A meta-analysis of patient outcomes with subcentimeter disease after chemotherapy for metastatic non-seminomatous germ cell tumor

Abstract

Background

Approximately a quarter of men with metastatic non-seminomatous germ cell tumor (NSGCT) have a residual mass, typically in the retroperitoneum, after chemotherapy. The management of small residual masses (≤1 cm) is controversial, with good outcomes seen with either post-chemotherapy retroperitoneal lymph node dissection (PC-RPLND) or surveillance. We sought to review our experience of surveillance and synthesize the cumulative findings with the current literature in the form of a meta-analysis.

Patients and methods

We searched PubMed, EMBASE and abstracts from ASCO and AUA to identify relevant, English-language studies for the meta-analysis. The DFCI (Dana Farber Cancer Institute) database was constructed from a database of men undergoing cisplatin-based chemotherapy for metastatic NSGCT. The outcomes of interest were the proportion with necrosis, teratoma or active cancer on histology at PC-RPLND (literature) and the total number of relapses, RP-only relapses and overall survival in men undergoing surveillance (literature and DFCI cohort).

Results

Three of 47 men undergoing post-chemotherapy surveillance at our institution relapsed over a median follow-up of 5.4 years. All three were alive at a median of 4.2 years after relapse. On meta-analysis, the pooled estimates of necrosis, teratoma and active cancer in the 588 men who underwent PC-RPLND were 71, 24 and 4%, respectively. Of the combined 455 men who underwent surveillance, the pooled estimate of the relapse rate was 5%, with an RP-only relapse rate of 3%. Of the 15 men who suffered an RP-only relapse on surveillance, two died of disease.

Conclusion

Surveillance is a reasonable strategy for men with minimal residual RP disease after chemotherapy and avoids an RPLND in ∼97% of men who are cured with chemotherapy alone.

introduction

Testicular cancer is the commonest solid tumor in men aged between 15 and 34 and notable for the fact that the vast majority of men with metastatic disease are cured with cisplatin-based chemotherapy [1]. However, around 20–25% of patients with non-seminomatous germ cell tumor (NSGCT) will have a residual tumor mass, most often in the retroperitoneum (RP) [2]. Surgery, in the form of post-chemotherapy RP lymph node dissection (PC-RPLND), is routinely carried out for patients with residual masses greater than 1 cm as ∼45% will harbor teratoma and 10% viable cancer [3–6].

The management of NSGCT with small residual masses (≤1 cm in short axial dimension in accordance with RECIST 1.1) post-chemotherapy is controversial. Men achieving serologic and radiologic complete remission are treated with PC-RPLND at certain institutions as routine policy, with its advantage of being able to resect potential low volume chemoresistant teratoma or viable residual cancer [7]. However, recent reports have suggested that surveillance may be an option in this situation [8, 9] and that the majority of men can be safely spared the potential complications of RPLND [10].

We therefore sought to review current literature on treatment strategies for this group of patients and combine it with the outcomes from a similar group of patients treated at the Dana Farber Cancer Institute (DFCI) in the form of a meta-analysis. The aim was to synthesize the cumulative findings and better define the clinical outcomes of this group of patients managed across institutions with and without a PC-RPLND.

patients and methods

DFCI surveillance cohort

This was selected from an IRB approved database and protocol of men undergoing cisplatin-based chemotherapy for testicular cancer at the DFCI between December 1997 and December 2007. The database was created by identifying all patients with ICD-9 code of 186.9 in the DFCI system during this period and data extracted from the electronic medical record using ICD-9 codes. Patients with seminoma and primary mediastinal disease were excluded. All patients with metastatic disease to the retroperitoneum at the time of diagnosis, achieved a complete serologic (i.e., marker negative status) and radiologic remission (i.e., resolution of all radiologic findings, including RP nodal side ≤1 cm in axial diameter [11]) after primary chemotherapy alone were identified. Surveillance is the DFCI policy for this approach. In three additional patients, RP node size was greater than 1 cm at the end of chemotherapy, but these men were observed given the degree of remission and anticipation their RP node sixe would decrease to <1 cm with repeat scanning; in all three cases, nodal size showed a continued interval response to ≤1 cm, and these patients were therefore included. One patient who presented with an initial presumed RP primary tumor subsequently had a testicular tumor found on scrotal ultrasound 15 months after the first tumor had been diagnosed. This patient had not undergone scrotal imaging at the time of first diagnosis, and hence, this testicular ‘recurrence’ was not included as a systemic relapse in our data.

