The Pollyanna Phenomenon and Non-Inferiority: How Our Experience (and Research) Can Lead to Poor Treatment Choices
Pollyanna, a popular children’s book written in 1913 by Eleanor H. Porter, introduced the world to one of the most optimistic fictional characters ever created. She always saw the good in people and her approach to life frequently involved playing “The Glad Game”, where she attempted to find something to appreciate in every situation no matter how unfortunate. She was glad about receiving crutches rather than a doll one Christmas because it was great that she didn’t actually need them. She teaches this philosophy to those around her, even her cantankerous Aunt Polly, and the entire town is transformed into a veritable Mayberry, USA. Later, when she actually does require the use of crutches, her resolve is tested but she triumphantly finds a silver lining.
The Pollyanna principle, first described by Matlin and Stang in 1978 and also known as positivity bias, is a psychological tendency for people to place greater importance on, and assume better accuracy of, descriptive statements about them that are positive. This goes on behind the scenes while our conscious brain tends to dwell on what is perceived as negative stimuli. Though many folks do come across as pessimistic, we are subconsciously biased to accept praise and reject criticism. Anyone who isn’t clinically depressed is on some level more like Pollyanna than Eeyore.
This positivity bias also plays a large role in how we remember past events. As has been covered extensively in prior posts here on SBM, and on Dr. Novella’s excellent Neurologica blog, memory isn’t a replayed video or audio recording of prior events and our interpretations of them, but rather is a reconstruction that is prone to errors during processing and editing that accumulate over time. This leads to false memories that feel no less real than our recollection of what happened five minutes ago.
In this case, the Pollyanna principle results in positive information being more accurately processed and recalled than negative experiences. It also causes our memory of negative events to gradually become less negative as the years go by. I couldn’t have done that terribly during my first high school trumpet solo because I remember people telling me it was pretty good afterwards, right?
So what does this have to do with the practice of medicine? Biases that affect memory also impact how physicians and patients interact. I once assumed the overnight care of a child who had undergone a lumbar puncture performed by one of my female colleagues earlier that day. I ordered no tests and performed no procedures during my brief exposure to the family—yet over a year later when I admitted the same child for a completely different reason I was accused of being the terrible doctor who had unnecessarily subjected their baby to a spinal tap during the last hospitalization. Even after I showed them the documentation which proved that I had nothing to do with that (very appropriate) decision, and that I did not put a needle in their child’s spine, they refused to accept the evidence and had great difficulty trusting my diagnosis and recommendations.
There is a phenomenon in medicine, again named after the overly optimistic heroine Pollyanna, that augments the negative impact of the fallibility of human memory on medical practice with the bias towards an overly optimistic interpretation of some types of medical research and of our own clinical experience. This Pollyanna phenomenon can lead physicians to make poor choices, especially regarding which antibiotic is the best choice for a particular infection. And this is particularly common in pediatric practice, in my opinion.
The concept of a Pollyanna phenomenon in medicine has primarily been used when critiquing the approach of using increasingly broad antibiotics for the treatment of pediatric otitis media—ear infection—but it applies to a number of other bacterial infections frequently diagnosed in children. These include community acquired pneumonia, strep throat and sinusitis, three common reasons children are prescribed an antibiotic. But ear infections stand out as a clear front runner in this regard, so that is what I’ll focus on.
Ear infections are extremely common in children, again being by far the most common reason that a young child will receive an antibiotic. And even though the total number of visits for ear infections, and subsequently the total number of prescriptions for antibiotics to treat these infections, has decreased significantly over the past 20 years, still more than 80% of children will be diagnosed with one by their fifth birthday. The reasons for the decrease are myriad and include insurance issues that keep kids out of clinics, the success of the pneumococcal (PCV7 and now PCV13) and influenza vaccines, and public education campaigns aimed at teaching parents about viral infections. It’s also possible, though I have my doubts, that doctors have helped a bit by better-acknowledging other reasons for there to be fluid behind the ear drum than an acute bacterial infection.
But it isn’t all good news. According to a 2010 study in Pediatrics, there has been essentially no change at all in the percentage of visits billed for otitis media that end in an antibiotic prescription. Today, 76% percent of children diagnosed with an ear infection get an antibiotic compared to 80% in the mid 90’s. This despite the significant clinical experience of most other developed countries, reams of research papers and the 2004 publication of joint AAP/AAFP clinical practice guidelines which strongly recommended that we simply observe children over 6 months who have uncomplicated ear infections. There should have been a huge drop in the percentage of kids with ear infections receiving antibiotics but we’ve pretty much just ignored the recommendations.
