April 10, 2012

A long time ago, in 2007 to be precise, my Halloween story showed how Real Big Brother would embrace rather than restrict Internet usage. For a rerun, read Evgeny Morozov's column, published a few days ago (*). It bemoans an "unnecessary and dangerous expansion of state power, in collaboration with companies, into our online - and offline - lives".

I do not claim my past fillip inspired the US and the UK governments. Nor do I seek applause for my predictive talents. I was but anticipating the obvious. As Evgeny Morozov's points out, one only needs "to force technology companies to build "back doors" into their services, making it possible to "wiretap" an exchange online as if it was a conversation by telephone". With current means of analyzing massive amounts of digital data, what at first may look like a mere "matter of degree" can indeed turn nationwide surveillance from a nightmarish dream into an efficient reality.

Since I started to ransack past outputs, why not quote Kevin Sack's article on kidney transplantation from living donors in the US (**)? "Leaders [...] recommended last week that a single nationwide registry should oversee the complex matching process needed to expand the use of {...] paired donations". For, three years ago as it happened, I proposed a nationwide matching system (1). In vain, but not before I learnt a few facts of life.

The overall case is well known. Fueled in part by the ever waxing tide of morbid obesity and diabetes in America, kidney failure condemns its patients to life on dialysis unless they find a compatible donor. As the pool of deceased donors of healthy organs is limited, short of a voluntary increase in the number of car accidents among other even more repellent recourses, their hope is to pair with a living donor.

This is not as simple as it sounds. Donating a replenishable body part, like blood or bone marrow, is one thing. Another to donate a kidney, an irreversible, inherently more risky decision. In line with the Declaration of Istanbul (2), the United States currently outlaws commercial markets for kidney procurement (3). Even so-called altruistic donors are routinely screened lest they fall victims to their better instincts, charitable but unfocused. Most patients therefore pair with a loved one unless blessed with having met a true neighbor, willing to act from compassionate but personal love.

Unfortunately blood and tissue types must match before a donor can give a kidney to a patient. Blood type alone fails more than 20% of all pairs. Hence the idea of pooling incompatible pairs to match registered patients and donors globally. Since it was his or her intent to give a kidney, every donor should be glad to do so even to a different patient, as long as some other donor gave to the patient who prompted this intent in the first place.

Rather than simply crossmatching two incompatible pairs, the donor in the first one giving to the patient in the other one and conversely, one can extend the search along whole chains, where the donor of one pair helps the patient in the next pair until the chain loops upon itself. Thus "in December, the National Kidney Registry [...] completed a record chain of 30 successful transplants, over four months at 17 hospitals".

Within a pool, the better the optimization algorithm, the larger the total number of matches. There is some irony in the fact that some of these algorithms can be traced to the study of so-called two-sided matching markets by Al Roth from Harvard University (4). But the use of the word market should not fool one into thinking a pool implements a "price market". On the contrary a kidney exchange is an instance of what I call a "value market", albeit an extreme one which eliminates all criteria linked to price or any other form of compensation besides the cross donation itself.

For a given algorithm, the larger the pool, the higher the probability of a match. A naive reader at this point may well ask why there is no universal pool to register all donors as there already is for patients (5). Wouldn't it deliver the highest possible number of matches? Who could object?

There is no shortage of reasons to explain the staying power of the status quo which includes "several national and regional registries [...] as well as a pilot program sponsored by the federal government". Why would the transplant center managers want to give up their decision power to some new, external bureaucracy? Before calling them callous though, we must grasp an essential fact. Living donors are just that, living human beings.

Contrary to the situation with deceased donors, a living kidney donation is not an emergency, quite the opposite. Not only is a donor's kidney available beyond a very short, impredictable window, but no transplant can be executed without getting the fully informed consent of the donor.

Yet today's registries freeze their membership each time they run their algorithms and all participating pairs must blindly abide by the global result. True, at the last minute, a donor may decline the match found by a registry but under the pressure of failing, not only the paired patient, but all registered patients, since the withdrawing of just one pair invalidates any claim to optimality (6). The larger the scale, the more brittle the process.

"Market segmentation" leads to "cherry-picking" which "undermines optimization" in the words of "Sommer Gentry, a mathematician at the Naval Academy". But why fight transplant politics and donor's remorse when the whole concept behind current and planned registries may be at fault?

To be truly efficient, the nationwide exchange pool should include all pairs, not just those which are incompatible (7). Then, if for argument's sake we disregard tissue compatibility, no longer a donor whose blood type is "universal donor" would "waste" a kidney on a paired patient whose blood type is "universal recipient", when it could be cross exchanged with a patient whose blood type is harder to match, such as "universal donor" (8).

Obviously this is ethical only if, all other things being equal, the outcome is no worse for crossed pair donations than for regular ones. To no avail I proposed to several transplant leaders to check this hypothesis. Still, were this correct, wouldn't the scale heighten the brittleness I count as a flaw?

What if the users remained in charge at all times, under a formal set of rules to enforce ethics and compute matches incrementally? In such a system, donors' data would never leave the office of their personal physician. No one but the latter would know of the search for a cross match until a result was accepted by all parties directly concerned. Within ethical bounds, compatible pairs would be free to seek better tissue matches than their own.

Certainly such a system would not deliver what current registries claim to be the optimum. But what does optimal mean if mathematics require no pair leave nor enter the pool being optimized? Unless the pool is tiny, and we know such tiny pools are not optimal to begin with, this is unrealistic.

What bureaucracy though would adopt such a decentralized solution which protects every one's eprivacy? Evgeny Morozov would think it unlikely.

If what counts were to get a discrete back door on all living donors, the current approach would be a must but a single registry only nice to have.

Philippe Coueignoux

• (*) ... Beware the unholy alliance of state and internet, by Evgeny Morozov (Financial Times) - Apr 4, 2012
• (**) . Experts Recommend Single Registry to Oversee Kidney Transplant Donations, by Kevin Sack (New York Times) - Apr 1, 2012
• (1) to request more information about NOAM (Network for Organ Automated Matching), please send an email to the author
• (2) see the Declaration of Istanbul for details
• (3) see NOTA in the wikipedia
• (4) see this introduction by Al Roth, Harvard University
• (5) with a touch of Newspeak, this patient registry, managed by UNOS, is often called the "deceased donor waiting list"
• (6) this is especially true when chains are allowed by the algorithm. A withdrawal within a chain breaks it, stranding pairs otherwise matchable.
• (7) in the US in 2011, there were 5769 kidney transplants from living donors (per UNOS), only 429 of those from cross exchanges (per Kevin Sack).
• (8) for more details see the presentation and the simulation of a simple-minded matching model based on blood type.