A total of 47 men met the criteria to be included in the analysis. Baseline patient and disease characteristics were obtained, and the International Germ Cell Cancer Collaborative Group (IGCCCG) risk classification [12] and disease stage were assigned. The details and outcomes of relapses were noted, with follow-up defined as the interval between the completion of primary chemotherapy and date of last visit to the hospital. Time to relapse was defined as the time from the completion of chemotherapy to the diagnosis of recurrence or censored at the last follow-up visit.

search strategy, study selection and data extraction for meta-analysis

Literature searches of PubMed (through November 2012) and EMBASE (1974 to November 2012) and abstract searches from the American Society of Clinical Oncology (ASCO) (1997–2012) and the American Urological Association (AUA) (2002–2012) were carried out using the following search terms: minimal residual disease, post-chemotherapy, testicular cancer (Figure 1). Searches were limited to English-language publications and conducted in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [13]. The key inclusion criterion was the presence of residual disease, up to 1 cm in maximal size, only in the retroperitoneum after chemotherapy for metastatic NSGCT. In total, 10 studies were included in the meta-analyses, with 6 describing data on PC-RPLND [7, 14–18] and 4 on surveillance for the post-chemotherapy minimal residual mass (three published studies [8, 9, 19] plus the DFCI experience). Summary characteristics of these studies are shown in Table 1.

Figure 1.

Flowchart of study selection.

Table 1.

Summary of studies included in meta-analysis

Citation	N	IGCCCG risk group (%; good/intermediate/poor)	Median (mean) follow-up (years)	Overall relapse-free rates (%)
PC-RPLND
Fossa [7]a	49	77/14/9b	Not available	Not available
Steyerberg [14]	199	Not available	Not available	Not available
Oldenburg [15]a	54	76/14/10	Not available	Not available
Karellas [16]	147	98/2/0	3.0	97 (5 years)
Steiner [17]	81	Not available	(7)	98
Pfister [18]	28	Not available	Not available	Not available
Surveillance
Kakiashvili [19]	106	Not available	Not available	100
Ehrlich [8]	141	77/9/14	15.5	92 (15 years)
Kollmannsberger [9]	161	94/3/3	4.3	94
DFCI (2012)	47	79/15/6	5.4	96 (5 years)

^aMedical Research Council group.

^bThe studies by Fossa and Oldenburg are from the same center, but report on independent datasets of patients managed among 1980–1990 and 1990–2000, respectively.

The outcomes of interest from the PC-RPLND studies were the histological rates of necrosis, teratoma and active malignancy, whereas those from the surveillance studies were the number of all relapses and RP-only relapses, as well as overall survival, where available.

statistical analysis

The patient and disease characteristics for the DFCI surveillance cohort were summarized using proportions. The proportion of patients with the outcome events and 95% confidence intervals (95% CIs) were derived from each study. For studies reporting zero events, a classic half-integer continuity correction was used for the calculation of incidence and its variance. For the evaluation of pooled estimates, the meta-analysis was carried out, where statistical heterogeneity was assessed using Cochrane's Q statistic [20], and inconsistency was quantified using I ² statistics [100%×(Q−df)/Q] that estimates the percentage of total variation across studies due to heterogeneity [21]. We consider an I ² value of greater than 50% as indicative of substantial heterogeneity. Summary incidences were calculated using fixed-effects or random-effects models depending on the heterogeneity of included studies. When substantial heterogeneity was not observed, the summary estimate was calculated based on the fixed-effects model using the inverse variance method. When heterogeneity was observed, the summary estimate was calculated based on the random-effects model using the DerSimonian and Laird method that considers both within- and between-study variations [22]. Publication bias was assessed using Begg's [23] and Egger's tests [24].

All reported P-values are two-sided. Analyses were carried out using SPSS version 20 (SPSS, Chicago, IL) and R 2.10.1 (The R Foundation for Statistical Computing, Vienna, Austria).

results

DFCI surveillance cohort

Table 2 shows the baseline characteristics of our cohort. Overall, three men relapsed over a median follow-up of 5.4 years, of whom two were intermediate risk and one good risk at initial diagnosis (Table 3). Median time to relapse was 12.7 months. One patient recurred solely in the retroperitoneum, another had RP and pulmonary relapse and the third relapsed in the mediastinum. All three were alive at a median of 49.8 months after relapse, having undergone RPLND plus salvage chemotherapy (one), high-dose chemotherapy (one) or high-dose chemotherapy followed by surgical resection (one). In the entire cohort, there was one death, arising from a road traffic accident. The 5-year overall survival was 97%, with a 5-year relapse-free survival of 96%.

Table 2.

