Sekeres, Mikkael A. MD, MS

doi: 10.1097/

It's the most important question on the minds of everyone in the examination room of our cancer clinics, though few are brave enough to actually ask it.

“How long do I have?”

It's the first thing I would want to know if I were diagnosed with cancer. Heck, I'd love to know the answer in the absence of any cancer. And we all have our ways of answering it.

“It's hard to tell.”

“Everyone is different.”

“You could get hit by a bus crossing Euclid Avenue tomorrow, I'd still look both ways.”

“You'll outlive me.”

And then, in the 1980s and 1990s, we got some help answering the question, because of two major advances in cancer. No, I'm not talking about monoclonal antibodies and the advent of PET scans (don't get me started on PET scans). I mean the systematic collection of disease-specific clinical data, and the application of sophisticated statistical analyses, such as multivariable regression models.


Well, think about it. Most cancers are relatively uncommon, when we compare them to other medical conditions, such as hypertension and heart disease. So, to be able to say something scientifically rigorous about subtypes who seem to do better than others, outside of personal experience (and, as George Canellos once told me, “the plural of anecdote is not data”) required a concerted effort to first document basic pathologic and demographic information about patients, often across multiple treatment centers. Next, these data had to be analyzed in a way that adjusted for variables that might have been related to each other, thus hiding (or exaggerating) an effect.

For example, if older age at diagnosis leads to a worse survival with lung cancer, and duration of smoking also leads to a worse outcome, and both smoking duration and age are related to each other, to analyze the impact of duration of smoking accurately, you have to adjust for age, as someone who is older will likely smoke for a longer period of time.

The other key to this equation was that computers came of age in a way that these sophisticated analyses became accessible to anyone, and not just to engineering departments or the rare statistical cores. What resulted were prognostic scoring systems that allowed us to actually give evidence-based estimates that, for the first time, could be approximately tailored to the person sitting in front of us. Two clear examples from the hematologic malignancies world are the International Prognostic Index (IPI) for non-Hodgkin lymphoma, and the International Prognostic Scoring System (IPSS) for myelodysplastic syndromes (MDS).

Back to Top | Article Outline

International Prognostic Scoring System

Let's take the IPSS as an example. In 1997, Peter Greenberg and colleagues from the International MDS Risk Analysis Workshop (IMRAW) reported on 816 untreated MDS patients derived from seven databases established in the 1980s and early 1990s. They examined a bunch of information about these patients, including their pathologic subtypes, presenting blood counts, and cytogenetic abnormalities, determined how much each factor contributed to a patient's risk of transforming to acute myeloid leukemia and to survival using regression analyses, and assigned a point value to these factors.

If your patient is assigned a score of 0 (let's say she has an isolated anemia, no excess bone marrow blasts, and normal cytogenetics), she will live, on average, almost six years. If, on the other hand, she is unfortunate enough to have pancytopenia, complex cytogenetics, and 18% bone marrow blasts, in all likelihood she will not be alive six months from now.

The IPSS has really stood the test of time, and has been used in ways its developers may not have anticipated. Sure, it helps in giving a rough estimate of survival to a newly diagnosed patient. But it also has been incorporated into drug labels.

In other words, rather than simply a prognostic system, it has become a method of determining what treatment a patient should be offered, which could, theoretically (if a treatment prolongs survival), alter its own ability to prognosticate!

Moreover, it was developed at a time when MDS diagnoses were relatively infrequent (compared to today, as evidenced by incidence rate trends), and when no therapy was approved to treat MDS. It is thus unclear whether it retains its accuracy in patients who have been treated, and/or who are not recently diagnosed. Yet, it is actively being used in potential registration studies of previously treated patients to determine study eligibility.

A bit of a sticky wicket, as the Brits might say.

Needless to say the IPSS, along with other prognostic scoring systems, needs to be updated as new therapies are introduced and as disease populations themselves evolve. But this is a hard sell—oncologists get used to a given prognostic system (meaning, we can remember it without having to consult an online textbook!), and there is the fear that introducing a new system will undermine the uniformity with which established prognostic systems are applied. So, to do so, requires a truly collaborative, international effort, with world-wide buy-in.

(Signal the cavalry trumpets) For the IPSS, help is on the way.

Once again, Peter Greenberg has spearheaded an international working group, this time to revise the IPSS, and presented the initial version at the MDS meeting in Edinburgh in May of this year. And this time, the IPSS-R (as it will be called) is based on over 15,000 patients from 17 databases worldwide, and will include versions for both untreated and for treated patients, with a goal of being dynamic—meaning, as a person's MDS evolves over time, that person's prognosis may also change, and disease characteristics will dictate that a given person jumps from one survival curve to another.

Imagine that—a personalized prognostic tool that is sensitive to the rate of disease evolution and to therapeutic evolution. If we are going to use prognostic scoring tools or indices to answer the most important question on the minds of everyone in the examination room, and to make treatment and regulatory decisions, we should make it a national goal to update them regularly. That way, we can share our successes in treating cancer with the people who have it.

Thanks to Paul Elson, ScD, for his critical input to this article.

© 2011 Lippincott Williams & Wilkins, Inc.
Home  Clinical Resource Center
Current Issue       Search OT
Archives Get OT Enews
Blogs Email us!