At a conference at Harvard Medical School the other day a speaker made a simple comment: earlier diagnosis improves outcomes. With some thought, I realized that this is unquestionably true.
Let’s imagine a disease that has no cure, and no therapy can slow its progression. The disease starts with, say, an environmental trigger of a genetic defect. At exposure (and before), we’ll call our patient’s symptoms zero. Eventually, the disease kills the patient, when the symptoms are, say, 10. Using current tests, we can’t diagnose the disease until the patient’s symptoms reach level 5, and because of the complexity, most people see several doctors before they are diagnosed and so the average patient at diagnosis is at level 6. Let’s add a couple more assumptions: the rate of exposure to the toxin is constant and patients progress up the levels at a rate of 1 a year. This means that the average patient has symptoms of 7.5 (ranging from 6-9 since the 10’s die) and 1/4 of those diagnosed today will be dead in a year (all the 9’s die next year).
Now, imagine that a diagnostic test is developed that diagnoses the condition with absolute certainty. At first, the specialists who know the disease order it, but, fairly quickly (remember, it’s just a thought experiment), it becomes a standard test ordered when primary physicians see the early symptoms. This means that most patients are diagnosed at level 5. There are real benefits to this: patients with symptoms the the peace of mind of quick and accurate diagnosis without a string of misdiagnoses and uncertainty. However, there are also some strange statistical benefits: without any improvement in care, this simple change results in real improvements in status and outcomes. The average patient has symptoms of 7, not 7.5, and only 20% die over the next year, not 25%. We achieve this by simply better diagnosing people who complain of symptoms.
The people who developed the diagnostic test in our story drink their own Kool-Aid. They believe that early diagnosis really does make a difference, based on the statistics, and lobby the AMA to recommend screening for this disease. The AMA reviews the numbers and decides to recommend that people be screened every 5 years. Let’s assume the test has perfect sensitivity: if you have been exposed to the toxin, the test will find it. In screening, the test will identify 1’s, 2’s , 3’s, and 4’s as well as symptomatic patients. So now we’re diagnosing everyone with symptoms and 80% of people before their symptoms express. Despite not changing the course of the disease at all, some simple math shows that the average patient now has symptoms of 6.4, and only 17% of those diagnosed die within a year.
As a final twist, let’s imagine that the test has perfect sensitivity (no false negatives) but imperfect specificity (some false positives). Then we get patients who are diagnosed with the disease but who don’t have it. Their disease doesn’t progress (because they don’t have it). Of course, they start taking any medications for the disease and increase the frequency of their clinic visits, increasing their utilization. These false-positives might also drive up the averages.
This is just a mathematical exercise, I don’t have any particular disease or screening test in mind, but I think that you can see that we need to be very cautious about claims that early detection saves lives or improves outcomes, because in fact, early detection does save lives and improve outcomes without necessarily providing any benefit. When you think about it, there are major cost drivers in this equation: a new (and likely expensive) test, therapy starting earlier, and false positives. Also, don’t forget that tests can have side-effects, too.
Finally, I think most of us have noted how, over the years, more and more people are taking more and more medications. I think a part of this is that we have defined as sick people who previously would never have been. A cynic might question how much of this is because these people actually need medical care and how much is because early diagnosis improves outcomes.