# Deriving life expectancy from FLIPI index data for FLIPI(3) High Risk

Deriving life expectancy from FLIPI index data for FLIPI(3) High Risk

I completely rewrote my very terribly simplistic initial question (currently erased) to take into account two key elements of advice from the original answer.
(1) Median years of survival is a good measure of life expectancy for patients with typical disease progression within a FLIPI score.
(2) Survival rates have drastically improved in the Rituximab era.

The above is from Figure 4 from this paper: Follicular Lymphoma International Prognostic Index 2004

The 6.16 years median life expectancy referenced by the answer below may be a very reliable measure of typical outcomes for those with typical FLIPI(3) disease progression.

Based on the advice in the answer I found a sufficiently large sample that recalibrated the overall survival rates for the Rituximab era: Follicular lymphoma in the modern era: survival, treatment outcomes, and identification of high-risk subgroups 2020

Because the number at risk for 10 years had 58% of the subjects drop out of the study we may not be able to trust the graph's 69% survival rate. Only 9% of the subjects dropped out at the 5 year mark so we can probably trust its 82% survival rate. This is a 55% improvement over the original FLIPI data 53%. The graph uses Kaplan-Meier to adjust for missing data.

When we simply multiply the above 6.18 year life expectancy by the 55% increased life expectancy provided by the Rituximab era we derive a 9.56 year life expectancy for patients with a FLIPI index score of 3 (eventually treated with Rituximab or equivalent) having typical disease progression.

Those calculations aren't correct. You can't reliably turn overall-survival percentages at any given year into a survival estimate at other times unless you know the shape of the survival curve over time and make certain assumptions. Sometimes there are sharp survival drops at early times followed by a plateau of long survival times, representing a situation in which some are "cured."

The paper you cite, however, shows survival curves for 3 groups based on FLIPI, in its Figure 4 reproduced below:

An index of 3 is in the "High" group. Although it's not strictly a "life expectancy," I think that a measure of survival like the median survival time (time at which half have died, survival probability of 0.5) is generally more informative.* That's just a bit beyond 60 months after diagnosis for the "High" group.

That's only a crude estimate for several reasons, and is unlikely to apply to you personally.

First, FLIPI doesn't take into account the details of the different risk factors (Age, Ann Arbor Stage, hemoglobin, LDH, and number of involved sites), just the number of those adverse features (Table 4 of the cited paper).

Second, survival curves often depend heavily on the particular patient population that was evaluated. The cited paper was based on an international study; the details of survival might well depend on residence location.

Third, and most important, that paper is now nearly 20 years old and is based on initial diagnoses from 30 to 37 years ago. Therapy and outcomes have improved substantially since then, particularly due to the development of rituximab, which was only approved in 1997 and thus was not available to most patients in the study you cite. See Freedman and Jacobsen, Follicular lymphoma: 2020 update on diagnosis and management, American Journal of Hematology, Volume 95, Issue 3, pages 316-327, for a more recent discussion.

This type of thing is always hard when you are personally involved. Make sure to discuss your concerns with your own clinicians, who are in the best position to evaluate your specific personal risks.

*For example, in a "cure" model the average survival time (life expectancy) can be quite high because of the numbers of individuals who survive the disease in question and die much later from other causes. The median survival time is when there's a 50/50 chance of having survived that long.

• Since I only wanted to derive the actual true measure I just realized that your answer may be better than my updated answer: 4.18 Average years of survival weighted by deaths per year simply ignores the 35% that live longer. Median 74 months until death does not ignore this. Feb 24 at 16:00
• @polcott I think that what you calculated is what's called "restricted mean survival," the average survival up to a specific time point. That has never made a lot of sense to me, although it's frequently used. Medians make a lot more sense: half have the event before the median survival time, half after.
– EdM
Feb 24 at 16:01
• Since my system ignores 35% of the data it is not as representative of the data. On the other hand my measure does make a reasonable guess of a plausible average worst case scenario. This give me a plausible range of what I can expect. This is much happier than my original 11 more months left based on my simplistic analysis. The much more precise estimate of my actual life expectancy is based on the degree of success of treatment. Treatment has not yet begun. Feb 24 at 16:05
• @polcott the thing that should make you even happier is rituximab, which was not available for most of the study on which you based your survival estimate. That survival curve was for a population that did not have rituximab available. You are not a member of that population. See the 2020 update linked in my answer.
– EdM
Feb 24 at 16:09
• I found another study and updated my analysis based on this (rituximab era) study. Although the 5 year survival rate is up to at least 75%, missing data seems make the graph of the table data totally out of sync for the 10 year data. The graph says 60% survival and the table says 29% survival (if we assume that all those the left the study died). Feb 24 at 22:59

The original answer was based on a very simplistic analysis of 53% five year overall-survival and 35% ten year overall-survival.

Those calculations aren't correct. You can't reliably turn overall-survival percentages at any given year into a survival estimate at other times unless you know the shape of the survival curve over time and make certain assumptions. EdM

I took two key aspects of this original answer to derive an updated analysis:

(1) My analysis changed its basis from 53% five year overall-survival and 35% ten year overall-survival to median years of survival. Median years of survival is a good measure of life expectancy for patients with typical disease progression. For patients with a FLIPI 3 score (based on the original FLIPI data) the median years of survival was 6.18 years.

(2) The above estimate can be made much more accurate when considering data from the Rituximab era. (shown above).

The 69% ten year survival is based on 58% of the subjects dropping out of the study. The median 13 years of survival is based on 98% of the subjects dropping out of the study. Because of this we will extrapolate median survival rate based on the 82% five year survival rate where only 9% of the participants dropped out.

A reasonable estimate of life expectancy for patients with a FLIPI 3 score (based on data from the Rituximab era) and typical disease progression is a 55% increase over the prior estimate: 6.18 * (82/53) = 9.56 years. (82% was interpolated directly from the graph pixels).

Patients with a FLIPI index score of 3 (eventually treated with Rituximab or equivalent) having typical disease progression have a life expectancy of 9.56 years.