The 2004 recommendations also included helpful guidelines on how to actually diagnose ear infections, particularly focusing on criteria stating that there should be more than just fluid behind the ear drum. They also required evidence of inflammation, such as redness and pain, and symptoms that are acute in onset. But there was too much wiggle room in the guidelines. They could be interpreted in such a way that children with another common condition, serous otitis media, could be diagnosed with an ear infection when presenting with fluid behind the ear drum and complaints of pain but no evidence of inflammation.
But even with this wiggle room, at the very least the days of the “asymptomatic ear infection”, an entity which both of my kids were diagnosed with at well child check-ups during infancy (both after 2004), but for which I refused treatment, should have been numbered. A new set of guidelines was published in 2013 which tightened up the criteria for diagnosis, requiring direct visualization of significant bulging of the ear drum from inflammation and purulent fluid. It will be interesting to see if we respond with a change in practice.
The 2004 guidelines also spelled out clearly what the most appropriate antibiotics for use in treating ear infections were. They further covered in what order antibiotics should be prescribed should an ear infection prove refractory to amoxicillin, the first line choice for the vast majority of cases. This was an effort to avoid overuse of unnecessarily broad agents and the risk of antibiotic resistance in the community.
Still more bad news, again coming from the widely publicized 2010 study in Pediatrics. Not only does it appear that we continued to prescribe antibiotics for any kid within arm’s reach that had fluid behind their ears, we’ve become increasingly careless in our antibiotic choices with very broad oral antibiotic agents, namely the third generation cephalosporin cefdinir (Omnicef) becoming increasingly popular. Use of cefdinir increased from 7% to 14% of all prescriptions for otitis media, frequently being used in place of amoxicillin and amoxicillin-clavulanic acid, the antibiotic recommended for severe infections or as second line for an amoxicillin treatment failure. Two silver linings did come out of the data however. Use of the recommended first line agent, the older, cheaper, more narrow spectrum and actually more effective drug amoxicillin did increase, and there was a 71% increase in the appropriate use of pain medications.
So why would a physician choose to prescribe an antibiotic, in this case a more expensive drug that has less efficacy killing the bacteria you want dead and actually increases the risk of future antimicrobial resistance in other types of infections? Well, there are a lot of reasons. Cefdinir may be more expensive, but it tastes better, can be dosed less frequently, and carries less risk of causing antibiotic associated diarrhea than amoxicillin-clavulanic acid. But in the case of ear infections, the downside of using it or a similar oral antibiotic in that class named cefixime (Suprax) outweighs the benefits.
There are more reasons, some decidedly more cynical. Drug company marketing, pressure from reps and the ease of use when there are readily available samples in the office all likely share some of the burden of blame. Still the most likely reason, if I had to hazard a guess, is more along the lines of something Pollyanna would come up with. Many physicians simply think that these antibiotics are as, or even more, effective than amoxicillin, and they really just want their patient to feel better. They want their patients’ caregivers to get sleep and be able to return to work. And they want return business too. Well maybe Pollyanna wouldn’t have come up with that last one.
But believing that an antibiotic is a better choice when it clearly isn’t is not an acceptable excuse. Are the doctors, nurse practitioners and physician assistants who write for cefdinir all incompetent? I have no doubt that some are but maybe there is reason to give them the benefit of the doubt. The deck is stacked against us a bit when it comes to these decisions after all. As I’ll get to, sometimes our own literature gives a false impression of efficacy.
Often there is a false perception that because an antibiotic is broader in its spectrum, meaning that it treats a wider variety of pathogenic bacteria, that it is better or “more powerful” than older and more narrow spectrum antibiotics. But as Mark Crislip once wrote,
the only thing in medicine that has 100% sensitivity and specificity is when a health care provider uses the terms “big-gun”, “strong” or “powerful” in relationship to antibiotics they know nothing, zip, zero, zilch about the treatment of infectious diseases.
Now I perhaps wouldn’t go so far as to say that they have no understanding of treating infectious disease. At least they are prescribing an antibiotic and not, say, a homeopathic remedy. I’ve just put The Glad Game to good use and I must say that I do feel a little less grouchy. But Dr. Crislip is absolutely right in that it’s not a good sign.
