The Adjacent Possible and the Buccal Mucosa Graft

Why It Took 125 Years to Move a Mouth Graft into a Ureter, and What That Tells Us About Innovation in Surgery

1. Sapezhko in Kyiv

As early as 1890, a surgeon named Kirill Mikhailovich Sapezhko, working in Kyiv, took oral mucosa from a patient’s mouth and used it to reconstruct the patient’s urethra. In 1894, he published a fuller account of this work, describing mucosa from the lip and mouth in patients with urethral disease. By the standards of the time, the operations appear to have worked.¹

And then the idea mostly disappeared.

This is worth pausing on, because we have a strong cultural narrative about innovation in which a clever idea, once discovered, propagates outward like a brushfire. Sapezhko’s idea was not just clever. In retrospect, it was correct. Today, oral mucosa is the AUA-guideline first-choice graft material for urethral reconstruction.² Buccal mucosa became the workhorse because it is hairless, accustomed to a wet environment, resistant to infection, easy to harvest, and well vascularized. The reasons it works now are basically the same reasons it worked in 1890.

So why did the brushfire fail to spread?

There is an instructive contrast in the other direction. In 1846, William Morton’s ether demonstration at Massachusetts General Hospital was an operation that succeeded socially and failed clinically: anesthesia entered the canon, but the patient’s tumor recurred within months. Sapezhko’s was an operation that succeeded clinically and failed socially. The technique worked, and was forgotten.

The Ether Dome got an amphitheater. Sapezhko got a footnote.

Standard explanations for that asymmetry, including language barriers, the disruptions of two world wars, and the personality politics of European surgery, are real but incomplete. The deeper answer is that Sapezhko had walked into a room whose connecting doors had not yet been built.

This is an essay about how innovations wait. About how you can have the right idea a hundred years too early. And about what happens when the rooms finally connect, which they did, slowly, in the case of oral mucosa for urinary reconstruction, between roughly 1990 and 2015. It is also an essay about why innovation in surgery looks the way it does: slow, late, sometimes simultaneous, sometimes inevitable, almost never heroic.

2. Kauffman’s Mansion

In the mid-1990s, the theoretical biologist Stuart Kauffman developed a concept he called the adjacent possible.³ The simplest version: at any given moment in evolution, technology, or culture, there is not an infinite range of next steps available. There is a bounded set of possibilities reachable from the current state, given the current pieces. Enter one of those possibilities and a new set becomes reachable. But you cannot leap across walls.

Steven Johnson later popularized this as an expanding house. I will call it a mansion, because surgery has too many rooms. You begin in a small chamber. Each room you enter unlocks doors to new rooms. The mansion is, in effect, larger than it was a moment ago, because possibilities exist only relative to the current configuration.

The carbon atoms of prebiotic Earth could not arrange themselves into a sunflower. They first had to arrange into amino acids, then proteins, then membranes, then self-replicating systems, then multicellular forms, then vascular tissue, before anything recognizable as a sunflower could exist. There is no wormhole from primordial soup to angiosperm.

This is a profoundly anti-heroic theory of innovation. No individual genius leaps over multiple rooms. By the late nineteenth century the lightbulb was inevitable: Swan, Edison, and others converged on it because the prior rooms (vacuum technology, electrical theory, filament materials) had finally opened. The story we tell about Edison is partly about invention, and mostly about patents, manufacturing, and marketing.

The reverse also happens. Some people leap too early. Their idea sits in a corner until someone else can reach it from a different direction. Mendel published his pea experiments in 1866, and the work then sat orphaned for decades. The cellular and statistical languages needed to make Mendel legible had not yet connected to his findings. When Mendelian genetics was revived around 1900, it was because those doors had opened.

Sapezhko was something like Mendel. He had walked into a room before the corridors existed. Oral mucosa worked. He could see that it worked. But the surrounding technical, biological, and social infrastructure was not yet there.

Humby revived the idea in 1941 for hypospadias. The same thing happened. It was only when Buerger and El-Kasaby brought buccal mucosa back into adult urethroplasty in 1992 and 1993 that the technique finally took.⁴ Three rediscoveries across roughly a century, and only one of them stuck.

