Development of Nomograms to Predict the Recovery of Erectile Function Following Radical Prostatectomy

Abstract

Introduction:

Given the number of confounders in predicting erectile function recovery after radical prostatectomy (RP), a nomogram predicting the chance to be functional after RP would be useful to patients’ and clinicians’ discussions.

Aim:

To develop preoperative and postoperative nomograms to aid in the prediction of erectile function recovery after RP.

Main Outcome Measures:

International Index of Erectile Function (IIEF) erectile function domain score-based erectile function.

Methods:

A prospective quality-of-life database was used to develop a series of nomograms using multivariable ordinal logistic regression models. Standard preoperative and postoperative factors were included.

Main Outcome Measures:

The nomograms predicted the probability of recovering functional erections (erectile function domain scores ≥24) and severe erectile dysfunction (≤10) 2 years after RP.

Results:

3 nomograms have been developed, including a preoperative, an early postoperative, and a 12-month postoperative version. The concordance indexes for all 3 exceeded 0.78, and the calibration was good.

Clinical Implications:

These nomograms may aid clinicians in discussing erectile function recovery with patients undergoing RP.

Strengths & Limitations:

Strengths of this study included a large population, validated instrument, nerve-sparing grading, and nomograms that are well calibrated with excellent discrimination ability. Limitations include current absence of external validation and an overall low comorbidity index.

Conclusions:

It is hoped that these nomograms will allow for a more accurate discussion between patients and clinicians regarding erectile function recovery after RP.

INTRODUCTION

Erectile dysfunction (ED) is an invariable sequela of radical prostatectomy (RP).¹ Classic predictors of recovery have been well delineated and include preoperative erectile function, patient age at time of surgery, and nerve-sparing status.² Other predictors proposed include postoperative erectile hemodynamic status, surgeon volume, surgeon as an independent predictor, and vascular comorbidity status.³ However, reported rates of postoperative ED in this population are tremendously variable, ranging between 10%–90% in the literature.¹ The reasons for this are complex and revolve around variability in the definition of adequate postoperative erectile function, discrepant study patient populations, and varied means of data acquisition. The question can, therefore, be asked: How does a clinician determine the likelihood of erectile function (EF) recovery for an individual patient?

Nomography has established itself in urology as a useful tool for the prediction of a variety of outcomes.^4–8 It is well documented that prediction models such as this are superior to guesswork by a clinician.^9–11 It is also well known that the complications profile from a specific treatment strategy can impact the patient decision-making process and that realistic expectations can impact patient satisfaction with an assigned treatment.^12,13 In an attempt to aid this process, we undertook this work to develop a series of nomograms to allow a more accurate, evidenced-based, individualized estimate of long-term EF outcomes after RP.

METHODS

Patient Population

We analyzed an institutional database containing data from an ongoing ethics committee–approved quality-of-life (QOL) study (IRB16–459), prospectively assessing, among other outcomes, erectile function recovery after RP performed by 1 of 3 experienced surgeons.

Only those patients with preoperative and ≥1 postoperative EF data were included in this analysis. We used a nerve-sparing scoring system: 1 = perfect nerve sparing, 2 = mild damage, 3 = moderate damage, 4 = complete resection. Each nerve was assigned a grade by the surgeon at the end of the operation, and a total score (right + left sides) was calculated (range 2–8; with 2 representing perfect nerve sparing and 8 complete non-nerve sparing). With regard to medical comorbidities, those present before treatment or that developed within the first year of surgery were assessed by the treating physician and included coronary artery disease, peripheral vascular disease, stroke, diabetes, hypertension, hyperlipidemia, hypercholesterolemia, obesity, and cigarette smoking status. Patients were considered as having the morbidity if it was noted as being present in the patient chart.

