My wife and I had Shabbat dinner Friday evening at the home of our colleagues Susan and Judah. After the wonderful singing and food the conversation drifted to work. At one point Judah mused on just how he liked to teach medical learners. Specifically, he got them to focus on the underlying biological mechanisms behind the treatment decisions being made in clinic. He gave an example of one time where he expanded on prednisone and how it counters specific inflammatory cytokines that their Lupus patients had precipitating inflammation in their various bodily organs.
Ever the disagreeable I countered that this was actually the opposite of what we at McMaster usually preach. As evidenced-based practitioners we were supposed to be nihilistic to the underlying mechanisms why something works and instead stick to making our clinical decisions based only on what has been proven empirically by randomized clinical trial or at least some acceptable level of evidence below that. Why would we waste time discussing how and why something worked rather than focusing on the studies done in humans that proved it actually did work? (Or in some cases didn’t work!) I mean, for every biologically plausible explanation one could come up with someone else could just make up a counter biological argument. Two people could look at the same basic science literature and select the studies that better supported their own but differing points of view and find themselves at an impasse.
But Judah was adamant that discussing biological mechanism was vital to his practice, and both his wife Susan and my wife Randi wholeheartedly agreed. I seemed to be alone on my ‘Island of empiricism’ and there were no rescue ships on the horizon. Judah was adamant that medical learners and patients themselves wanted to know why we were asking them to take these treatments. To just say it was because a clinical trial had a p-value < 0.05 or an absolute risk reduction of 2.7% was insufficient for them. People wanted to know what these drugs were doing inside their body. While I thought that was fair, I countered that many times we didn’t know exactly what these medications were doing inside their bodies and what we were saying they were doing was more than just a little made up. After more than thirty years working as a doctor and a scientist I was aware of many instances where such biologic mechanistic explanations were used to promote therapies that had actually been empirically proven to be ineffective. My recent work during Covid had provided several such examples. As I was caring for many people dying from COVID on our hospital wards we were at the same time enrolling them in RCTs on a revolving platform of many therapies ranging from anti-malarials and anti-fungals to anti-inflammatories. Ironically the thing that worked best was prednisone and drugs like it. How? I don’t know. All I knew was if you had COVID pneumonia and your oxygen levels were dropping when I gave you a really high doses of glucocorticoid likes prednisone or dexamethasone you were much less likely to die for whatever biological reason. The less dying was good enough reason for me.
But Judah was right to some degree. Humans by their nature are curious and more than a little apophenic; driven to overfit data and see patterns that aren’t really there. We have this addiction to causation and an even worse aversion to uncertainty, both to a fault. To better make my point I started to discuss my own research career working with Hertzel for 20 years trying to prevent heart disease in people with diabetes. One problem we faced was an impasse with the people that had run the UKPDS. Their studies were too underpowered to show heart benefits, but they followed their patients for 10 more years and this showed that their blood sugar lowering medications eventually did prevent heart attacks and cardiovascular death. But all their work was in sharp contrast to our ACCORD study that had shown no heart protective effects at all for glucose lowering drugs. Indeed, we had to stop our study early as the intervention had been found to actually cause harm (specifically increased death) and so a data and safety monitoring committee (thank God for them picking up on that early) shut us down. So how does one reconcile the conflicting evidence between the heart protective effects seen in the old UKPDS study patients vs. the patients seen in modern studies like our ACCORD trial and others? The consensus at the time was that glucose lowering must have some kind of ‘vascular legacy effect’ that was protective even years later, but you had to give these therapies early, before atherosclerosis had set in and already did its damage to the blood vessels. That was the proposed biological mechanism to explain the contradiction.
I distinctly remember the fall of 2015, sitting in a chemo suite getting an infusion of R-Benda while reading my paper copy of the NEJM and the Empa-REG trial within. I nearly fell out of my seat. For nearly two decades we’d been banging our heads against the wall failing to stop heart disease in diabetes and this SGLT2 inhibitor trial was doing just that in spades. From a biological mechanism perspective I had always been skeptical of the SGLT2i class. Their mechanism of action was to induce urinary loss of sodium and glucose. For years I had taught medical learners that such osmotic diuresis was the biological mechanism behind diabetic nephropathy as well as a physiologically steppingstone to glycemic emergencies like hyperosmolar non-ketotic coma (HONK). How could SGLT2i be a good thing then? All the biological mechanisms would predict the opposite. Well, many studies later using many other SGLT2i therapies there’s a clear answer that drugs in this class not only reduce heart failure but also lower kidney failure and mortality. Then came along the GLP1 drugs, a class in which a quarter century earlier I had studied their biological mechanisms in mice and cells during my time working in a molecular biology lab. Similar to SLGT2i the GLP1 studies showed equally impressive vascular benefits, mortality reduction and kidney protective effects. Not to mention dramatic weight loss benefits that even Hollywood movie stars were clamoring for.