Here is the central claim of this essay: The adjacent possible is structural, but the realization of the adjacent possible is sociological. Whether a given room can be entered is determined by which other rooms are open. Whether it actually gets entered, and by whom, and how quickly the door gets propped open behind them, is determined by the people in the corridor.

3. Which Rooms Had to Open

What had to change between Sapezhko’s oral mucosa urethroplasty and the routine practice of robotic buccal mucosa graft ureteroplasty? The sequence is instructive because each step looks modest in isolation. The drama is in the accumulation.

The biological room. Before antibiotics, free grafts in the urinary tract carried a real risk of infection and abscess. Penicillin and its successors did not make grafts magically safe, but they made infection controllable enough that surgeons could plan reconstruction around healing rather than around the next septic event. Surgeons also had to understand graft take: imbibition, inosculation, neovascularization. Once they understood that a free graft survives by absorbing nutrients from a vascular bed before developing its own blood supply, they could predict which beds would work. The omentum, with its vascularity and mobility, became a useful supporting bed for ureteral reconstruction.

The technical room. Fine reconstruction depends on tools that do not fight the operator. Silk and catgut had limitations: inflammation, variable absorption, tissue drag. The late-twentieth-century maturation of synthetic absorbable monofilament sutures gave surgeons more predictable handles on delicate tissue. Magnification mattered too. Loupes made it possible to see graft edges and watertight closures at a useful scale.

Buccal mucosa urethroplasty itself then had to mature. Buerger and El-Kasaby reintroduced it in the early 1990s. Morey and McAninch refined harvest and deployment in adult urethroplasty. Barbagli helped define dorsal onlay techniques. By the early 2000s, buccal mucosa urethroplasty had become routine. By 2016, when I sat on committee for the AUA urethral stricture guideline, oral mucosa was specified as the first-choice graft material for urethroplasty without much controversy.

Once that room was unlocked, the next door became visible: perhaps the same trick would work in the upper tract. In 1999, J. H. Naude in South Africa reported six patients with complex ureteric lesions reconstructed using buccal mucosa.⁵ Patency and drainage were maintained at follow-up. It was proof of concept. But it was not yet scalable. Open upper-tract reconstruction still meant a large incision, prolonged recovery, and a morbidity profile that made the operation hard to generalize. The room was open. The corridor was still unpleasant.

The minimally invasive room. The da Vinci platform received its first FDA clearance for assisting surgery in the late 1990s and broader operative clearance in 2000. Through the 2000s it diffused widely. The robot did not invent new operations. It lowered the activation energy for operations that already existed. Wristed instruments at depth made suturing in the retroperitoneum credible in a way pure laparoscopy never quite did for most surgeons.

Robotic pyeloplasty became the training procedure for robotic upper-tract reconstruction. By the early 2010s, many programs were doing it fluently. At NYU, Michael Stifelman had built a high-volume robotic upper-tract reconstruction program. A 250-patient consecutive series published in 2015 captured what had happened: the upper urinary tract had become familiar robotic territory.

The epistemic room. Intraoperative flexible ureteroscopy allowed the surgeon to pass a scope through the bladder and up to the stricture during the robotic case. The scope could identify the lumen, transilluminate the ureter, and force the map to match the territory in real time. Near-infrared fluorescence imaging with indocyanine green addressed a related problem: perfusion. Ureteroscopy helped define the lumen. Fluorescence helped assess the tissue. Together, they narrowed the gap between what the surgeon thought was there and what was actually there.

Ureteral rest mattered too. Decompressing the kidney and removing the stent before reconstruction allows inflammation to settle and the true stricture length to declare itself. This is a small, almost embarrassingly simple insight, derived partly by analogy from urethroplasty. It is also surprisingly important. A stented ureter is either an optimist or a liar.

By 2014, these four rooms were open simultaneously. The door to robotic buccal mucosa graft ureteroplasty was sitting there, unlocked.