Erectile Function Assessment

Preoperative and postoperative EF was assessed using the International Index of Erectile Function (IIEF). The IIEF instrument is a well-validated inventory that has been used extensively in the evaluation of EF, particularly in response to pharmacotherapy.¹⁴ The questionnaire has 15 questions; questions 1–5 and 15 are organized into an EF domain (EFD), which for the sexually active man is scored 6–30. A score ≥24 indicates robust EF (average per-question response of “4,” for most questions representing a “most of the time” response) and a score ≤10 defines severe ED. Patients completed the first IIEF before surgery and as part of the QOL study completed the questionnaires serially at 3, 6, 9, 12, 18, and 24 months. Patients completed the questionnaire for their sexual activity for the 4 weeks before the visit, regardless of whether they were using phosphodiesterase 5 inhibitors(PDE5i). All data were entered by research staff into the departmental database. PDE5i use was assessed by a single question, which asked the patients whether they had used PDE5i never, sometimes, or regularly.

Database

We used an open-source software program (www.caisis.org) designed for the collection and analysis of clinical data in the healthcare field, and its security system adheres strictly to the guidelines set forth by the Health Insurance Portability and Accountability Act. The application and the source code are freely available for download under the General Public License, an open-source license approved by the Open Source Initiative. To ensure accuracy of data entry, the database was reviewed and verified.

Nomogram Development

A series of nomograms were constructed using multivariable ordinal logistic regression models. Both ordinal and continuous variables were fit using restricted cubic splines to relax linearity assumptions. No variable selection was performed, because we included all clinically relevant variables. All nomograms were internally validated, with bootstrapping used to correct for optimism. Predictive accuracy of the nomograms was evaluated by 2 methods. First, discrimination was calculated by the area under the receiver operating characteristic curve, with values ranging from 0.5 (no discrimination) to 1 (perfect discrimination). Second, calibration was assessed visually by plotting the nomogram-predicted probabilities against the observed event proportion over a series of equally spaced values within the range of the predicted probabilities. All analyses were performed using S-plus 2000 Professional software (Statistical Sciences, Seattle, WA, USA) or the open source software R (R Development Core Team, 2008) with the Design and Hmisc libraries added.¹⁵ This research was funded in part through the NIH/NCI Cancer Center Support Grant P30 CA008748.

RESULTS

Patient Population

From 328 eligible patients from the RP QOL study, data were complete and full for evaluation in 272, with the remaining being excluded because of lack of erectile function data or wrong questionnaire filling. The mean patient (SD) age was 59 (7) years. The comorbidity profile included diabetes (5%), hypertension (29%), hyperlipidemia (34%), coronary artery disease (4%) and cigarette smoking (past or present) (24%).

Erectile Function Outcomes

The number of patients completing the IIEF questionnaires at baseline, 12 and 24 months after surgery were 272, 256, and 264, respectively. Table 1 represents the EFD scores at each of the 3 time points. Only 14% of patients had IIEF EFD scores ≤10 at baseline, but the percent increased to 49% at 1 year after RP, then dropped to 38% at 2 years after surgery. Figure 1 is a histogram representing changes in IIEF EFD score from baseline to 2 years after surgery. Table 2 represents the PDE5i use scores. The percentage of men who used PDE5i regularly rose from 6.3% at baseline to 21.6% at month 12 after surgery and then decreased to 13.9% at month 24. Given the inconsistency in use of PDE5i, we believe this nomogram would be better described as a predictor of EF recovery with or without the use of PDE5i.

Table 1.

Erectile function outcomes (with or without PDE5i)

	IIEF EF Domain Score
Time	Median (IQR)	≥24 (%)	11–23 (%)	≤10 (%)
Baseline	29 (22.30)	74%	12%	14%
12 months	11 (5.16)	24%	26%	49%
24 months	22 (17.27)	36%	26%	38%

EF = erectile function; IIEF = International Index of Erectile Function; IQR = interquartile range; PDE5i = phosphodiesterase 5 inhibitor.

Figure 1.

