaging protocols for different management strategies.
MATERIALS AND METHODS
A systematic review of the literature was performed for prospective studies that included phase 2 studies, phase 3 clinical trials, or prospectively maintained databases with uniform management policy. Only studies published after 1980 were considered eligible, as prior to this period, computerized tomography access would have been rare, and, hence, staging accuracy was potentially an issue.
Medline, Embase, and Cochrane databases were searched with the aid of a health sciences librarian by using the exploded keyword “Germinoma” (which includes seminoma and the pre-1994 medical subject heading of dysgerminoma) combined with “Radiotherapy”, “Drug Therapy”, “Surveillance”, or “Watchful Waiting.” Furthermore, abstracts from 2000–2005 of meeting proceedings of the American Society for Clinical Oncology (ASCO), European Society for Therapeutic Radiology and Oncology (ESTRO), American Society for Therapeutic Radiology and Oncology (ASTRO), and the European Cancer Conference (ECCO) were examined, as were key oncology textbooks. The 706 abstracts identified by this search were reviewed by 2 authors (JM and DZ). Any article that was considered to report outcomes for stage 1 seminoma was obtained and read. Reference lists were scrutinized, and any further potentially relevant articles were also obtained.
Studies (or specific patient groups, if defined) were excluded for the following reasons.
There was no explicit statement of the prospective nature of the study.
A more up-to-date publication from the same patient population was available.
Each article was reviewed by using a predetermined and piloted Microsoft Excel XP (Microsoft, Redmond, Wash) spreadsheet into which information was directly entered. All studies had to explicitly define the nature of management strategy as well as the follow-up schedule. A negative computerized tomogram of the abdomen and pelvis and a chest x-ray were considered to be minimum staging investigations. Information extracted included management strategy, minimum, median, and maximum follow-up, number of patients treated, number at risk during follow-up, number of relapses for each treatment, time-to-relapse for each patient, median, minimum and maximum time to relapse, and locations of relapse.
Three-year and 5-year recurrence-free rates (RFRs) were estimated for each study. Within each management modality, a weighted average of these statistics was calculated. Individual study weights were calculated by using the reciprocal of the variance of the 3-year and 5-year RFRs. The variance of the estimated pooled 3-year and 5-year RFRs was calculated by using the reciprocal of the sum of the weights across each study.3
A chi-square test of heterogeneity, with degrees of freedom equal to number of studies minus 1, was calculated under the null hypothesis of an identical effect in every study within each management modality.4 However, as there were only a few studies within each modality, the test may not have been sensitive to detection of statistically significant heterogeneity when it existed.4
Nonetheless, a fixed-effects model was assumed, ie, it was assumed that the estimated 3-year and 5-year RFRs for each study within each of the 5 management strategies estimated the same true effect and that study differences were attributable to random error.
The hazard function was implemented to graphically depict the risk of recurrence over time, as it is an instantaneous measure of risk. Also, as an instantaneous measure of risk, the hazard is not necessarily always increasing or decreasing.
Semiannual hazard rates were calculated by using actuarial RFRs coupled with the fact that the negative of the natural logarithm of the RFR equals the cumulative hazard.5 So, for example, the hazard at 2 years was determined by subtraction, the cumulative hazard at 2 years minus the cumulative hazard at 1.5 years. Semiannual hazard rates were plotted against time for each study arm. To obtain a better sense of the nature of the relation between these 2 variables, a locally weighted least squares regression (lowess) smoothing curve was fit to the data.6 Several smoothing spans (ie, the proportion of points that influence the smooth at each value) were used to produce smoothing curves. A smoothing span of 0.45 was thought to provide a reasonable balance between smoothness and inherent association. For each management strategy, the different curves were compared to ensure relative homogeneity between studies. The data for each modality were then combined into a single graph with a superimposed lowess smoother relating the semiannual hazard with time.
Given that the RFRs of interest were generally close to unity, and the knowledge that the Greenwood formula is known to underestimate the variance in this context, an alternative expression from Peto et al was used to calculate the variance, and thus the confidence intervals for the RFR effects at 3 years and 5 years.5, 7
For strategies with a very low risk of relapse (ie, EFRT, PART, C1, and C2), pooled median follow-up was calculated by using linear interpolation (using numbers of patients at risk) of the 2 middle-most 6-monthly follow-up intervals. For Surveillance, the pooled median follow-up was estimated from the sum of the products of initial patient numbers (as numbers at risk subsequently were considered less reliable) and median follow-up for individual study arms divided by the total number of patients.