Michael Stifelman and I performed the first cases together at NYU between late 2013 and 2014, and we reported a four-patient series the following year.⁶ Other groups walked through the same open door. A 19-patient multi-institutional cohort was published in 2018. An intermediate-term update with 54 patients followed in 2021. A decade-of-experience retrospective, published online in 2024 and appearing in print afterward, reported 169 patients.⁷

The technique stopped being a curiosity. It became a standard option for proximal and mid-ureteral strictures that, ten years earlier, would often have led to ileal ureter, renal autotransplantation, or indefinite ureteral stenting with exchanges every few months.

This is what the adjacent possible looks like in slow motion. The same basic idea Sapezhko had in 1890, take epithelium from one place and use it to reconstruct a tube somewhere else, was waiting for the rooms to connect. The 1890 room and the 2014 room are not identical, but they resonate.

4. Dallas, New York, Taipei

So far I have described the structural side. The mansion has rooms; rooms have doors; doors connect when the prior rooms are open. But the mansion does not actually contain rooms. It contains people.

Doors connect when people trained in different rooms find one another and recognize that what each knows is the missing piece of what the other is trying to do. This is the part of innovation that is hardest to schedule and easiest to underrate.

From 2012 to 2013, I trained at UT Southwestern in Dallas under Allen Morey. Morey, with Jack McAninch in San Francisco, helped make buccal mucosa part of modern adult urethral reconstruction. By the time I left fellowship, I had harvested dozens of buccal grafts. The movement had become routine: mark the dimensions inside the cheek, infiltrate, incise, raise the graft, defat, close. I did not think of myself as having done anything special. I had inherited a skill that others had spent decades making teachable.

In 2013, I joined NYU. Michael Stifelman was the director of robotic surgery there. By then, he had spent years building one of the highest-volume robotic upper-tract reconstruction practices in the country. He had the retroperitoneum mastered with an arsenal of wristed instruments. He had the team, the docking, the flow. He had been waiting for a graft material that solved the problem of the long proximal stricture without requiring bowel or renal autotransplant.

I had walked out of one room. He was standing in another. The door between us did not need to be invented. It was already there.

We did the first cases together. Neither of us alone could have done them as quickly or as confidently. The 2015 paper has my name first because I brought the graft technique with me, but the operation is unimaginable without his program and platform.

In September 2014, Stifelman and I flew to the World Congress of Endourology in Taipei. I was working on the manuscript that would become the 2015 Urology paper. Louis Kavoussi was on the same flight. Kavoussi had performed early laparoscopic ureteral and renal operations and had written extensively about minimally invasive approaches to benign ureteral disease. The paper I was working on was, in part, a small attempt to make some of that more morbid work unnecessary.

Two surgeons, working on different ends of the same problem and separated by a generation, could end up a few rows apart on the same flight to a urology meeting in Taiwan. That density is not nothing. It is much of what enables the adjacent possible to become real: not a single conversation, but a field crowded enough that the relevant people are already near one another.

At the meeting, we met Daniel Eun, who ran a high-volume robotic reconstructive practice at Temple in Philadelphia and had been thinking along similar lines. Eun adopted the technique quickly. His own series was published in 2017. Out of those overlapping efforts grew the Collaborative of Reconstructive Robotic Ureteral Surgery, the multi-institutional database that produced the major large series of robotic buccal mucosa graft ureteroplasty.

I am not saying the technique would not have happened without these specific people. It probably would have. I am saying it happened on the timeline it did, with the dissemination pattern it had, because of those encounters.

Sapezhko in Kyiv had no one to meet. There was no conference where a reconstructive surgeon who routinely harvested oral mucosa and a roboticist with hundreds of upper-tract reconstructions could find each other. By 2014, there was. The mansion had become populated enough that the right encounters were no longer rare. They were close to inevitable, given enough Taipeis.

When you read about a technique that suddenly became standard, look for the conference where its developers first met. Almost every time, you will find one.

5. The Case Against

Two objections deserve serious treatment.