So I was wrong about the biological mechanism of SGLT2i being bad. And the consensus biological mechanistic explanation behind the ‘legacy effect’ was clearly wrong too. All these SGLT2i and GLP1 trials had been done in people with already well established atherosclerosis. You didn’t have to start glucose lowering drugs early before atherosclerotic damage had occurred. You just had to use the right glucose lowering drugs that offered true atherosclerosis protection. The so-called legacy effect was wrong and the long-term data from UKPDS, which was clearly not the primary outcome measure from the original trial, was simply not true. A statistical play of chance if I can use the UKPDS author’s own vernacular. An important reminder that secondary analyses should always be limited to hypothesis generation and that luck is always a factor both in science and in life.
“I find it hard to believe Will that you truly think so little of biological mechanisms” said Susan. “You worked in a lab. Surely you comprehend the value of understanding how things work at a biological level. These new drugs aren’t conjured out thin air. The ideas behind them come from basic science and their creation and design if anything are becoming even more based on our understanding of biology.”
“Exactly right Susan” said Judah. “Just look at the advances in biologics and check-point inhibitor therapy and how they’re revolutionized my field of rheumatology as well as neurology and oncology. Have there been no examples in your own career where the understanding behind biology mattered most?”
I admitted there was one instance in my career when biological mechanism had actually carried more weight than empirical evidence. A representative from the Pfizer company had approached me about their new experimental drug torcetrapib that raised blood levels of healthy HDL cholesterol. Pfizer was banking on this “CETP inhibitor” being their next blockbuster drug to replace Lipitor, a drug that mainly lowers bad LDL cholesterol but was soon going off patent meaning a loss of millions in future revenue sales for Pfizer. They wanted my thoughts on torcetrapib and perhaps considering being a local PI if they were to start a Canadian study site. I explained to them that I didn’t think their CETPi would work because it’s mechanism of action was to inhibit HDL particles from dropping cholesterol off at the kidney for renal disposal. In this way, though it raised HDL cholesterol levels, it actually impeded HDL from doing its job of taking cholesterol out of atherosclerotic plaques in the blood vessel wall and disposing off it at the kidney. I said it was like closing the dump but allowing trucks to still pick up the garbage. You’d have a lot of garbage trucks driving around town and eventually they wouldn’t be able to pick up anymore garbage from people’s homes as they’d be too full already. In population cohort studies high HDL levels correlate with less heart disease because having more HDL particles means greater removal of cholesterol from blood vessel plaques. But for people on CETPi the higher HDL levels reflect instead HDL dysfunction and It was my opinion that the drug wouldn’t work and would likely cause more heart disease not reduce it. But Pfizer was dead set on continuing their clinical trials. They didn’t care about the biological mechanism concerns I was raising as they had already invested hundreds of millions on this drug and the Lipitor patent expiration was hanging over their head like the sword of Damocles. They hoped their empirical clinical trials would prove my biological mechanism argument wrong. But that’s not what happened. Pfizer’s long track record of blockbuster drug discovery stretching back to 1990s came to a screeching halt and they had to reverse course and diversify into more boring low margin OTC business like their competitors JNJ did to help them get through the lean years when drug discovery attempts went more bust than blockbuster. A couple of years later Merck came by to talk to me about their own CETPi anacetrpib. Unlike Pfizer’s torcetrapib their drug didn’t cause hypertension. Merck believed it was the hypertensive side effect which was what had doomed torcetrapib and so their anacetrapib would be different. “Nonsense” I told them and said it was the CETPi mechanism on HDL that was flawed not any blood pressure effects. And the Merck trial failed just as bad as the Pfizer one did, wasting hundreds of millions of dollars and maybe a more than a few human lives chasing a dream of raising HDL that turned into a nightmare of doing it in exactly the wrong way.
“So it’s both” said Judah. “Both biological mechanism and clinical trial empiricism are necessary then. Equally important.”
“I guess” I said acquiescing, “it depends on the situation I guess.”
“Well, the dessert situation is berries and ice cream.” said Susan.
“My favorite!” agreed Randi
We all laughed. Susan was wise to cut things off there. Judah and I both have a tendency to go on endlessly given the chance. And Susan has a long track record of such wisdom during her storied career at McMaster. I guess that’s how you get to stick around for so long. I’ll have to remember that I thought to myself but now It was time for dessert. Susan got that right too.
Komentáře