Karim Touijer, Memorial Sloan-Kettering Cancer Center, Sidney Kimmel Center for Prostate & Urologic Cancers, 353 East 68th Street, New York, NY 10021, USA. e-mail: firstname.lastname@example.org
Study Type – Therapy (case series)
Level of Evidence 4
What's known on the subject? and What does the study add?
In lymph node dissections for cancer, the more extended the dissection, the higher the number of lymph nodes removed. In addition, the higher the number of nodes retrieved, the better the staging. This leads many investigators to set a threshold of a minimal number of nodes below which the dissection is considered inadequate.
Although the minimal threshold concept is generally good, it is not based on very objective data. a number of factors might influence the final number of nodes removed: (i) the surgeon and the surgical technique; (ii) the pathologists and tissue processing technique; (iii) the patient; and (iv) the audit effect and feedback to the surgeons about the number of nodes removed.
• To examine the number of lymph nodes removed over time for men undergoing a standard pelvic lymph node dissection (PLND) during radical prostatectomy.
PATIENTS AND METHODS
• In total, 2119 consecutive patients with clinically localized prostate cancer were scheduled for non-salvage radical prostatectomy between February 2005 and September 2009.
• All patients underwent PLND, including the external iliac, hypogastric and obturator fossa nodal groups.
• We tested whether the number of lymph nodes increased over time by including the date of each patient's surgery into a linear regression model using nonlinear terms.
• From 2008 onward, there appears to be a large increase in the number of nodes removed.
• Date of surgery was a significant predictor of the number of nodes removed (P < 0.001).
• The anatomical template of dissection, the specimen submission and pathological assessment were reportedly unchanged.
• The nodal yield increase in the later part of the study coincides with an increase in the academic interest in PLND and nodal metastasis in prostate cancer at the institutional level and worldwide.
• Without any intentional change in surgical technique or pathological processing, the number of lymph nodes removed in our radical prostatectomy experience increased.
• This change coincided with an increased academic interest in the subject and highlights the positive feedback effect.
• The change also raises concerns about unaccounted for confounding factors that could affect multi-institutional datasets and surgical clinical trials.
In many cancers, the presence of lymph node metastases portends a poor prognosis and the quality of the lymphadenectomy has been confirmed as an independent predictor of outcome . In gastric cancer and testicular cancer, meticulous anatomical work has helped develop lymph node dissection templates specifically adjusted to the severity of the disease [2–4]. In bladder cancer, an association between the number of lymph nodes retrieved and survival was shown and, because of the lack of a true objective intra-operative index to measure the quality of the lymphadenectomy, a minimal number of nodes was chosen to determine the lowest threshold of adequacy of the dissection [5–7].
In prostate cancer, it is well established that the wider the anatomical template of dissection, the higher the number of lymph nodes retrieved, and the higher the nodal yield, the better the detection of metastasis [8–10]. However, no consensus has been obtained on the specific number of nodes required to define the quality of dissection for prostate cancer . In reality, how objective is a qualitative threshold or description of the anatomical template in determining the adequacy of a dissection? In our routine practice of lymphadenectomy during radical prostatectomy for prostate cancer, we dissect the external iliac, hypogastric and obturator fossa nodal groups; however, it was our subjective impression that, in recent years, the mean number of lymph nodes removed during radical prostatectomy at our institution has been steadily rising. In the present analysis, we aimed to objectively verify this impression and, if confirmed, to explore the possible causes such an increase.
PATIENTS AND METHODS
The present study is an institutional review board-approved retrospective analysis of prospectively collected data.
From February 2005 to September 2009, 2318 consecutive patients with clinically localized prostate cancer underwent a standard pelvic lymph node dissection (PLND), including the external iliac, hypogastric and obturator fossa nodal groups, during non-salvage radical prostatectomy. Those patients who were missing data on the number of nodes removed (n= 18) were excluded, leaving 2087 patients available for analysis. An open retropubic approach was performed by one of six surgeons on 1159 (56%) patients and, in 928 patients, the approach was laparoscopic transperitoneal, as performed by one of five surgeons.
PREOPERATIVE TREATMENT PLANNING
All surgeons used a uniform preoperative evaluation and risk assessment. Preoperative clinical parameters, including patient age, body mass index, 2002 TNM clinical stage, preoperative PSA level and Gleason sum on prostate biopsy, were prospectively collected.
PLND INDICATIONS AND ANATOMICAL LIMITS
Regardless of their disease risk stratification, all patients underwent a standard PLND, including the external iliac, hypogastric and obturator fossa nodal groups. All surgeons submitted the lymph nodes in two packets (right and left), except for one surgeon who submitted the lymph nodes in six separate packets (right and left: external iliac, hypogastric and obturator fossa, respectively).
The specimen processing in pathology remained unchanged during the study period. All lymph node specimens removed during surgery were counted by either a technologist or a pathologist, and then fixed in neutral buffered 4% formaldehyde for 24 h, separated from the adjoining adipose tissue and counted manually. Each node retrieved was cut in 3-mm slices, which were separately embedded in paraffin, stained with haematoxylin and eosin, and examined microscopically.