Change in IIEF EFD score between baseline (before surgery) and 24 months after surgery. EFD = erectile function domain; IIEF = International Index of Erectile Function.

Table 2.

PDE5 inhibitor use

Time-point	Never	Sometimes	Regularly
Baseline (n = 272)	79%	15%	6%
12 months (n = 256)	38%	40%	22%
24 months (n = 264)	53%	34%	14%

PDE5 = phosphodiesterase 5.

Nomogram Development

Nomograms were developed predicting the 24-month EFD score. The preoperative nomogram (Figure 2) was based on baseline parameters. The early (3-month) postoperative nomogram was based on preoperative and intraoperative nerve-sparing status (Figure 3). Finally, a third nomogram (delayed post-operative) was developed using 12-month EF data (IIEF-EFD score) to predict 24-month outcomes (Figure 4). To use any of those nomograms, the reader must see where a given patient fits in each line (factors), see how many points are added by drawing a perpendicular line upward, and see the intersection on the top line labeled “Points.” A protective factor, such as younger age, translates into more points than older age. After checking how many points each factor adds to a given patient, the reader should go to the “Total Points” line and see, drawing a perpendicular line downward, the probability of meeting each specific outcome.

Figure 2.

Preoperative nomogram. EFD = erectile function domain; IIEF = International Index of Erectile Function.

Figure 3.

Early postoperative nomogram. EFD = erectile function domain; IIEF = International Index of Erectile Function.

Figure 4.

12-month postoperative nomogram. EFD = erectile function domain; IIEF = International Index of Erectile Function.

The bootstrap-corrected concordance indexes for the preoperative, early postoperative, and late operative models were 0.76, 0.78 and 0.87, respectively. Calibration plots appeared good.

DISCUSSION

Prostate cancer is one of the most common cancers in men.¹⁶ A number of factors have resulted in a far greater number of men being diagnosed with prostate cancer at the earliest stages, and only 3% of men die of this disease. Thus, in the modern era, the vast majority of men live for a long time after diagnosis and treatment. This has translated into a greater number of younger men being diagnosed with prostate cancer and surviving their disease. Given these factors, long-term sexual function has become an even more significant issue.

The incidence of ED after RP varies dramatically between 10%–90%.^2,17–22 Could it be that 1 center performs RP so much better than others? The answer to this question is not a simple one but probably lies just as much in how data are reported as it does in the technical expertise of any given surgeon.

Given the variability in EF outcomes in the literature, the clinician having a pre-RP discussion with a patient has a dilemma. Does the urologist cite the figures in the literature and, if so, which ones? Should they quote the very best figures, or should they cite the results in their particular practice? For most clinical urologists, structured tabulation of post-RP EF outcomes is not routine. Thus, although the clinician will likely discuss patient age, baseline EF, and nerve-sparing status as predictors of EF recovery, with the patient sitting in front of them, they are, therefore, likely to guess what the EF outcomes might be. The available literature and tools for predicting EF recovery can estimate the chance of recovery globally and, at best, factors that can influence it favorably or adversely. Nevertheless, the complexity of association between those factors is not encompassed by the current tools, creating a scenario where the individual chance and net effect of all pros and cons is unknown. As discussed briefly, on the patient decision-making process for any treatment, the perception of side effects is very relevant. This perception comes from non-medical sources, such as blogs, and family and friend experiences, but, importantly, it usually comes from counseling with >1 doctor.¹² It seems reasonable to state that the heterogeneity of estimates for EF recovery between doctors can negatively impact the whole process, and, therefore, the existence of an available tool to generate an evidence-based individualized estimate would be of great use.

