Erectile function (EF) recovery after bilateral nerve-sparing radical prostatectomy (nsRP) is one of the most emotionally charged “numbers” in modern urology. Not because clinicians enjoy questionnaires, but because patients do. A man can accept a scar, tolerate transient fatigue, and even make peace with a temporary urinary pad. But uncertainty about sexual recovery has a special talent for hijacking attention—often months before surgery and long after the cancer is gone.
The uncomfortable truth is that EF recovery is not a simple consequence of “nerve-sparing: yes/no.” It is an interaction between baseline sexual health, psychological state, surgical technique, healing biology, and time. Some men recover surprisingly well; others do not, despite an apparently perfect operation. This variability creates two recurring clinical failures: unrealistic preoperative expectations and “one-size-fits-all” postoperative rehabilitation plans.
The paper you provided offers a different way to think about prediction. Instead of searching for one dominant variable (age, nerves, drug, surgical approach), the authors used exploratory decision-tree modeling on data from the large randomized REACTT trial to identify hierarchies of predictors—the kind of “if this, then that” logic clinicians actually use at the bedside. The key message is both practical and slightly ironic: in a study designed around tadalafil, the strongest predictors of EF recovery were often not the drug at all, but baseline sexual desire, confidence, and intercourse satisfaction, plus the quality of nerve sparing and surgical approach.
Why Prediction Matters More Than Another “Penile Rehabilitation” Debate
Recovery of erections after nsRP is not only a quality-of-life issue; it is a counseling issue. When counseling is good, men plan realistically, partners stay engaged, and postoperative disappointment is less likely to evolve into avoidance, blame, or silent resignation. When counseling is vague, men either expect miracles (“robot surgery means no ED”) or expect catastrophe (“I’ll never have sex again”), and both extremes are clinically unhelpful.
Penile rehabilitation strategies—especially PDE5 inhibitors—are commonly used early after surgery. Yet the literature has repeatedly shown what many clinicians see in practice: drug-assisted erections during treatment and unassisted recovery after washout are not the same outcome. REACTT itself was designed with that distinction: a 9-month double-blind period of tadalafil once a day (OaD), tadalafil on demand (PRN), or placebo; then a 6-week drug-free washout; then 3 months open-label tadalafil OaD for all patients. The primary REACTT endpoint was unassisted recovery after washout (IIEF-EF ≥22), and tadalafil did not significantly improve that endpoint—while daily tadalafil did improve drug-assisted EF at the end of the double-blind period.
That nuance is not academic. It changes the conversation in clinic. If a patient asks, “Will tadalafil bring me back to normal?” the honest answer depends on whether “normal” means improved function while taking medication or spontaneous function after stopping it. The decision-tree analysis does something even more useful: it shows which men, based on baseline sexual profile and surgical factors, are most likely to land in better outcome branches—so counseling can be individualized rather than generic.
And yes, there is a mild irony that deserves acknowledgment. In a trial with hundreds of men and a major pharmaceutical intervention, the most predictive variables ended up being simple questions about desire, confidence, and satisfaction—things we often rush through because they feel “subjective,” even though they clearly predict “objective” recovery better than we like to admit.
The REACTT Dataset: A Rare Combination of Scale, Structure, and Follow-Up
The REACTT trial enrolled 423 men younger than 68 years with localized prostate cancer (Gleason ≤7, PSA <10 ng/mL) and normal preoperative erectile function (IIEF-EF ≥22). Surgery was performed across 50 centers in nine European countries and Canada, and men were randomized postsurgery (1:1:1) to tadalafil 5 mg OaD, tadalafil 20 mg PRN, or placebo for 9 months, starting within 6 weeks after nsRP.
From a modeling standpoint, this dataset is unusually rich because it includes:
- Multiple baseline IIEF domain scores and all 15 single-item IIEF responses,
- Surgical approach categorized as robot-assisted laparoscopy, conventional laparoscopy, open surgery, or other,
- A standardized nerve-sparing score (NSS) graded per neurovascular bundle (1–4 each side), with total NSS = 2 defined as “perfect,”
- Distinct assessment time points corresponding to drug-assisted EF (end of double-blind), unassisted EF (after washout), and re-challenge EF (end of open-label tadalafil for all).
The decision-tree modeling used the IIEF-EF domain score as a continuous outcome rather than a binary “recovered/not recovered” endpoint. Statistically, this matters because it preserves information—an IIEF-EF of 21 and 10 are both “not recovered” by the ≥22 rule, but clinically they are very different experiences.
