Statistics and the Prognosis

The Median Isn't the Message

One of the best essays on statistics, as applied to cancer, was written by Steven Jay Gould. He was diagnosed with abdominal mesothelioma in 1982. Mr. Gould fought his cancer for 20 years before succumbing to it in 2002. I urge anyone diagnosed with any cancer to spend a few minutes reading The Median Isn't the Message found on Steve Dunn's CancerGuide.

Statistics, what are they?

From a post to the EC-Group by Dot Clear on October 28, 2001

The question is as follows: how long do patients live after esophagectomy surgery?

Say there are 15 subjects in a study, and these 15 subjects are followed for 7 years.

Number of subjects in the study: 15

Years lived after esophagectomy surgery: 1,1,1,1,1,1,1,2,4,5,5,5,5,5,7

Seven subjects live 1 year, one lives 2 years, one lives 4 years, five live 5 years, one lives 7+ years.

Mean (the average): Sum up all the values and divide that sum by the number of subjects.

Mean example:
(1+1+1+1+1+1+1+2+4+5+5+5+5+5+7)/15 = 45/15 = 3

The mean average is 3 years

Median (the middle): Find the value where 50 percent lie above and 50 percent lie below that value. If there's an even number of subjects, take the mean of the middle two values.

Median example:
1,1,1,1,1,1,1 -> 2 <- 4,5,5,5,5,5,7

The medan is 2 years

Mode (most popular): Find the value that shows up most often from all the values considered. If certain values appear the same number of times, there's more than one mode.

Mode example:
1 appears seven times, the most common value

The mode is 1 year

There are other methods to try and find a statistical value, not just by those methods given herein. Statistics tries to offer an overall value. This overall value is meant to represent the population, regardless of what method is used. Obviously an overall value cannot predict the real outcome for any one individual person. Each time an individual value "fails" to agree with the overall value, there is statistical error. Statistics come with errors; in fact, there is a whole sect of statistics that deals with errors.

So, how long do patients live after esophagectomy surgery: 1, 2, or 3 years?

The answer is... who knows! It could be 1, 2, 3, or any number of years.

Alas, the point is that no statistical value can tell you what is going to happen to you.

— Dot

Seeing Past the Statistics

Based on a post to the EC-Group by Cecily Dixon on November 7, 2001

One thing to bear in mind when it comes to statistics for survival rates is that they usually reflect a median (i.e., a midway point, as opposed to an overall mean or average). Also, EC survival medians reflect a wide range of patients, including patients with very advanced/inoperable cancer, patients who choose not to pursue treatment, patients who get treatment at subpar medical facilities, and patients who have other serious health problems. Another crucial point is that EC median survival rates do not differentiate between squamous cell EC and adenocarcinoma EC.

Statistics for survival rates are usually broken down by EC stage. However, there is a debate in the medical community about how accurate the current EC staging system is. Staging information can be obtained from websites like Medscape, CancerNet, Oncolink, etc. One of those websites (I think it was the website for the American Joint Committee on Cancer [AJCC]), makes the following disclaimer:

The current staging system for esophageal cancer is based largely on retrospective data from the Japanese Committee for Registration of Esophageal Carcinoma. It is most applicable to patients with squamous cell carcinoma of the upper- and middle-third of the esophagus, as opposed to the increasingly common distal esophageal and gastroesophageal junction adenocarcinoma. In particular, the classification of involved lymph nodes as M1 disease has been criticized. The presence of positive abdominal lymph nodes does not appear to carry as grave a prognosis for metastases to distant organs. Patients with regional and/or celiac axis lymphadenopathy should not necessarily be considered to have unresectable disease due to metastases. Complete resection of the primary tumor and appropriate lymphadenectomy should be attempted when possible.

After my dad underwent a series of diagnostic tests at Memorial Sloan Kettering (MSK), the surgeon at MSK staged my dad at 2a (T3N0M0). The current staging system for EC breaks down stage 2 into two sub-stages: 2a and 2b. The median five-year survival rate is 40 percent for stage 2a and 20 percent for stage 2b. Do I think that my dad has no more than a 40 percent chance of surviving for another five years? No. Many EC Group members who faced a worse prognosis than my dad now faces have passed the five-year mark with flying colors. And because cases of adenocarcinoma in the United States were extremely rare until ten years ago, almost no five-year-survival data had been collected on that form of EC prior to then. The latest survival rates that have been drawn from clinical trials initiated five years ago (involving patients who'd been diagnosed with EC at least five years ago) do not reflect the years beyond the five-year-mark that many patients will continue to live through. More years will have to pass before further statistics can be compiled on longer-term survival rates and begin to reflect recent and ongoing improvements in treatment. So the statistics are fairly limited in what they can tell us.

My father's oncologist (who administered my dad's chemo and radiation) thinks that my dad's cancer is in fact stage 1, not stage 2. (Based on the criteria under the current staging system, my dad is definitely stage 2a, since his tumor has penetrated the adventitia - the outermost wall of the esophagus; but because my dad has adenocarcinoma EC, and not squamous cell EC, his oncologist may have revised the staging based on what he believes to be my dad's survival outlook - i.e., it may be that the oncologist thinks that my dad's survival prospects are as good as those of a patient with stage 1 squamous cell carcinoma, as reflected by the Japanese staging system, and has revised my dad's staging accordingly). The oncologist has reviewed MSK's test results, but unlike the MSK surgeon, the oncologist has been seeing my dad on a weekly (sometimes daily) basis. So I'm feeling optimistic. To ease some of the distress that certain doctors cause by dryly telling a patient that the numbers donÕt look good for him or her, a patient should bear in mind that the current staging system for EC did not originate in the United States and has not been updated to specifically reflect adenocarcinoma (cases of squamous cell carcinoma have been quite common for decades in places like Japan, India, and Iran, which is why there is a staging system geared toward that form of EC and not toward adenocarcinoma EC, which has begun to be more prevalent only in the past decade. Perhaps this is due to America's love for junk food, since adenocarcinoma tends to stem from gastric reflux, whereas squamous cell carcinoma is associated more with long-term alcohol and cigarette consumption, the latter of which has been declining in America over the past couple of decades).

These may seem like minor points, but I don't think that they are. Many doctors tend to base their prognostic remarks on the overall median five-year survival rate for ALL patients with EC, which is 12 percent (24 percent for localized disease) as of 1999. That figure is way too broad for individual patients to go by, since it is based on innumerable factors. EC patients who have serious underlying health problems and do not pursue adequate treatment (or perhaps undergo no treatment at all) skew the statistics, making the midway point / median look grimmer than it is (and probably the median is much grimmer than the average). Chuck the figures out the window, I say. I might sound like Pollyanna if it weren't for the many EC Group members who are evidence of how little the statistics really mean. When I see messages confirming this, I print them out for my dad, since it's all good and well for me to make obligatory remarks about how he's not a mere statistic, but it doesn't carry nearly as much weight as a message from a survivor. I've printed out at least a dozen such messages since his diagnosis in mid-August, and they've been a tremendous morale-booster for my dad. This group has been an absolute god-send.

— Cecily in Park Slope, Brooklyn