For PART, a chi-square test (with continuity correction) of association was performed between location of failure (pelvic vs nonpelvic) and time of failure (cut point at the median time to relapse). All analyses were conducted using R.8
Twenty-six eligible study arms were identified in 17 reports (Table 1).9–27 Three were randomized controlled trials, 9 were phase 2 trials, and 5 were from prospective institutional databases. For purposes of summary data generation, these studies included a total of 5561 patients (Table 2). In 1 study, actuarial data were available from an older publication with only summary data available in a later abstract.9, 10 The more mature information is included for summary figures in Table 2, and actuarial data were used for the remainder of the analysis. One study arm from a randomized trial was not included in the analysis because it did not explicitly define the modality used for all patients who recurred, stating only “radiotherapy” when both extended-field radiotherapy and para-aortic radiotherapy were used.11 A published individual patient data meta-analysis of surveillance studies included updated time-to-recurrence data from 3 prospective cohorts without reporting the location of failure.12 Hence, this meta-analysis was used for time-to-recurrence data, but the original publications were used for the rate of failure in each location.9, 10, 13
|Ref||Author||Year||Strategy||Median follow- up||No, of patients||No. of relapses||Actuarial relapse data available||Summary location data available||Actuarial location data available||Notes|
|23||Schulz||1984||Extended Field RT||48||424||13||N||Y||N||National prospective database|
|24||Fossa||1999||Extended Field RT||54||242||9||Y||Y||Y||Randomized Trial of EFRT v PART|
|9||Oliver||1994||Extended Field RT||51||79||4||Y||Y||Y||Multiarm phase 2 trial|
|10||Oliver||2001||Extended Field RT||208||78||5||N||N||N||Update of Oliver 1994, Summary figures|
|22||Jones||2005||Extended Field RT||61||72||5||Y||Y||Y||Randomized Trial of RT 20 v 30Gy|
|14||Warde||2005||Extended Field RT||120||283||14||Y||Y||Y||Institutional prospective database|
|21||Classen||2004||Para-aortic RT||61||675||26||Y||Y||Y||National prospective database|
|20||Niazi||2005||Para-aortic RT||75||71||1||Y||Y||Y||Institutional phase 2 trial|
|13||Horwich||1992||Surveillance||62||103||17||Y||Y||N||Institutional phase 2 trial|
|19||Von der Maase||1993||Surveillance||48||261||49||Y||Y||N||National phase 2 trial|
|18||Daugaard||2003||Surveillance||60||394||69||Y||Y||N||Institutional phase 2 trial|
|17||Aparicio||2003||Surveillance||52||143||23||Y||N||N||National multiarm phase 2 trial|
|16||Choo||2005||Surveillance||145||88||17||Y||Y||N||Institutional phase 2 trial|
|15||Aparicio||2005||Surveillance||34||100||6||Y||N||N||National multiarm phase 2 trial, risk adaptive|
|Carboplatin x2||34||214||7||Y||Y||N||Treatment allocation|
|12||Warde||2002||Surveillance||76||412||84||Y||N||N||Includes update time to relapse data from ref 9,13,27|
|14||Dieckmann||2000||Carboplatin x1||48||93||8||Y||Y||Y||Sequential phase 2 trials.|
|11||Oliver||2005||Carboplatin x1||48||560||29||Y||Y||N||Randomized trial of C1 v RT|
|25||Krege||1997||Carboplatin x2||28||43||0||Y||NA||NA||Institutional phase 2 trial|
|13||Argirovic||2005||Carboplatin x2||48||163||3||Y||Y||N||Institutional phase 2 trial|
|Surveillance||Extended field RT||Para-aortic RT||1 Cycle of carboplatin||2 Cycles of carboplatin|
|No. of study arms||8||5||4||3||6|
|Total no. of patients||1558||1100||1535||799||569|
|Total no. of relapses||263||45||52||38||13|
|Crude relapse rate, %||16.9||4.1||3.4%||4.8||2.3|
|3-y relapse-free rate, % (95% CI)||84.7 (80.4–89.0)||96.6 (93.5–99.7)||96.6 (94.8–98.4)||94.4 (91.9–96.8)||97.5 (94.5–100)|
|5-y relapse-free rate, % (95% CI)||83.0 (78.0–88.0)||95.3 (91.6–99.0)||96.5 (94.5–98.6)||94.1 (90.3–97.9)||97.5 (93.4–100)|
|Actuarial 5-y relapse rate, %||18.6||4.8||3.6||6.1||2.3|
|Pooled median FU||77 mo||68 mo||65 mo||50 mo||40 mo|
|Relapse time; actuarial data available, %||100||61.5||100||84.9||99.3|
|Relapse location data available as percentage of all relapses||90.5||100||100||97.4||76.2|
|Relapse time and location reported together as a percentage of all relapses||0||71.7||100||21.6||7.7|
By using both 3 and 5-year relapse free rates as effects, nonsignificant chi-square values (all P > .1) were obtained for PART, EFRT, Surveillance and C1. Two studies in the C2 cohort had a relapse-free rate of 1, which indicated no events, a standard error of 0, and, hence, an inability to perform chi-square testing. The remaining 4 studies had P > .2, and their confidence interval included 1. Hence, all studies were eligible for inclusion in the pooled analysis.