Nomograms are prediction models that have been consistently shown to be better than a clinician’s guess regarding a particular outcome. In 1 study, nomograms out-predicted urologists in their prediction of biochemical recurrence in men with prostate cancer.²³ In a breast cancer study, nomograms were more accurate than a multidisciplinary team of physicians when predicting lymph node status.²⁴

We decided to develop 3 nomograms, which we believe will be useful for clinicians in their discussion with the patient with prostate cancer undergoing RP: specifically a preoperative nomogram for the pre-RP discussion; an early postoperative nomogram for the discussion with the patient in the first months after RP surgery; and a late postoperative nomogram to be used in the second year after surgery, with 12-month EF outcomes used to predict long-term EF.

We decided to use as an outcome the prediction of the IIEF EFD score and categorized outcomes into 3 groups: severe ED, moderate dysfunction, and robust function (≤10, 10–24, or ≥24). The IIEF is a highly validated instrument and is familiar to most urologists.¹⁴ It has been our clinical experience that most patients are primarily interested in knowing whether they will be, first, capable of having sexual intercourse and, second, having severe ED, which might preclude them from responding to oral therapy and possibly even intracavernosal injection therapy. An IIEF EF domain score ≤10 represents severe ED.¹⁴ We chose a score ≥24 to represent robust EF. This score equates to a mean score of 4 per question on the EF domain, representing an answer “most of the time” for most domain questions. Although a score ≥26 has been used to define no ED, it has been shown that most men with scores between 22–25 have excellent self-reported erectile function.²⁵ Indeed, 2 major post-RP ED studies have used scores of 24 and 22 to define success.^26,27

With regard to the time point at which to predict the EFD score, we chose 24 months. It has been shown that EF recovery continues out to this timepoint and sometimes later.²⁸ Rabbani et al² have shown continued EF improvement at 24 months and beyond. Glickman et al²⁸ have shown that some men continue to see EF improvement between 2–4 years. It may be that the very late change in EF is likely not related to cavernous nerve regeneration but rather that at least some of the men who experience EF improvement past 2 years do so on the basis of the psychological concept of confidence restoration.

Although patients are not familiar with and are not using the IIEF questionnaire to score their function, a brief explanation of the structure of this questionnaire permits the patient to get a clearer sense of where his EF may end up. In our preoperative nomogram, we used standard factors to predict EF recovery, including patient age, preoperative function as defined by using the IIEF EF domain score, and comorbidity status as defined by the number of medical conditions chosen from those that have been associated with ED, particularly diabetes, hypertension, hyperlipidemia, cigarette smoking, and coronary artery disease. Numerous studies have demonstrated that patient age and pre-operative EF are predictive of EF recovery.²

In the early postoperative nomogram, nerve-sparing status has been added to the factors used in the preoperative nomogram. 1 of the basic tenets of EF recovery is that it is related to the volume of cavernous nerve sparing as introduced by Walsh and co-workers²⁹ in the early 1980s. For the purposes of this work, we used the 4-point nerve-sparing grading system we have been using for several years at our institution. This grading system assigns a score of 1 for perfect nerve sparing, 2 for mild damage, 3 for moderate damage, and 4 for complete resection. Thus, the range of scores is 2–8, with the higher number representing worse nerve sparing. We decided to avoid using the historical system of bilateral, unilateral, and non-nerve sparing, because this presupposes that nerve sparing is an all-or-none phenomenon, when in fact, this is not the case. In the late postoperative nomogram, we added in the 12-month IIEF EF domain score as a further factor to aid in the 24-month IIEF EF domain score prediction. It has been our clinical experience for a number of years that EF recovery begins to manifest itself at the end of the first year or at the beginning of the second year after RP and that a man’s EF at the 12-month time-point gives a rough yet reliable indication of long-term EF recovery.

Nomograms are familiar tools to most urologists because, for some years now, they have been used in prostate and other urologic cancers to predict stage, recurrence, and survival. In essence, a point system is used, where points are assigned to different factors (patient age, baseline IIEF EF domain score, etc) depending on their relative contribution to the outcome. Thus, the wider the line assigned to each factor, the greater the potential contribution it has to the outcome. For example, in the preoperative nomogram, the contributions of patient age and baseline EF are far greater than comorbidity status, with patient age being the most potentially influential contributor to EF recovery. This holds true for the early postoperative nomogram, although nerve-sparing status represents a significant contributor also. Of note, in the late postoperative nomogram, the most significant contributor to EF recovery is the 12-month EFD score.