The authors created three decision-tree models, each targeting a different clinical reality:
- EF while still on randomized treatment (end of 9-month double-blind period),
- EF after stopping treatment (end of 6-week washout),
- EF after all men received daily tadalafil open-label for 3 months (end of month 13.5).
This structure is not just elegant—it is clinically recognizable. It mirrors what patients ask: “How will I do while I’m taking something?”, “What happens if I stop?”, and “Will daily medication later still help?” Decision trees are particularly good at mapping these layered realities into an interpretable pathway.
Decision-Tree Modeling: A Clinical Mindset in Statistical Form
Most clinicians do not think in regression coefficients. They think in branching logic: “If he had strong function before, and nerves were spared well, and he’s motivated, he’ll likely do better.” Decision-tree analysis formalizes that instinct by repeatedly splitting the population into subgroups using thresholds that best separate outcomes.
In this analysis, trees were built using the Gini index as the split criterion, pruned with leave-one-out cross-validation, and handled missingness with surrogate splits. Importantly, this was exploratory and post hoc—the models were meant to describe and visualize patterns rather than declare definitive causal predictors.
The key advantage of decision trees is interpretability. Figures in the paper literally show the “first split,” the most informative variable, and then secondary splits that refine predictions. For example, in the end-of-double-blind model (Figure 1, page 13), the first split is based on baseline sexual desire (IIEF item 12 threshold ≥3.5). Men with higher baseline desire landed in higher mean IIEF-EF branches at 9 months than those with lower desire. Subsequent splits included confidence to get and maintain an erection (IIEF item 15), surgical approach (robotic vs not), nerve-sparing quality, and finally whether the man received tadalafil OaD.
The important clinical interpretation is not “sexual desire causes recovery.” Rather, sexual desire and confidence are proxy signals for a broader phenotype: baseline sexual vitality, relational engagement, psychological resilience, and possibly underlying vascular health. These factors likely influence rehabilitation behavior (frequency of attempts, adherence, willingness to persist), which in turn influences functional recovery trajectories.
In other words: the decision tree doesn’t just predict tissue biology. It predicts human behavior interacting with tissue biology. Medicine often pretends those are separate. They are not.
What Predicted Better EF at 9 Months: Desire First, Confidence Second, Then the “Surgery + Daily Tadalafil” Layer
The first model, examining IIEF-EF at the end of double-blind treatment, included 332 men with complete data. The dominant predictors were baseline IIEF single items, not age. Baseline sexual desire (IIEF item 12) split the population into a higher outcome group (mean IIEF-EF ~14.9) and a lower group (~11.1). Within the higher desire branch, confidence to get and maintain an erection (IIEF item 15) further separated outcomes (~15.4 vs ~7.1).
Once those baseline psychological/sexual vitality variables were satisfied, the model began to “care” about clinical interventions. In the higher desire + higher confidence branch, tadalafil once daily appeared as a meaningful next split, associated with higher mean IIEF-EF (the paper reports ~17.6 in that branch). This is consistent with REACTT’s primary trial finding: daily tadalafil improved drug-assisted EF during treatment.
For men not receiving daily tadalafil (PRN or placebo), surgical and technical factors emerged more strongly. The model highlighted robot-assisted laparoscopy and better nerve sparing (lower NSS) as predictors of higher mean outcomes in certain branches. In Figure 1 (page 13), branches show mean IIEF-EF values rising substantially in small subgroups where robotic approach and close-to-perfect nerve sparing align with favorable baseline sexual profiles.
This layered structure is clinically meaningful: daily tadalafil may enhance function while it is present, but high baseline sexual desire and confidence appear to determine who is positioned to benefit most from “conserving” surgery and early rehabilitation. The model implies a simple counseling idea: if a man enters surgery with high desire and confidence, he is more likely to convert excellent surgical conditions and daily pharmacologic support into better mid-term sexual function.
What Predicted Unassisted Recovery After Washout: Satisfaction Beats Medication, and Surgery Matters Again
The second decision tree evaluated IIEF-EF after the 6-week drug-free washout (month 10.5), using 320 men with complete data. Here, the dominant predictor shifted: baseline intercourse satisfaction (IIEF domain score threshold ≥12.5) became the first split, associated with higher IIEF-EF at washout (~15.6 vs ~11.4).
