Moneyball is a movie about the 2002 Oakland Athletics, the year after they lost to the New York Yankees in the 2001 American league playoffs. In 2001, the Yanks had a $114-million-dollar payroll and the “A’s” had a $39-million-dollar team. Before the 2002 season, the power of A’s team is gutted as they loss their 3 star players to free agency. Not surprisingly, each left for mega-dollar contracts, leaving the A’s (a small market team with a cheap owner) to struggle with very little money, a weak team and numerous positions to fill.
The movie is a David versus Goliath tale with the A’s (David) struggling to compete with Goliath teams like the Yankees. In planning for the next season, the teams General Manager (Brad Pitt) meets with a room full to the team’s top scouts. They cuss and discuss, evaluate the available players, use their collective 200 years of major league experience and prognosticate on future player choices. Their crusty, seasoned scout advice is based on experience, knowledge and "gut" feelings.
However, the general manager reminds them it’s not business as usual. They can’t afford what they want and their recommendations have no real certainty. His mantra to the scouts is, “Adapt or die!”
The movie pits hunch-driven “expertise” (convention) against a mathematic approach to decision making (Moneyball). The scouts protest saying you don’t put a team together with a computer. “Google boy can’t do what we do; he and his computer - they don’t know what we know, don’t have our experience, intuition and devotion to the game.”
The rest of the movie proves the scouts to be wrong as the geeky number-crunchers and “big data” analyses replace high price talent with cheaper players that have a statistical likelihood of producing the same success but at a much lower cost. The A’s struggle at first, but go on to win, set records and go to the playoffs. Are the radical changes proposed in Moneyball something current day rheumatology should embrace?
Would you trust a big Whopper computer printout of next best drug(s) to give to Mrs. Hawking who has psoriatic arthritis and needs to start a DMARD? (Hey remember we are in the movies, suspending disbelief and not considering real world issues of money, insurance, manage care mismanagement, etc.).
I believe most of you would huff and scoff at a formulaic or number driven approach. Why take the highly complex task of choosing a drug, (negating the expertise of a seasoned rheumatologist) and trust a computer-crunching geek monkey? Foolish, unproven and destined to fail the sick patient, say you.
But, with all your expertise, knowledge of the patient and the research results, you still have no certainty that methotrexate is preferred over adalimumab or secukinumab in Mrs. Hawkins. Odds yes, but certainty, no.
Sure, given a neverending cavalcade of uncontrolled Mrs. Hawkins, your recommendations will work well for the majority. Everyone expects the 60-40-20 percent ACR (20/50/70) responses seen when starting a new therapy. Yet, when I surveyed 495 US rheumatologists about their outcomes, they claim to achieve an ACR20 response in 82% of their patients and remission (akin to an ACR70) in 60%. Obviously these self-inflated recall numbers are nowhere near reality, suggesting we may skew our reality to fit our choices.
In the movie, the general manager says, “It’s not about buying players, it’s about buying wins; but really it’s about buying runs. Plus runs from batters and minus runs from pitchers and defenders”. Can you analogize this line to how you manage a rheumatoid arthritis patient?
I don’t wish to advocate too strongly for numbers driving decisions, but I’m having a hard time understanding why rheumatologists:
- Are slow or fail to change DMARDs in the face of moderate to high disease activity. Several studies show it takes 6-12 months for a new DMARD to be introduced despite repeated measures of high activity.
- There are 5 TNF inhibitors, an IL-1, IL-6, B cell and T cell blockers on market to treat RA. All have the same trial designs, the same ACR responses and relatively minor differences in safety, yet each manufacturer want us to use their drug next? Pharma goes to great lengths to influence your behavior without providing any research or evidence to ensure that the treatment decision doesn’t favor, but instead almost ensures patient success.
- Our best biomarkers are ESR, CRP and CCP. Your fathers’ rheumatologist relied on the ESR, CRP, and SCAT. Now that’s what I call going nowhere slow. Expensive wannabees, like MRI and Vectra, don’t make you smarter; they make you feel good about yourself.
- We infrequently grapple with uncertainty in the 15% of (RA) patients you see who have failed most of your “best” therapies and your left to choose the next best. At this point, therapeutic choices are an unnerving combination of experience, guesswork and faith, and less about the science (evidence). Must there be diminishing odds of success with repeated failures?
- RA survival rates haven’t changed substantially, even though our therapies have changed dramatically.
Baseball and rheumatology can be "an unfair game" to those who play. The question is, how are you going to compete against ill-defined odds? What forces will you bring to bear to optimize the fate of your patients? How can we become "card counters at the Vegas black jack table"?
Many will point out that despite the success of the A’s in 2002, “Moneyball” as a management style never caught on or was reproduced in major league baseball. Yet, it did revolutionize the game. Nearly all MLB teams have a team of analytic, computer-crunching geek monkeys to shape the personnel decisions and day to day operations in baseball. Just as analytics are a driver for modern baseball, so shall it become an important tool for the practicing physician?
I grew up knowing all the “stats” for my favorite players and team – AB, BOB, AVG, ERA, etc. But in the last decade the stats collected, reported, bought and sold have become too complex for most fans to follow. You have to leave those to the experts.
I would propose we need stats, evidence and new numbers by which we can make more intelligent, timely and scenario-specific decisions. This may come with biomarker development and personalized medicine – but when will be? Or will it take “big data” to guide us with another level of certainty.
Although confident, I still have some uncertainty about my next DMARD or biologic choice. But I am very certain about the following:
- I have the metrics (CDAI, GAS, etc.) and a target goal needed to know when to change therapy.
- A new drug deserves a therapeutic trial, but only for 6-8 weeks and then it’s time for a “go-no go” decision on treatment success. If the drug takes longer than 8 weeks to succeed, it goes to the end of my list. If the patient is better at 6 or 8 weeks – that’s not success that’s a “foul ball” and an invitation to change my next best treatment.
- I will continue to look to available “big data” for the edge needed. This includes very large databases like CORRONA and analyses of commercial or claims data. Randomized clinical trials with done in 200-5000 patients are designed to prove efficacy. We are more likely to get insight on patient management issues, best practices and safety measures from tens or hundreds of thousands of patients followed prospectively and studied by analytic, computer-crunching, geek monkeys (God Bless them).
Therapeutic decision making - it just sounds good and smart. Mix a bunch of clinical trial results with safety stats and an appreciation of comorbidities, in the context of patient preference, coached by prescriber experience and wahlah! You have an obvious treatment choice, or two or three.
Someone needs to break the model on therapeutic decision making. My experience in management is invaluable, but I would prefer to avoid experience and faith-based guesses and depend on better odds for my patients.
“It’s unbelievable how much you don’t know about the game you’ve been playing all your life”
– Mickey Mantle