Baseline characteristics of DFCI surveillance cohort

Characteristics (N = 47)	Number (%)
Median age at diagnosis (years)	31
Range	16–52
Stage
IIA	19 (40)
IIB	10 (21)
IIIA	8 (17)
IIIB	6 (13)
IIIC	4 (9)
IGCCCG risk group
Good	37 (79)
Intermediate	7 (15)
Poor	3 (6)
Induction chemotherapy
BEP × 3	30 (64)
BEP × 4	5 (11)
BEP/EP combination	6 (13)
EP × 4	1 (2)
Other BEP combination	4 (9)
BEP × 2 + HDCT/ASCT	1 (2)
Teratoma in primary
Yes	16 (34)
No	31 (66)
Median follow-up (years; range)	5.4 (0.2–12.4)

Table 3.

Characteristics and outcome of DFCI patients who relapsed after achieving a complete remission to primary chemotherapy

Patient	Histology of primary	Stage/risk	Initial chemotherapy	Relapse site	Time to relapse (months)	Treatment of relapse	Status at last follow-up
1	E, YS, T, S, CCa	IIIB/intermediate	BEP × 3	RP	78.9	Surgery, chemotherapy	Alive
2	E, T	IIB/good	BEP × 3	Lungs, RP	3.3	Chemotherapy	Alive
3	E, S, CCa	IIIB/intermediate	BEP × 2, VIP × 1, EP × 1	Mediastinum	12.7	Chemotherapy, surgery	Alive

CCa, choriocarcinoma; E, embryonal carcinoma; YS, yolk sac tumor; T, teratoma; S, seminoma; BEP, bleomycin, etoposide, cisplatin; EP, etoposide, cisplatin; VIP, etoposide, ifosfamide, cisplatin; RP, retroperitoneum.

meta-analysis

For the incidence of histological findings of necrosis, teratoma and active cancer, six studies of patients treated with PC-RPLND for a subcentimeter residual mass were included, representing a total of 558 patients. Figure 2 summarizes the histological findings for each study as well as the combined cohort. By using a fixed-effects model (heterogeneity test: Q = 1.0; I ² = 0; P = 0.96), the pooled incidence of necrosis was estimated to be 71% (95% CI: 67–75%); similarly, the pooled incidence of teratoma was estimated to be 24% (95% CI: 20–27%) using a fixed-effects model (heterogeneity test: Q = 1.9; I ² = 0; P = 0.86), and the pooled incidence of active cancer was estimated to be 4% (95% CI: 1–7%) using a random-effects model (heterogeneity test: Q = 13.9; I ² = 64%; P = 0.02).

Figure 2.

Forest plots showing overall rates of necrosis (A), teratoma (B) and active cancer (C) at PC-RPLND.

The four surveillance studies (including the DFCI cohort) are summarized in Figure 3, comprising a total of 455 patients. Twenty-five of 455 men, or a pooled estimate of 5% (95% CI 0–10%) using a random-effects model (heterogeneity test: Q = 19.2; I ² = 84%; P = 0.00), undergoing surveillance suffered a relapse. Fifteen men [3% (95% CI 0–5%)] had an RP-only relapse, and of these 15, 2 died of disease (0.4%).

Figure 3.

Forest plots showing overall proportion of relapses (A) and RP-only relapses (B) in men undergoing surveillance.

A summary of the results of the meta-analyses is given in supplementary Table S1, available at Annals of Oncology online. There was no evidence of publication bias for either the PC-RPLND (Begg's P = 1.000, Egger's P = 0.665) or surveillance data (Begg's P = 0.750, Egger's P = 0.180).

discussion

the role of PC-RPLND for small residual masses

The rationale for RPLND in the management of residual masses greater than 1 cm after chemotherapy is well established. It is well documented the histology of a residual mass will be necrosis, mature teratoma and viable cancer in ∼45, 45 and 10% of patients, respectively [3–6, 25]. If a residual mass harbors active malignancy, complete resection via PC-RPLND along with adjuvant chemotherapy will produce a 5-year survival of 50–80% [26–28]. Since incompletely excised or unexcised residual active cancer has the potential to relapse, and teratoma is chemo- and radioresistant, with the capacity to undergo malignant transformation, PC-RPLND has been accepted as a means of optimizing oncologic outcomes in this patient group. However, patients with lesions less than 1 cm in short-axis diameter post-chemotherapy form a unique cohort.