Summary pooled data is presented in Table 2. The risk of recurrence at 5 years was highest with the surveillance strategy. The 95% confidence intervals for the remaining 4 strategies all overlap. The actuarial data for time to relapse as well as location of relapse were available in most cases.
Recurrence Hazard Rates
Figure 1 shows the semiannual hazard rates for the 5 management strategies. The vertical lines represent the earliest point where the smoothed curves are first less than an annual hazard rate of 5% and 1% respectively. A similar line is also drawn for 0.3% and takes into account sporadic late recurrences and number of patients at risk >100.
Individual time-to-relapse information grouped with location data was available for the majority of patients treated with extended-field RT and para-aortic RT. Figure 2 displays this time and location of relapse data for those management strategies. There is no statistically significant interaction between locations of relapse over time for para-aortic radiotherapy (pelvis vs nonpelvic failure, before and after median time to relapse of 14 months. Chi-square = 0.16; 1 degree of freedom; P = .69.) Table 3 displays the crude risk of relapse by location for the 5 management strategies. Actuarial risk-of-relapse information is also included for extended-field radiotherapy and para-aortic radiotherapy. Considering a stated relapse site of “stage 2 not otherwise specified” as representing relapses primarily in the retroperitoneum, and “stage 3 not otherwise specified” as representing mediastinal relapse, we estimated that locations of relapse fell into 4 categories: abdomen, pelvis, chest, and palpable lymph nodes (groin and neck). For surveillance, all 4 locations have a >1% crude risk of failure. For 1 and 2 cycles of carboplatin, abdomen and pelvis combined have a 4.1% and 1.8% crude failure risk. For PART, the pelvis has a 2.5% actuarial failure risk, and for EFRT, there is an actuarial 2.9% risk of failure in the chest and 1.7% in palpable lymph nodes. All other failure locations were at <1% risk for the nonsurveillance management strategies.
|Surveillance||Extended field RT||Para-aortic RT||Carboplatin x1||Carboplatin x2|
|Crude RR||Crude RR||Actuarial RR||Crude RR||Actuarial RR||Crude RR||Crude RR|
|Abdomen||147 (10.3%)||3 (0.3%)||0.20%||13 (0.8%)||0.50%||8 (1.2%)||1 (0.2%)|
|Pelvis||14 (1%)||1 (0.1%)||0.20%||29 (1.9%)||2.50%||0||0|
|Stage 2 NOS||79 (5.5%)||0||—||0||—||20 (2.9%)||0|
|AP NOS||5 (0.3%)||5 (0.5%)||—||0||—||0||8 (1.6%)|
|Lungs||2 (0.1%)||6 (0.5%)||—||4 (0.3%)||—||0||1 (0.2%)|
|Mediastinum||2 (0.1%)||11 (1%)||—||7 (0.5%)||—||0||0|
|Stage 3 NOS||5 (0.3%)||3 (0.3%)||—||0||—||3 (0.4%)||0|
|Neck/SCF||0||10 (0.9%)||—||3 (0.2%)||—||0||1 (0.2%)|
|Groin||4 (0.3%)||5 (0.5%)||—||7 (0.5%)||—||0||0|
|Overall palpable lymph nodes||0.30%||1.40%||1.70%||0.70%||0.50%||0%||0.20%|
|Other||5 (0.3%)||3 (0.3%)||0.20%||7 (0.5%)||0.30%||3 (0.4%)||1 (0.2%)|
Stage 1 seminoma represents, in many ways, an ideal population in which to formulate evidence-based guidelines for follow-up policy. Long-term survival is excellent in this relatively young population, thus reducing the influence of competing risks. The quality of relapse reporting in the prospective literature is also conducive to an analysis of failure patterns.
