Friday Reflection 43: The Absence of Reassuring Counterfactuals in Clinical Medicine

KN is a 90-year-old man I visit at an inpatient hospice. His son warned me that my visit might “be a waste of time” because he sleeps much of the day and “can’t even get out of bed.” When I arrive, however, I find him in good spirits. He is pain free, cognitively intact, and genuinely pleased to see me. He is not only happy for the company but touched that I made the effort to come. He is cachectic and his skin is pale and waxy. He looks much worse than he did two years earlier when we made the decision not to treat his cancer.

On the L train home, I think that this seems like a pretty good way to approach the end of life. I also worry that we made the wrong decision in not treating his cancer.

The first time I heard the term “counterfactual” was in a meaningless argument I was having on twitter. This was five or so years ago – when I still had meaningless arguments on twitter (and when twitter was still called twitter). I made a definitive statement, and my sparring partner responded, “you can’t say that, you don’t know the counterfactual.”

He was right, and I was impressed with this new term he’d invented in the heat of a social media argument. When I looked it up, I found that “counterfactual” was coined by Nelson Goodman in 1947. After learning that, I just felt bad that I didn’t take more philosophy in college.

The way we now use the term is somewhat different than Goodman intended. Most of the time, people are simply making the point that we know only how things turned out given the decisions we made. We don’t know how things would have turned out had we made different decisions.

When the medical outcomes of my patients are bad, I am sometimes haunted by the lack of a counterfactual.

I diagnosed KN with cancer when he was 88. Given his symptoms, making the diagnosis was not difficult, the decision making around his management was. When we first met to discuss the diagnosis, treatment for cure was possible, but would have entailed further evaluation, major surgery, and chemotherapy. It was likely that he would also require treatment of another, albeit minimally symptomatic, disease to allow chemotherapy.

We were in the fortunate position of having known each other for decades. I also knew his family well, most of them were my patients. I knew how he usually approached medical decisions and his thoughts about end-of-life care. In the end he chose, with my strong guidance, to forgo therapy. He was 88 years old, had other serious medical issues, and his cancer was not causing problems. Choosing not to begin therapy, leaving an opening for future palliative care, seemed like the right choice.

Two years later, I am not so sure. Sitting with him in hospice I wonder whether, if we had treated him, he’d be home, cancer free, with years left to live. Of course, I don’t know the counterfactual. Maybe therapy would have ruined or even killed him. The lack of a counterfactual leaves me only with second guesses and no possibility of an answer.

So much of medicine is uncertain. Any difficult decision I make with a patient is more art (or maybe a gamble) than science. Studies don’t exist for the scenario at hand, usually because the situation is unique. The infinite variation of our biology and histories guarantees this. (Maybe this is a reason my AI dystopia will never be realized.)

Even when literature clearly supports a decision, and things turn out badly, the lack of a counterfactual allows endless second guessing. A few cases:

• A patient with established cardiovascular disease is started on a statin. Withholding the drug could be considered malpractice. She develops statin-induced necrotizing autoimmune myopathy and suffers through years of pain, weakness, and immunosuppressive treatment. Had I not treated her, would she have had a cardiovascular complication?

• A 72-year-old patient, having already received four COVID vaccines, calls with 24 hours of upper respiratory symptoms. A home COVID test is positive. Given recent literature, I opt not to treat him. Four days later he is admitted with dehydration and delirium. After discharge, he spends two weeks in a subacute rehab facility. Three months later, he is still not back to his premorbid level of functioning. Would he have done just as badly had I started Paxlovid?

• I see a 52-year-old woman as a new patient. I suggest she have a screening colonoscopy. It reveals a colon cancer. Liver metastases are found at the time of diagnosis. After eight years of intensive treatment, she succumbs to her disease. Would she have lived if she had heeded her previous doctor’s advice and had a colonoscopy at 50? Would the end of her life been better had I not encouraged her have that colonoscopy. (She reminded me, late in her life, that I had told her that the worst we would find on the scope would be a polyp.)

When things go well in medicine, we never second guess our decision making. If we reflect on the absence of an outcome at all, we are likely to congratulate ourselves for having prevented it. “I convinced this man to stop smoking when he was 50 and now, here he is at 70, alive and well!” When things go poorly many of us are apt to beat ourselves up. There is never a counterfactual to let us off the hook.

Photo is of Nelson Goodman.