This is a subtle but important change. At washout, the model is predicting unassisted function—what the man can do without pharmacologic assistance. In this context, baseline satisfaction with intercourse likely reflects more than psychological optimism. It may reflect better baseline erectile rigidity, better penile vascular status, better relational stability, and more positive reinforcement from sexual activity. Those factors can influence rehabilitation behavior and tissue conditioning after surgery.
Crucially, in the washout model, randomized treatment (tadalafil OaD, PRN, placebo) had no predictive effect and is not included in the tree (as described in the figure caption on page 15). That finding aligns with the main REACTT outcome: tadalafil did not improve unassisted EF recovery after washout compared with placebo.
Instead, surgical approach—especially robot-assisted laparoscopy—appeared as a key predictor among men with higher baseline intercourse satisfaction. Younger age (<61) had a small predictive role, but again the model’s center of gravity was not age. As visualized in Figure 2 (page 15), branches combining higher baseline satisfaction with robotic surgery showed notably higher mean IIEF-EF at washout, including a small subgroup with very high mean values.
For counseling, the message is refreshingly direct: if a patient wants to know his odds of spontaneous recovery, daily tadalafil during the early months is not the dominant predictor in this dataset. Baseline sexual satisfaction and the quality/approach of surgery matter more.
After Open-Label Daily Tadalafil for All: Intercourse Satisfaction Still Leads, and “Robotic + Good Nerve Sparing” Wins the Highest Branches
The third decision tree examined outcomes after 3 months of open-label tadalafil once daily for all men (month 13.5), using 307 men with complete data. Once again, baseline intercourse satisfaction was the primary split (threshold ≥11.5), associated with higher mean IIEF-EF at the end of open-label treatment (~18.9 vs ~14.1).
Within the higher satisfaction branch, robot-assisted laparoscopy emerged as the main predictor of higher mean IIEF-EF (reported mean ~22.3, which falls around the threshold for “recovery” by the ≥22 definition). In the best-performing branches, a combination of robotic surgery and close-to-perfect/perfect nerve sparing produced mean IIEF-EF values in the mid-20s, approaching full recovery. Figure 3 (page 17) visually depicts these favorable paths.
Notably, in this open-label model, both age and earlier randomized treatment had only minor predictive effects. That does not mean they are irrelevant; it means that within this selected population (all younger than 68, all with normal baseline EF, and predominantly perfect nerve sparing), baseline sexual satisfaction and surgical approach dominated the model.
This is an uncomfortable message for anyone hoping that “the right pill” is the central determinant of post-prostatectomy sexual recovery. The paper does not dismiss pharmacologic rehabilitation. Instead, it places it in its correct role: adjunctive, potentially valuable, but not the main driver of unassisted recovery when baseline sexual vitality and surgical preservation are not aligned.
Turning These Insights into Better Preoperative Counseling: A Practical, Patient-Ready Framework
One of the strongest reasons to care about these models is that they produce an actionable counseling style. They encourage clinicians to stop treating “preoperative EF” as a single yes/no checkbox and instead evaluate baseline sexual functioning as a multi-domain profile—desire, confidence, satisfaction—not only hardness.
That matters because men with “normal EF” by IIEF-EF ≥22 are not homogeneous. In the REACTT baseline table (Table 1, page 19), mean IIEF-EF is high (~28), but sexual desire and satisfaction domains still vary. Those variations are predictive in the trees. In other words, a man can technically qualify as “normal EF” and still sit in a lower-predicted recovery branch if desire and confidence are low.
The models also reinforce that surgical quality remains central. The nerve-sparing score was simplified into “perfect” vs “not perfect,” with most men in REACTT having perfect scores, yet nerve sparing and robotic approach still emerged as predictors in favorable branches. That is a reminder that even within a high-quality trial, technical differences still matter—and that “robotic vs open” can be a proxy for multiple variables: visualization, surgeon experience, and consistency of nerve preservation. The authors explicitly caution that the trial was not designed to compare surgical approaches and did not account for surgeon volume or learning curves.
If you want a clinically useful translation of the decision-tree logic (without pretending it is a magic oracle), here is a compact heuristic grounded in the paper’s hierarchy:
- Start with baseline sexual vitality: desire, confidence, and intercourse satisfaction are not “soft” variables—they are leading predictors in this dataset.
- Then evaluate surgical preservation potential: robotic approach and high-quality nerve sparing repeatedly associate with better outcomes, especially in men with favorable baseline profiles.