The annualized hazard plots allow definition of cut points in the annual-relapse hazard, which can help to determine frequency of follow-up. A suggested follow-up frequency is described in Table 4. Surveillance is the only strategy that has an annualized hazard rate of >5%. We suggest 4 monthly appointments in this setting, and with such an approach at the Princess Margaret Hospital, there have been no seminoma-related deaths in the last 20 years and no increased risk of requiring chemotherapy for disease control compared with EFRT treatment.14 Hence, more frequent assessment would appear to not be beneficial. Six monthly review appointments are recommended when the annualized hazard rate is between 1% and 5%.
|Annual hazard rate, %||Frequency, y||Surveillance, y||Extended field RT, y||Para-aortic RT, y||Carboplatin x1, y||Carboplatin x2, y|
|0.3–1||1x||5th–10th||4th–6th||4th–6th||Limited data||Limited data|
|<0.3||Cease||After 10||After 6||After 6||Limited data||Limited data|
Defining the time period for annual review before ceasing follow-up all together is controversial. We have suggested doing so in the time interval during which the annualized hazard rate is between 0.3% and 1% and after which isolated later relapses are very rare. Relapse more than 6 years after radiation therapy and after 10 years on surveillance is extremely rare. In a population-based study of >1100 patients with seminoma, the incidence of such late failure was <0.5%.28 Although some patients will relapse beyond our recommended time of follow-up cessation, ongoing routine imaging of all patients to detect very small numbers of relapses is of doubtful value. Of the current management strategies, there is a relatively short experience with the use of carboplatin in prospective trials. In a letter, Oliver states that in his extensive phase 2 experience following 1 cycle of carboplatin, no relapses have been observed beyond 39 months.29 However, this data set is still maturing, and for this reason and until the long-term relapse risk with this strategy is established, we have not made recommendations beyond 3 years of follow-up. At the time of discharge from routine follow-up, it is important that patients are made aware of the need to present for any concerning symptoms at any stage in the future.
There is little evidence to provide guidance for investigations required. Clinically assessable regions such as the neck, supraclavicular fossa, inguinal region, scrotum, and contralateral testis should be examined at every visit. We recommend computed tomography of the pelvis ± abdomen if the cumulative crude relapse risk for failure in these locations is >1%. This will detect disease at an early stage and allow treatment with radiation therapy, thus sparing most patients the need for multiagent chemotherapy. Data from the Christie Hospital in Manchester where no routine evaluation of the pelvic nodes is carried out after para-aortic radiation alone have shown that the median size of the pelvic lymph nodes at time of detection of relapse was 5 cm (range, 2.5 cm to 9 cm).30 To detect relapse in the chest, we recommend chest x-ray rather than chest computed tomography, given the lower radiation dose and lack of evidence of a significant impact on prognosis of discovering metastases at an earlier stage. For surveillance and adjuvant carboplatin strategies, a chest x-ray at alternating visits initially, and then annually when the hazard rate is <1%, would reflect the lower risk of relapse at this site. Given the low frequency of an isolated pelvic relapse on surveillance, a similar intermittent approach with pelvic computed tomography imaging would also be reasonable.14 This may also be the case with adjuvant carboplatin, although detail is lacking in published results to date to allow this recommendation to be evidence based. Recommendations for each management strategy are summarized in Table 5.