The first is the gadget critique. Robotic buccal mucosa graft ureteroplasty, the critic says, is a fancy version of an open operation. The robot does not materially improve the outcome. It mostly enables surgical complexity creep, drives up case costs, and produces the illusion of innovation where there is mostly novelty. Published success rates around 85 to 90 percent are not obviously better than what Naude reported with open technique in 1999. This is the Red Queen Effect dressed in haute couture: hospitals running faster to stay in the same place.

The second is the survivorship critique. It says: you celebrate robotic buccal mucosa graft ureteroplasty because it worked. The graveyard of failed surgical innovations is enormous, and we do not write essays about the operations that quietly disappeared. The adjacent-possible framing makes every successful innovation look inevitable in retrospect, which is just hindsight bias with extra steps.

The gadget critique is correct that the robot is not magic, and that costs are real. It is wrong, I think, about the comparator.

The relevant comparator was not open buccal mucosa graft ureteroplasty. For many patients with a long proximal ureteral stricture, the realistic prior alternatives were chronic drainage, ileal ureter, or renal autotransplantation. Chronic ureteral stenting means a plastic tube indwelling from kidney to bladder, exchanged every few months, accumulating encrustation, infections, pain, and emergency visits. Ileal ureter means bowel in the urinary tract, with mucus, stones, and metabolic consequences. Renal autotransplantation means a major operation with vascular reconstruction and ischemia, selecting heavily for patients healthy and motivated enough to accept it.

Robotic buccal mucosa graft ureteroplasty is not merely a fancy version of open ureteroplasty for these patients. It is a substitute for either a lifetime of tubes or a category of operation much more morbid than the stricture itself. The cost arithmetic looks different when the comparator is the right one. Naude’s open technique was elegant, but it was never a real option for most patients in most American hospitals. The recovery and risk profile selected against it.

The survivorship critique is stronger. Most surgical innovations fail. Most are forgotten. The published literature is a censored sample, biased toward the cases that worked, the surgeons who persisted, and the patients who remained in follow-up. The denominator is everyone who tried. The numerator is who reported. If we count only the survivors, we overweigh inevitability and underweigh luck.

But the adjacent possible, properly understood, does not say every innovation is inevitable. It says certain innovations cannot happen until their rooms are open. It is silent about which open rooms turn out to be useful. We discover usefulness only by walking in.

Some rooms are dead ends. Some lead to durable techniques. We cannot tell from the threshold which kind we are entering. We can only walk in, look around, and report back.

So both critiques are partly right. The robot is not magic. Most innovations fail. The adjacent possible is not a guarantee of progress. It is a structural explanation for timing and for simultaneous discovery. Many people walked toward the buccal mucosa graft ureteroplasty room at roughly the same time because the doors had finally opened. The fact that the technique worked is a separate question. We did not know the answer in 2014.

6. Wicked Mansions

In an earlier essay, The Wicked Problem of Surgical Failure, I argued that surgery is a wicked learning environment in Robin Hogarth’s sense: feedback is delayed, noisy, and biased, with rules that change faster than expertise accumulates.⁸ The adjacent possible interacts with wickedness in a specific and uncomfortable way.

In a kind learning environment, innovation is fast. A new chess opening can be tested in thousands of games within a year. The feedback loop is short. The rules are stable. The outcome is visible. The room you walked into is quickly evaluated and either occupied or abandoned.

In surgery, the feedback loop is measured in years. By the time you know whether your novel reconstruction holds up at five years, you have done hundreds more cases. The signal-to-noise ratio is poor enough that even good techniques can look ambiguous in small series. Innovations get adopted on small series with short follow-up, refuted on larger series with longer follow-up, and sometimes adopted again when the technique is refined.

The adjacent possible expands in surgery like a vine through fog. You can feel for the next branch, but you cannot see far ahead.

This has several implications.

First: surgical innovation will always look slow compared with innovation in software, chemistry, or consumer technology. The validation cost is high. That is not a moral failing. It is the structure of the problem. Surgeons are conservative because the cost of being wrong is paid by someone else’s body.