In all 3 nomograms, the comorbidity status is the least influential contributor. Of course, this is not to say that it is not important, but its relative contribution is low. However, the patient population included in this analysis was relatively healthy, with few patients having >1 comorbidity, and only rare patients having ≥3 comorbidities. Thus, this fact may have minimized the contribution of comorbidity status to EF recovery in this analysis.

The clinical implication of these nomograms is obvious; they will facilitate a more accurate discussion between the urologist and the patient with prostate cancer. For example, using the early postoperative nomogram, for a 55-year-old man with perfect baseline EF (EFD score = 30), no comorbidities, and perfect nerve-sparing status (score = 2), the predicted probability of severe ED at 24 months is 14%; the probability of recovering EF to an EFD score ≥24 is only 58%; therefore, the chance of having mild-moderate ED is 28%. For a 60-year-old man with a baseline EFD score of 24, no comorbidities, and complete non-nerve sparing surgery (score of 8), the probability of having severe ED is 75%, and a 24-month EFD score ≥24 is only 6%. There are 2 reasons that this patient has even a small probability of retaining functional erections. First, it may be related to the fact that his baseline EFD score may have underestimated his true function; that is, his erectile machinery may have functioned better than he scored, perhaps due to the stress response to a prostate cancer diagnosis.³⁰ Second, although he was given a nerve-sparing score of 8 by the surgeon, the latter may have overestimated the amount of damage he/she caused at the time of the RP. Similar factors may explain why a 60-year-old man with a baseline EFD score of 10, 3 comorbidities, and partial nerve sparing (score of 4) has an 85% probability of severe ED but still has a 3% chance of have recovery of erections to an EFD score ≥24. It must also be mentioned that sexuality has different degrees of importance for each own and couple life, with men coping differently with ED and being more or less open to pursuing treatment. Therefore, although the nomogram is an accurate estimate of EF, it cannot be used to estimate global sexual well-being or satisfaction.

The strengths of this study include the use of a prospective database to generate the data; the nomogram is well calibrated and has remarkable discrimination ability; nerve-sparing was not assessed in an “all-or-none” fashion, and there is the inclusion of a significant number of patients for analysis. There are, however, limitations to this study, which include the fact that these nomograms have not yet been externally validated; our data came from only experienced surgeons; our patients had overall a low comorbidity index; and, most importantly, the data presented here include men who were using PDE5i after their RP. However, the proportion of the patients who admitted to using PDE5i at all at 24 months after RP was 50%, and only 1 in 7 used them regularly at that time point. Saying that, this use of PDE5i means that these nomograms should probably be used to predict the recovery of EF “with or without the use” of a PDE5i.

With regard to the low comorbidity index in this population, we believe that the first set of nomograms should be a best-case scenario group, to avoid underestimating the potential for recovery. The only group of men for whom these nomograms might not function well are those with >3 comorbidities. The 4-point grading system has been used at our institution for several years but may not be applicable to other centers. However, it is highly intuitive (1–4) and corresponds generally to intrafascial, interfascial, extrafascial, and resection, a system with which most prostatectomists are familiar. Finally, those clinicians using the IIEF-5 (Sexual Health Inventory for Men) may believe that these nomograms are not applicable to their patients. However, there is data available that, to convert from SHIM to EF domain score of the IIEF, a multiplication factor of 1.2 can be used.³¹ Although this would not be a perfect approach, it would give a reasonable figure for the doctor to discuss with patients.

CONCLUSIONS

These are the first published data delineating nomograms for the prediction of EF recovery after RP. It is hoped that the nomograms will facilitate a more realistic and comprehensive discussion with the patient before and after surgery.