The pathological results were reported as the number of lymph nodes involved and the total number of lymph nodes retrieved per packet; such computerized worksheet, detailed measurement and reporting was instituted in 2004 and has therefore remained unchanged during the study period.
INSTITUTIONAL AND INTERNATIONAL ACADEMIC INTEREST IN PLND
To gauge the international urological community's academic interest in PLND for prostate cancer, we reviewed and counted all the accepted abstracts on the topic each year at the annual meetings of the AUA and the European Association of Urology (EAU) during the study period, using the search engines available on their corresponding websites. We choose the keywords ‘lymph node dissection’ OR ‘lymphadenectomy’ AND ‘prostate cancer’. We also measured our institutional academic interest in PLND in prostate cancer by reviewing and counting all of the abstracts submitted by our group on the topic for the AUA and EAU annual meetings during the study period.
The primary aim of the present study was to examine the number of lymph nodes removed over time for men undergoing radical prostatectomy. To test whether the number of lymph nodes increased over time, we entered the date of each patient's surgery as a predictor in a linear regression model. To allow for a potential nonlinear relationship between the date of surgery and the number of nodes removed, the date of surgery was entered into the model with nonlinear terms (i.e. restricted cubic splines with knots at the tertiles). To illustrate the association between the number of lymph nodes removed over time, we plotted the predicted mean number of nodes removed by the date of surgery. To illustrate the variability in the number of nodes removed, we additionally plotted the predicted 25th and 75th percentile of the lymph node distribution, as obtained from a quantile regression model.
To account for potential differences in the patient population over time, we adjusted for the probability of having a positive lymph node using a Memorial Sloan-Kettering-developed nomogram . We adjusted for within-surgeon clustering using the ‘cluster’ option in STATA (StataCorp; College Station,TX, USA).
The data audit confirmed our suspicion that the number of lymph nodes retrieved during radical prostatectomy has increased over time. Figure 1 illustrates the number of nodes removed over time, after adjustment for the probability of having a positive lymph node (P < 0.001). The mean number of nodes removed over time appears relatively constant from 2005 to 2008; in 2008, there appears to be a pronounced increase in the number of nodes removed that continues through 2009. In 2005, the median number of nodes removed was 13 (interquartile range, IQR, 8–17) compared to 18 (IQR, 12–26) in 2009 (Table 1). We were concerned that only reporting the mean number of nodes removed over time would ignore the large overall variability in the number of nodes removed. For example, in 2005, the number of nodes removed was in the range 2–48; in 2008, it was in the range 1–58. To reflect this wide variation, we included shaded bands that indicate the 25th and 75th percentiles of the distribution of the predicted number of nodes removed in Fig. 1. These lines show that there is a wide range in the number of nodes removed; however, importantly, they also follow the same trend over time.
Table 1. Number of lymph nodes removed over time, both overall and separately by approach (laparoscopic and open radical prostatectomy)
Year of surgery
Total number of nodes removed
Laparoscopic radical prostatectomy (N= 928)
Open radical prostatectomy (N= 1159)
All values are presented as the median (interquartile range).
To explore the causes of the rise in number of nodes retrieved, we explored several possible contributing factors.
PATIENTS AND DISEASE
Patient characteristics of the 2087 men who were treated by radical prostatectomy are shown in Table 2. The median (IQR) patient age was 60 (55–65) years and the median (IQR) PSA level was 5.1 (3.7–7.2) ng/mL. Except for Gleason score, we found no evidence that any clinical feature significantly changed over time. In 2005, 31% of patients were Gleason 6 or lower, whereas, in 2009, the proportion was 19%; the changes in the proportion of Gleason 8+ cancers were not consistent over time. These findings are suggestive of a greater use of active surveillance over time for patients with lower Gleason grade disease.
Table 2. Summary of patient characteristics (N= 2087)
P value for change over time
Only includes surgeries from February 2005.
†Only includes surgeries up to September 2009. IQR, interquartile range.
All surgeons reported performing a standard PLND, encompassing the external iliac, hypogastric and obturator fossa, during the study period and denied any deliberate extension of the limits of their dissection. To test the possibility that a more extended PLND was performed in more recent years, we evaluated the change in operating time by date of surgery. We found no evidence that operating time significantly changed over time (P= 0.7) (Table 3). When we looked at operating time separately by each of the four high volume surgeons, it appeared to be stable for three of the four experienced surgeons; operating time decreased slightly over time for one of the surgeons.
Table 3. Operating time throughout the present study
Operating room time (min)
All values are presented as the median (interquartile range).
We were also interested in whether the overall trend in the number of nodes removed over time was different by surgeon or surgical approach (laparoscopic vs open). Figure 2 shows the trends for the four experienced surgeons in our cohort. Two of the surgeons (B and D) performed laparoscopic radical prostatectomy, whereas the other two (A and C) performed open radical prostatectomy. The same sharp increase in recent years was seen in all four four surgeons and in both surgical techniques.
As stated above, the pathological specimen-processing institutional procedures have not changed during the study period. The pathology reports were signed by 12 different pathologists.