Second: early adopters bear most of the epistemic risk. The rooms that look open might not lead anywhere useful, and only patient follow-up at five and ten years can tell. The early adopter has incentives to claim more confidence than the data warrant, because confidence attracts patients, referrals, industry attention, and grants. So we should expect early reports of new techniques to be more optimistic than they ought to be, including my own early reports of robotic buccal mucosa graft ureteroplasty.

There is a related observation about the cost of being early. The early adopter pays a tax in money, operative time, failed cases, and reputational risk, in exchange for an option on whatever lies behind the next door. The static cost-benefit analysis usually says the option is not worth the price. The dynamic analysis, run over enough years, sometimes says it is. We do not know in advance which kind of room we are walking into. We pay the tax anyway because the alternative is leaving the door closed.

Third: the most valuable infrastructure for surgical innovation is often not the gadget itself, but the registries, networks, and follow-up systems that let us see which rooms turned out to be useful. The robot unlocked many rooms. The multi-institutional collaborative tells us which rooms were worth entering. The technology and the epistemics are not the same room. Neither substitutes for the other.

Sapezhko walked into a room that would not connect to much else for more than a century. He could not have known that. He had the idea, a few patients, and the willingness to try.

The mansion has grown since then. The rooms are more numerous. The doors are better lit. But the work has not really changed.

Look at what is currently possible. Look at what your patient needs. Find the door that nobody has opened. Walk through it carefully. Report back honestly.

Then, decades later, someone will write about the room you opened as if its discovery were inevitable, when really it was just you, or me, making a calibrated bet in the fog.

That is what innovation in a wicked environment actually looks like.

Not a leap.

A walk.

Notes

1.     Sapezhko, K. M., “Toward the question of the surgical treatment of urethral defects” [in Russian], 1890; expanded as “On the urethroplasty by oral mucosa” in 1894. The historical recovery of these papers and the case for Sapezhko’s priority are reviewed in Korneyev I., Ilyin D., Schultheiss D., Chapple C., “The First Oral Mucosal Graft Urethroplasty Was Carried Out in the 19th Century: The Pioneering Experience of Kirill Sapezhko (1857–1928),” European Urology 62, no. 4 (2012): 624–27. The Korneyev paper is itself an artifact of the rooms-connecting story I am telling: it took a multilingual urologic-historical team to translate Sapezhko’s Russian-language case reports into the English literature more than a century after they were written. Without that translation, Sapezhko remained a name that surface-level histories of urethroplasty mentioned in passing or omitted entirely.

2.    Wessells H., Angermeier K. W., Elliott S., Gonzalez C. M., Kodama R., Peterson A. C., Reston J., Rourke K., Stoffel J. T., Vanni A. J., Voelzke B. B., Zhao L., Santucci R. A., “Male Urethral Stricture: American Urological Association Guideline,” Journal of Urology 197, no. 1 (2017): 182–90. doi:10.1016/j.juro.2016.07.087. Although published in print in 2017, this is the original 2016 AUA Male Urethral Stricture Guideline. The guideline specifies oral mucosa as the first-choice graft material when substitution urethroplasty requires graft tissue. The biology is well-established and explains the durability of the technique across very different surgical eras: oral mucosa has a thick, nonkeratinized epithelium adapted to a wet environment; a thin, vascular lamina propria that supports imbibition, inosculation, and neovascularization; and a contracture profile favorable compared with skin.

3.    Stuart A. Kauffman, At Home in the Universe: The Search for the Laws of Self-Organization and Complexity (Oxford University Press, 1995); the more developed treatment, including the explicit framing of the adjacent possible as a generative principle in evolutionary and economic systems, is in Kauffman, Investigations (Oxford University Press, 2000). Most readers first encounter the term through Steven Johnson, Where Good Ideas Come From: The Natural History of Innovation (Riverhead, 2010). The expanding-house metaphor is Johnson’s; the surgical mansion in this essay is mine and is meant to do specific work that Johnson’s house does not, namely to keep the people in the corridors visible.