|Investigation||Surveillance||Extended field RT||Para-aortic RT||Carboplatin x1||Carboplatin x2|
Legitimate concerns have been raised concerning the high radiation exposure that patients are subjected to while being followed with serial imaging. A conventional computed tomography scan of the abdomen and pelvis gives a radiation exposure of 6 mSv to 14 mSv, and patients are, therefore, exposed to a significant amount of radiation from computed tomography imaging over the course of follow-up after carboplatin or on surveillance.31, 32 Whereas, to date, there has been no direct study that shows harmful effects from the x-ray exposure associated with computed tomography scanning, concern about secondary malignancies has been raised in literature.32, 33 In 1 study, the risk of cancer death from a single computed tomography scan of the abdomen was estimated to be 12.5 in 10,000.32 This estimate may appear excessive, but certainly there is good evidence that low-dose radiotherapy is associated with some increase in second-occurrence cancer risk. Ongoing research is attempting to validate low-dose computed tomography or magnetic resonance imaging to further limit radiation exposure.34
The role of tumor markers in the follow-up of pure seminoma patients can be questioned. Because modern histopathological techniques have increased the accuracy of initial diagnosis, relapse with nonseminomatous elements is unusual. There is little robust evidence to support the routine use of tumor markers in addition to follow-up imaging, as there is unlikely to be significant diagnostic yield.
A potential limitation of this study is the arbitrary nature of cutoff times and frequencies that were used to derive recommendations summarized in Tables 4 and 5. Given that our final recommendations mirror, in large part, the protocols used in many prospective studies, there should be some familiarity and reassurance derived from the data presented. The data presented may allow individual centers to customize their own follow-up policies, building in their own perceptions of risk based on reported patterns of failure. A further limitation is the heterogeneous timing of follow-up appointments and the accompanying investigations performed in each study, which may be expected to alter the pattern-of-relapse detection. For most studies included, such differences were small. Furthermore, the routine, but judicious, use of targeted imaging and assessment should largely avoid the scenario of patients presenting with large volume tumors and potentially morbid disease that require multiagent salvage chemotherapy. Hence, our recommendations tend to be conservative, but they well within the spectrum of policies used in the prospective literature.
In this review, we have concentrated on the use of follow-up to detect relapse. There are other reasons to maintain patient follow-up, including detection and management of a contralateral testicular primary tumor or other coexisting conditions such as infertility and hormone deficiency. In addition, the detection and management of late toxicities of treatment, including second malignancies and cardiac disease, is of considerable importance. However, routine follow-up is likely not the optimal strategy to manage these issues, and as discussed by Raghavan in a recent review, programs for surveillance for likely late complications of therapy urgently need to be defined, as the population of cured germ cell tumor patients after chemotherapy and radiotherapy is rising rapidly.35
There are many options for management in stage 1 seminoma, and this review was undertaken to develop recommendations for appropriate follow-up policies depending on the management approach pursued. These recommendations, based on evidence currently available, may offer the possibility of maximal patient convenience and optimal healthcare resource allocation without compromising disease control.
The authors thank John Jackson for assisting with the systemic literature review, Anthea Lau for data management, Betty Tew-George for database entry, and Eleni Sachinidis for manuscript preparation
- 2Testicular cancer. In: Clinical radiation oncology. GundersonL, TepperJ, eds. Philadelphia: Churchill Livingstone; 2000: 844–862., , .
- 3Biostatistics in clinical medicine. 3rd ed. New York: McGraw-Hill; 1994., et al.
- 4HigginsJPT, GreenS, eds. Cochrane handbook for systematic reviews of interventions 4.2.5 [updated May 2005]. In: Cochrane library, Issue 3, 2005. Chichester, UK: John Wiley; 2005.
- 5Modelling survival data in medical research. 1st ed. London: Chapman and Hall; 1994..
- 6Robust locally weighted regression and smoothing scatterplots. JASA. 1979; 74: 829–836..
- 8R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2005..
- 10Fifteen-year follow-up of the Anglian Germ Cell Cancer group adjuvant studies of carboplatin as an alternative to radiation or surveillance for stage 1 seminoma [abstract]. Proc Am Soc Clin Oncol 20: 2001. Abstract 780., , .
- 14Should surveillance be considered the standard of care in stage I seminoma? [abstract]. J Clin Oncol. 2005; 23(16 suppl ). Abstract 4520.. , , et al.
- 22Randomized trial of 30 versus 20 Gy in the adjuvant treatment of stage I Testicular Seminoma: a report on Medical Research Council Trial TE18, European Organisation for the Research and Treatment of Cancer Trial 30942 (ISRCTN18525328). J Clin Oncol. 2005; 23: 1200–1208., , , et al.
- 26Germ cell testicular tumors in clinical stage A and normal values of serum tumor markers post-orchiectomy: the experience in the management of 300 consecutive patients. J B.U.ON. 2005; 10: 195–200..