Figure 3 shows the number of abstracts accepted at the AUA and EUA and the number of abstracts submitted by investigators at our institutions that addressed lymph node dissections during the study period. The curves representing the institutional and international academic interests in PLND for prostate cancer showed a parallel increase over time, with a sharp increase in 2007. This rise preceded the increase in the number of lymph nodes retrieved during radical prostatectomy at our institution.
Finally, we were concerned that the results of the present study could have been unduly influenced by the choice of nonlinear terms used. Briefly, we used restricted cubic splines, which allow the curve to bend at prespecified locations (called knots), the tertiles. As such, we conducted a sensitivity analysis using locally weighted scatterplot smoothing (Lowess) to graphically illustrate the association between the number of lymph nodes over time. Lowess is a non-parametric method that plots a smoothed curve of the relationship between two variables. Overall, the main results obtained very similar. We are confident that the results are not unduly influenced by the placement of knots. The analyses in the present study were also not changed markedly when adjusted for clinical stage, PSA level and type of surgery.
In oncological surgery, the quality of lymph node dissection is measured either quantitatively or qualitatively. In testis and gastric cancers, for example, risk-adjusted anatomical templates have been carefully developed based on well-established primary landing zones [2–4]. Therefore, the total number of nodes retrieved is less frequently used as a surrogate of quality. In bladder cancer, however, the number of lymph nodes retrieved during cystectomy was shown to be an independent predictor of outcome, which led investigators to propose a threshold below which lymph node dissection is deemed inadequate [6,7]. In prostate cancer, on the other hand, no consensus has been reached with regard to the anatomical limits of PLND and no lower limit of acceptable quality has been set [11,12]. Several studies report an association between the number of lymph nodes removed and a better staging of nodal metastasis. Comparative analyses of limited vs standard or extended PLND, including the external iliac, hypogastric and obturator fossa nodal packets, have shown up to threefold increases in the detection of lymph node metastasis [8–10,13]. Consensus on such a template would theoretically allow us to perform large clinical trials with little variability and compare multi-institutional datasets with less heterogeneity. In reality, unfortunately, we know that variability exists between surgeons, surgical approaches , mode of specimen submission (en bloc vs separate packets) and pathological specimen processing . The present study confirms such variation, whereby, at the same institution, and without any deliberate change in the standardized surgical technique and pathological processing, the number of lymph nodes retrieved at the time of radical prostatectomy has sharply increased over the course of the study period. In the present analysis, we were unable to detect any change in the patients and their disease characteristics, in the surgical technique, or in the pathological processing of specimens. Indeed, the change (i.e. the increase in number of lymph nodes removed in the present study) affected the results obtained for all surgeons, regardless of whether they performed open or laparoscopic surgery, or whether they submitted the specimen in packets or as right and left PLND. More importantly, the significant increase did not occur gradually over time but rather followed a sharp change at almost the same time for all surgeons. This is in accordance with the exact pattern well-described by Gladwell , whereby change occurs in an epidemic fashion with three characteristics: it is contagious, happens dramatically rather than gradually, and is driven by small causes. In our experience, the number of lymph nodes removed did just that. Although causation is not established, we found it interesting that an increased awareness or academic interest in lymph node dissection and prostate cancer, both at our institution and worldwide, coincided with the time of increased nodal yield. The highest academic productivity preceded the increase in lymph nodes removed by ≈12–18 months, creating an indirect audit or feedback effect. It is obvious that this nodal count increase is a result of surgeons performing wider dissections and/or pathologists being more judicious about counting lymph nodes.
The interest, attention, focus and perhaps even the expectation of feedback in this case supports the observations that patients enrolled in clinical trials often have better outcomes than that of those not participating . The lessons learned from this analysis are outlined below.
First, although surgeons and pathologists are not reporting any objective change in the surgical- and specimen-processing techniques respectively, the increase in the number of lymph nodes retrieved shows that some imperceptible changes have taken place. Either the surgeons have performed a more meticulous or more extended dissection without directly realizing it and/or the pathologists have been more diligent about counting nodes without actively deciding so.
Second, in a general sense, the findings of the present study highlight a limitation of the prospectively maintained databases. All the PLND in the present study are labelled ‘standard type of dissection’ but, from the different nodal count removed over time, we know that they are not the same. Hence, any statistical analysis using this type of databases should take into consideration the year of surgery or adjust to some coefficient of improvement over time.
Third, quality improvement is possible through the implementation of surgical outcomes metrics measurement. In the example of the present analysis, measurement of the number of nodes removed during radical prostatectomy, as well as feedback provided to surgeons and pathologists through research projects focusing on the subject, has had a positive effect on the quality indicator measured. The feedback provides an incentive for improving, particularly with the awareness that an indivdual's performance is measured and analyzed.
In conclusion, the findings obtained in the present study illustrate the heterogeneity of surgical data and the potential role of unaccounted for confounding factors. The incorporation of objective quality metrics in surgery may help to improve performance, as well as reduce learning curves and variability, amongst surgeons, and also allow for the exchange of data and the provision of uniformity in surgical clinical trials.