4.    Humby G., Twistington Higgins T., “A One-Stage Operation for Hypospadias,” British Journal of Surgery 29, no. 113 (1941): 84–92. Humby used full-thickness oral mucosa for pediatric hypospadias repair and reported good results, though he treated buccal grafting as a tool for a specific pediatric problem rather than as a general substitution material. Bürger R. A., Müller S. C., el-Damanhoury H., et al., “The Buccal Mucosal Graft for Urethral Reconstruction: A Preliminary Report,” Journal of Urology 147, no. 3 (1992): 662–64. El-Kasaby A. W., Fath-Alla M., Noweir A. M., et al., “The Use of Buccal Mucosa Patch Graft in the Management of Anterior Urethral Strictures,” Journal of Urology 149, no. 2 (1993): 276–78. Bürger and El-Kasaby reintroduced buccal mucosa for adult urethral stricture nearly simultaneously, in different countries, without (as far as the published record indicates) explicit knowledge of Sapezhko or full engagement with Humby’s prior work. The same idea arrived three times before it stuck.  The technique kept being reachable from the existing surgical state, which is what one would expect of an open room that the corridors had not yet linked.

5.    Naude J. H., “Buccal Mucosal Grafts in the Treatment of Ureteric Lesions,” BJU International 83, no. 7 (1999): 751–54. Six patients, open technique, with patency and drainage maintained at the reported follow-up. The cohort was small and the morbidity of the open approach kept the technique from generalizing. The proof of concept sat for roughly fifteen years before the minimally invasive room opened wide enough to make the operation practical at scale, which is itself a useful data point on how long the latency between open-room and reliable-occupancy can be in a wicked environment.

6.    Zhao L. C., Yamaguchi Y., Bryk D. J., Adelstein S. A., Stifelman M. D., “Robot-Assisted Ureteral Reconstruction Using Buccal Mucosa,” Urology 86, no. 3 (2015): 634–38. Four patients, single institution, short follow-up, conclusions appropriately tentative on paper and probably not tentative enough in retrospect. As I note in the body, early reports of new techniques are systematically more optimistic than they ought to be, and our own first paper was not exempt from that pattern. The honest description of the 2015 series is that it was a credible proof of concept that we were correct to publish and that no one, including us, should have read as proof of durability.

7.     Zhao L. C., Weinberg A. C., Lee Z., et al., “Robotic Ureteral Reconstruction Using Buccal Mucosa Grafts: A Multi-institutional Experience,” European Urology 73, no. 3 (2018): 419–26. (19 patients across multiple centers). Lee Z., Lee M., Koster H., et al., “A Multi-Institutional Experience with Robotic Ureteroplasty with Buccal Mucosa Graft: An Updated Analysis of Intermediate-Term Outcomes,” Urology 147 (2021): 306–10 (54 patients, longer follow-up). Chao B. W., Raver M., Lin J. S., Zhao K., Lee M., Gelman S., Stifelman M., Zhao L. C., Eun D. D., “Robotic Buccal Mucosa Graft Ureteroplasty: A Decade of Experience From a Multi-institutional Cohort,” Urology 197 (2025): 174–79. (169 patients). The denominator grew roughly an order of magnitude across these series and the headline success rates held in the 85–90 percent range. Whether they will hold at twenty-year follow-up is a question we cannot yet answer. Daniel Eun’s independent series and the subsequent Collaborative of Reconstructive Robotic Ureteral Surgery registry are what make these multi-center denominators possible; the registry is the part of the infrastructure that gets the least credit and did the most work.

8.    Robin M. Hogarth, Educating Intuition (University of Chicago Press, 2001); and Hogarth, Lejarraga, and Soyer, “The Two Settings of Kind and Wicked Learning Environments,” Current Directions in Psychological Science 24, no. 5 (2015): 379–85. I treat the surgical case at greater length in “The Wicked Problem of Surgical Failure” (leezhaomd.org, 2026), and the implications for surgical training in “The Endless Residency” (leezhaomd.org, 2026). The relationship between wicked learning and the adjacent possible is, I think, the core of the present essay: structurally reachable rooms in a wicked environment carry an unusually long latency between first entry and reliable evaluation, and this latency is not a defect of surgical culture but a feature of the underlying problem of working on human bodies.