Data addressing the outcome of patients with post-chemotherapy residual masses less than 1 cm in the form of several single-center series have reported outcomes after RPLND for small residual masses. Meta-analysis showed that the rates of active cancer or teratoma in these small residual masses were 4 and 24%, respectively, and hence, it appears that subcentimeter residual masses specifically are far more likely to represent necrotic tissue (with an estimated rate of 71%) rather than harboring teratoma or active cancer, when compared with the typical residual mass (of any size). In other words, PC-RPLND may benefit up to 30% of men with subcentimeter residual masses by removing viable cancer or teratoma. In the remaining 70%, surgery would have been unnecessary from an oncologic perspective. Nevertheless, it should also be appreciated surgical removal of viable cancer and, to a lesser extent, teratoma in subcentimeter masses does not provide a guarantee that the patient will not relapse. As such, the benefit of PC-RPLND is likely to be less than 30%, since a proportion of these patients will relapse after surgery.

surveillance for the small residual mass

PC-RPLND is a technically challenging procedure with a complication rate of 20–30% [29, 30]. Even a nerve-sparing procedure can lead to ejaculatory failure in more than 20% of patients [31]. An alternative to PC-RPLND for men with a subcentimeter residual mass is observation. There is an increasing amount of evidence to support this as a reasonable option without compromising survival. The study with the longest follow-up, and therefore offering the most persuasive evidence, comes from Indiana University [8], where 141 men who achieved a complete remission (residual mass <1 cm) were followed over a median time of 15.5 years. Twelve (8.5%) relapsed at a median time of 11 months from completion of chemotherapy, with five recurring solely in the retroperitoneum (3.5%). Two of these five men died of disease, and there were two other deaths from bone and brain relapse. The remaining eight men were in remission at a median follow-up of 11 years after recurrence, having been managed with surgery alone (three), chemotherapy and surgery (three), chemotherapy alone (one) or a combination of surgery and radiotherapy (one).

Data from DFCI are similar to the other surveillance literature, with an overall relapse rate of 6.4% and an RP-only relapse rate of 2.1% over a median follow-up of 5.4 years. All three men who relapsed at our institution were alive at a median of more than 4 years after recurrence. The other reported series broadly concur with these findings with an overall 5% risk of relapse on meta-analysis.

It is important to note that the patient characteristics of the surveillance studies in which such data are reported (Indiana, BC/Oregon and DFCI) are similar to those from the PC-RPLND literature. The BC/Oregon series had 94% good-risk patients, 3% intermediate and 3% poor-risk, and for the Indiana cohort, the corresponding numbers were 77, 9 and 14%, and for the DFCI cohort, the figures were 79, 15 and 6%. These are broadly equivalent to the disease characteristics in the Norwegian and Memorial Sloan Kettering PC-RPLND series [15, 16], which facilitates comparisons among the data.

Taken together, these studies provide good-quality evidence that surveillance is a feasible option in managing patients with subcentimeter post-chemotherapy radiographic abnormalities, though it must be appreciated that only the Indiana series had the longest follow-up to detect late recurrences beyond 5 years. Only 3 of the 12 patients in the Indiana series who relapsed did so after 5 years. This would suggest that most of the relapses that are destined to occur would have been captured by the median follow-up periods of 4.3, 5.4 and 7 years in the non-Indiana series. Further, if we extrapolate the late relapse rate seen in the Indiana series (i.e., 25% of all relapses) to the other studies, the estimated total number of relapses is higher (28 versus 25 of 455 patients); however, this higher rate of 6.2% remains within the 95% CIs of the estimated rate from the meta-analysis.

In essence, when combining all surveillance data, the overall relapse rate was 5%, with an RP-only relapse rate of 3.3%. Importantly, of the 15 men who experienced an RP-only relapse, two died of disease (both of whom had active cancer on histology). While accepting that these two deaths out of a cohort of 455 patients may theoretically have been prevented by PC-RPLND, the cumulative surveillance data indicate that PC-RPLND would have been unnecessary in nearly 97% of men with a subcentimeter residual mass.

We can also apply the relapse rates observed with surveillance to the PC-RPLND data in order to assess the likelihood of relapse outside of the retroperitoneum in PC-RPLND series. Ten (2.2%) of 455 patients observed post-chemotherapy had relapsed disease beyond the retroperitoneum; this rate (2.2%) translates to 12 men experiencing an extra-RP relapse in the cohort of 558 PC-RPLND patients. While these 12 patients are potentially curable with salvage therapy, the likelihood is that not all would be cured.

weighing up the pros and cons of either approach

Despite the attractiveness of surveillance, there are downsides associated with this strategy. An often-neglected issue is the radiation exposure from frequent computerized tomography (CT) scanning. The National Comprehensive Cancer Network (NCCN) guidelines recommend a minimum of six or seven CT scans over the first 5 years of follow-up for men with a complete response to chemotherapy for non-seminoma; hence, a man undergoing surveillance for a subcentimeter residual mass may be exposed to up to 100 mSv of radiation (each abdominopelvic CT carries a radiation dose of 14 mSv [32]). Such levels may be associated with increased cancer risk [33], though computing the precise magnitude of this risk is difficult.

