Treating cancer is a battle against exponential growth of mutating cells, so even breakthrough drugs may offer only incremental increases in survival time before residual tumors rebound. But as we learn more about the fundamental mechanisms of cancer, we can target those processes more directly. When will people diagnosed with the most lethal cancers more often than not survive for years? I will examine progress in cancer treatment and speculate on the path toward managing intractable cancer types.

## Cancer’s complexity

Why are several-year survival rates the currency of progress in fighting cancer? It is easy to wonder, especially if you have a personal connection to cancer, why we keep seeing breakthroughs that result in a small bump in survival times rather than cures. In short, cancer cells are human cells and they evolve. It is easy to kill cancer cells, but difficult to kill them without destroying the patient’s body. And it is extremely difficult to kill every one of them before they evolve again.

Most mutations either have no effect or they kill the cell and therefore self-correct. Even when mutations self-perpetuate against all odds, the body has [many defenses against them](https://www.ncbi.nlm.nih.gov/books/NBK9894/). The chances of a combination of mutations evading all this is vanishingly small, but there are trillions of opportunities for it to happen in the body. By the time it is diagnosed, a tumor has already bested a solid wall of defenses.

That means there is no easy fix for cancer, and a variety of treatments are used—primarily different types of surgery, radiotherapy, and chemotherapy. The relative importance of each of these will continue to evolve.

## Predicting survival rates

What would it mean for us to finally ‘control cancer’? As long as our DNA can mutate, there will be some uncontrolled cell growth. The best we can do is to identify it early enough that those cells can be destroyed or fixed before they cause problems. 

My predictions focus on the more intractable cancers to reflect our abilities relative to cancer more generally:

Note that five-year survival rate in this essay and question refers to [five-year](https://ncci.canceraustralia.gov.au/outcomes/relative-survival-rate/5-year-relative-survival) [*relative*](https://ncci.canceraustralia.gov.au/outcomes/relative-survival-rate/5-year-relative-survival) [survival rate](https://ncci.canceraustralia.gov.au/outcomes/relative-survival-rate/5-year-relative-survival), which is the fraction of patients alive five years after diagnosis normalized by the survival rate of similar people without a cancer diagnosis. That means a five-year relative survival rate of 100% indicates no lethal effect of cancer on that timescale, even though the patients might die of unrelated causes.

While the rate of pancreatic cancer incidence in the US has [increased slightly](https://seer.cancer.gov/statfacts/html/pancreas.html) over the last few decades, the survival times have increased dramatically. Five year survival was around 2% in the 1970s, and has been rising at an accelerating rate since then, crossing 10% in the 2010s. If the increase since 2000 were linearly extrapolated, five year survival would reach 50% by the year 2092. And if we look at cancer types that have longer survival times right now, can steady increases last for decades? They can, as five year survival rates have generally increased steadily since 1975 (for example [melanoma](https://seer.cancer.gov/statfacts/html/melan.html), [kidney and renal pelvis cancer](https://seer.cancer.gov/statfacts/html/kidrp.html), and [female breast cancer](https://seer.cancer.gov/statfacts/html/breast.html)), though sometimes they plateau ([colorectal cancer](https://seer.cancer.gov/statfacts/html/colorect.html)).

But I predict that the survival rate for the more lethal cancers will eventually increase faster than before, and five year survival for pancreatic cancer will reach 50% long before 2092. This is due to two factors and how they interact: 1) the trends in treatment strategy and 2) the nature of survival statistics. Early therapies were often effective only in a subset of cases, and this uncertainty was compounded with the probability that the cancer had spread too far in the body to be treatable at all. Over time, earlier detection and a clearer diagnosis of the particular mutations responsible will lead to a more predictable, and even routine, personalized treatment.

As a result, improvements in treatment will more universally benefit cancer patients rather than helping only those with one subtype. Breakthroughs that can push survival up a few months *for a high proportion of cases* can have a large effect on five year survival. This is especially true with early detection: An earlier diagnosis means more time until death, regardless of the ability to actually treat the cancer.

I would therefore start at the linear extrapolation of 2092 and bring the estimate down for each of several factors. The direct treatment-agnostic benefit of earlier detection, the improved treatment resulting from earlier detection, and the broadening of successful treatments thanks to personalization could each accelerate the five year survival increase by 40%. These adjustments result in an estimate of 2047, which seems plausible. But, I'm not confident in the magnitude of each of these adjustments, and so I would still include a 25% to 75% probability range of 2038-2062 corresponding to +/-20 percentage points from those 40% adjustments.

## **Predicting cancer incidence**

There are different metrics of progress in fighting cancer, and for some metrics it is not always clear which direction is good or bad. Survival rates for a cancer type can increase while the [incidence of new cases also increases](https://seer.cancer.gov/statfacts/html/melan.html), and the latter could happen for reasons as benign as an increase in detection. And with complete cancer cures as elusive as they are, the metric I find most difficult to speculate on long-term is, what will the trend look like for prevalence, or number of people living with cancer? This, however, seems too difficult to measure, since patients whose cancer was treated effectively generally won’t know in the following years whether it is truly gone.

So as a more testable forecast question: *What will the rate of new pancreatic cancer cases be in the US in the year 2038?*

While the definition of a cancer diagnosis may be fuzzy in the case of early detection, I predict 15.3 per 100,000 people, or a modest increase to 115% of the 2018 rate. The trends in pancreatic cancer risk factors such as [obesity](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6685923/) could inform us on this going forward.