- Use daily tadalafil as a support tool, not a guarantee: it appears most relevant for improving drug-assisted EF during treatment and may help certain subgroups, but it did not dominate unassisted recovery prediction after washout.
That is one list—kept intentionally short—because the point is not to overwhelm patients with “algorithmic medicine,” but to structure counseling around the variables that actually predicted outcomes in a large randomized dataset.
Limitations You Must Respect: Why These Trees Guide Conversations, Not Verdicts
This was a post hoc exploratory analysis. Decision trees are sensitive to which variables are included and can shift structure if new predictors are added. The authors acknowledge that although pruning and cross-validation improve robustness, the trees should be viewed as descriptive rather than confirmatory.
There are also population constraints. REACTT enrolled a selected group: relatively young men (<68), localized cancer (Gleason ≤7), normal baseline EF, and exclusion of significant comorbidities such as diabetes in many cases. This limits generalizability. The authors note that limited age range may explain why age was not strongly predictive here; in a broader cohort including older men and more comorbidity, age might re-emerge as a stronger predictor.
Baseline EF assessment timing is another challenge. The IIEF was collected after cancer diagnosis and biopsy, and psychological stress can influence sexual desire, confidence, and reporting. The authors explicitly raise concern about baseline reliability under cancer-related anxiety. That critique is not a weakness of the paper; it is an honest reminder that patient-reported sexual metrics are dynamic, especially in the context of a new cancer diagnosis.
Finally, nerve-sparing score (NSS) is surgeon-rated and subjective, and patients with very poor nerve sparing were excluded. Most men had perfect NSS, reducing variability in a variable that might otherwise be strongly predictive. Despite that, nerve-sparing quality still appeared in favorable branches—suggesting that even modest differences in preservation can matter.
The practical conclusion is straightforward: use these trees as structured guidance for counseling and expectation management, not as deterministic predictions. The value is not precision; it is clarity.
Conclusion: The Best Predictor Is Not a Pill—It’s the Patient Profile Plus Surgical Preservation, with Medication as a Rational Adjunct
This decision-tree modeling of REACTT data identified three baseline sexual characteristics as repeatedly influential predictors of erectile function recovery after bilateral nsRP: sexual desire, confidence to get and maintain an erection, and intercourse satisfaction. These factors often preceded surgical and treatment variables in predictive hierarchy, suggesting that who the patient is at baseline—psychosexually and functionally—strongly shapes what he can recover after surgery.
Surgical approach (especially robot-assisted laparoscopy) and the quality of nerve sparing also appeared as meaningful predictors in better outcome pathways, particularly among men with favorable baseline profiles. This reinforces a reality clinicians already suspect but patients sometimes overlook: erectile recovery is partly a surgical outcome, and partly a biologic and behavioral recovery process that extends long beyond the operating room.
Tadalafil once daily showed relevance in improving drug-assisted EF at the end of the double-blind treatment period in certain subgroups, yet randomized treatment did not drive prediction of unassisted recovery after washout—consistent with the primary REACTT endpoint. The rational clinical stance is therefore balanced: early daily tadalafil may support function and engagement, but the “recovery ceiling” is set mainly by baseline sexual vitality and the degree of anatomic preservation.
If you want a final line with a little professional irony: the decision tree suggests that the fastest way to improve post-prostatectomy counseling is not a new molecule—it is asking better questions before surgery, and then doing a truly conserving operation when oncologically safe.
FAQ
1) Does daily tadalafil guarantee spontaneous erectile recovery after nerve-sparing prostatectomy?
No. In REACTT-based modeling, tadalafil once daily was associated with better drug-assisted erectile function during treatment for some subgroups, but randomized treatment did not predict unassisted recovery after washout in the decision tree, consistent with REACTT’s primary outcome.
2) What were the strongest predictors of erectile recovery in this analysis?
The most consistent top-level predictors were baseline sexual profile measures: sexual desire (IIEF item 12), confidence (IIEF item 15), and intercourse satisfaction (IIEF satisfaction domain). Surgical approach and nerve-sparing quality became important predictors in favorable branches after those baseline factors.
3) How can clinicians use this information in real practice?
Use it for preoperative expectation management: assess desire, confidence, and satisfaction—not only “can you get an erection.” Discuss how surgical preservation (robotic approach, high-quality nerve sparing when feasible) and early rehabilitation may benefit men with favorable baseline profiles, while being honest that recovery remains variable and time-dependent.