Moreover, the requisite frequency of CT scans for subcentimeter residual masses post-chemotherapy is unknown. Authors from the Royal Marsden Hospital reported findings on CT scans in men being followed post-chemotherapy for germ cell tumor [34]. In their study, an abnormal CT scan at 5 years after primary chemotherapy was reported in 13 of 89 men who had achieved a complete remission for non-seminoma after chemotherapy with or without RPLND. Only four were clearly abnormal and two of these men had teratoma resected from retroperitoneum and both had teratoma at prior PC-RPLND. The two other cases had ‘unnecessary’ surgery for a Histoplasma lung nodule and a reactive lung nodule. There were nine cases with abnormalities that prompted further testing but were found to not be associated with cancer. As such, 87 of 89 patients had an ‘unnecessary’ CT scan at 5 years post-chemotherapy. Indeed, the Indiana series [8] is notable for lack of routine follow-up scans once a remission with all lymph nodes less than 1 cm was documented.

It is difficult to reconcile the 28% chance of finding teratoma (24%) or cancer (4%) on PC-RPLND for <1 cm residual disease with the 5% relapse rate on surveillance over an observation period of ∼5 years. The relapse rate of 5% suggests that the vast majority of patients could be spared the morbidity of surgery. There are a number of potential reasons for this divergence. First, the possibility that disease characteristics may be different between the two groups of studies needs to be carefully investigated. While the risk categories were broadly equivalent, the Indiana surveillance series [8] in fact had a larger proportion of patients with intermediate or poor-risk disease (23%) compared with the MSKCC PC-RPLND series (2%) [16]. Of the 12 men who relapsed in their study, seven were intermediate or poor risk, with three of the four deaths occurring in patients of intermediate risk. That they were able to see a relatively low relapse rate (similar to the rate observed in the BC/Oregon series [9]) over a long follow-up despite having a somewhat higher risk population strengthens the argument that surveillance is a reasonable strategy.

Further, it is important to make note of overall survival with both approaches. Five-year recurrence-free survival in the MSKCC PC-RPLND series was 97%, and the 5-year progression-free survival in the Norwegian PC-RPLND dataset (which included post-chemotherapy masses up to 2 cm) was 96%. In comparison, the Indiana surveillance data showed 5-year recurrence-free survival of 96% for good-risk disease, and 73% for patients with intermediate- and poor-risk disease combined. Since the MSKCC data were almost entirely composed of good-risk patients (144 of 147), evaluation of the good-risk data from Indiana shows equivalent survival to good-risk patients at MSKCC treated with PC-RPLND. Five-year relapse-free survival at our institution, where the patient make-up was similar to the Indiana series, was also similar (96%). Therefore, despite the lack of randomized data, survival appears comparable between PC-RPLND and surveillance. The literature on long-term outcomes are limited to the information provided here as four of the six RPLND series post-chemotherapy do not relapse and survival data. It is recognized a limitation of the literature and this paper is that four of the six datasets describing the outcomes with RPLND are presented only in abstract form and do not provide indepth details on patient populations.

In conclusion, the management of the subcentimeter residual mass after chemotherapy for metastatic germ cell tumor is controversial. Our meta-analysis suggests that 24% of these masses harbor teratoma, which concerns clinicians owing to its unpredictable and potentially malignant nature and that 4% contain viable cancer. However, robust data over a long follow-up period indicate that 5% of men on surveillance will go on to relapse, and about 3% will relapse in the retroperitoneum alone. This suggests that 97% of men in complete remission can be spared a PC-RPLND and the associated risks and complications. For the 5% who do relapse, cure is possible with salvage surgery and/or chemotherapy. The surveillance data therefore suggest that the biologic nature of microscopic residual teratoma is benign in the majority of cases.

The survival figures suggest that both RPLND and surveillance are equally efficacious for the management of men in this situation, which is reflected in the current European [35], American [36] and Canadian guidelines [37]. While our meta-analysis is unable to unequivocally answer the question of whether surgery or surveillance should be the preferred treatment modality, it does help to more accurately quantify the risks and benefits of both approaches. This will be of value to physicians when counseling patients in this situation, enabling both to jointly arrive at an informed decision of which option to pursue.

funding

No funding was used for this project.

disclosure

The authors have declared no conflicts of interest.