An antibiotic that is best at killing the bacteria causing the infection, and I mean really gets in there and makes them wish that their mommy-daddy had never undergone binary fission, may be extremely narrow spectrum. A staph-specific penicillin such as oxacillin is a much better choice for a sensitive staph infection, even a life threatening one, than vancomycin. It just doesn’t matter if a drug can kill 50 other types of bacteria if it can’t kill the one eroding into your mastoid process.
But sometimes it isn’t so clear that one antibiotic is better or worse than another. Both our clinical experience and the results of our research can be misleading. And so, finally, we’ve come full circle and are back to the Pollyanna phenomenon.
In 1992, a paper was published in The Journal of Pediatrics that looked at “the comparative efficacy of antibacterial agents for acute otitis media.” The authors looked at randomized, double-blinded trials of antibiotic therapy for ear infections that used both clinical and bacteriologic outcomes. Clinical success was determined by the patient’s signs and symptoms of illness resolving and bacteriologic success by the use of tympanocentesis. This involves using a needle to retrieve, in a sterile fashion, fluid from behind the ear drum that can be sent to a lab for culture of pathogenic bacteria if present. In many of the studies, tympanocentesis was performed both before and during treatment, which allowed for measuring the rates of sterilization of the fluid.
What they found is the Pollyanna phenomenon, where even the worst choices don’t seem that bad after all. For antibiotics that cleared bacteria out of the inner ear 100% of the time, the patient’s symptoms resolved 93% of the time when it was known to be caused by a bacterial infection based on tympanocentesis and 89% of the time when the ear infection was diagnosed only clinically. So roughly 1 out of 10 kids being treated with a drug that wipes the floor with the potentially offending bacterial pathogen won’t get better as expected in the real world, where tympanocentesis is rarely performed outside of an ENT office. And when a drug was used that had virtually no bacteriologic efficacy, the patient’s symptoms resolved in 71% of proven bacterial ear infections and in 74% of those diagnosed clinically.
In a nutshell, when we measure how well an antibiotic works for otitis media based on what is essentially subjective clinical improvement, drugs that kill the bacteria seem to be less effective than they really are while drugs that are one step above placebo appear to be work pretty well. This is a pattern that will always be seen when the condition being treated is largely self-limited, as ear infections have long been known to be, and may be actually caused by a viral infection which won’t respond to any antibiotic. This is why so many other countries stopped treating most ear infections with antibiotics long before our recent attempts to change, and it is why chiropractors think that their spinal adjustments work.
So it should be easy to see how an individual physician might incorrectly believe that the antibiotic they have chosen has worked, and why they might continue to use it time and time again despite it being a poor means of actually doing what they intend it to do. But there is more to the Pollyanna phenomenon. It also amounts to what is essentially a methodological “Glad Game” when designing studies that test the efficacy of a new antibiotic.
When attempting to see if a new antibiotic works, it can be challenging to design a trial that fits the mold of the randomized controlled trial that we all know and love. Ethically there can’t always be a placebo group that would leave subjects with a diagnosed bacterial infection untreated. In pediatric research, historically most antibiotic studies have not had a placebo control group and were designed as non-inferiority trials. This means exactly what it sounds like. Their goal is to demonstrate that the new drug is not unacceptably worse than the standard of care treatment, which may have also been based on non-inferiority trials. But non-inferior does not necessarily mean equivalent.
So in the case of cefdinir (#1, #2), comparisons to established treatments for ear infections have largely shown it to be comparable in effectiveness. But, as you can likely imagine, its efficacy is actually unclear because of the spontaneously-resolving nature of otitis media. When pharmacokinetics and pharmacodynamics are also looked at, true effectiveness can be triangulated.
In general, oral cephalosporins like cefdinir are not absorbed well, are easily rendered inactive by sticking to proteins in the blood, and don’t last long in the body. This leads to the level of drug available to kill the offending bacteria dropping below what is needed faster than other choices, which is a set up for resistance issues. For this and several other reasons, such as poor postantibiotic effect and tissue accumulation, cefdinir and cefixime are never better than amoxicillin or amoxicillin-clavulanic acid for killing susceptible bacteria. And the bacteria that cause the overwhelming majority of ear infections (and pneumonia) almost always are.
Non-inferiority trials are not all bad. They are necessary in many types of infections, particularly when they are not self-limited or frequently blamed when the actual culprit is a virus. It would be horribly unethical to compare the use of a new antibiotic for MRSA meningitis to placebo, and naturally a study such as that would never be approved. And if two antibiotics truly were determined to be equivalent in effectiveness, factors such as palatability and ease of dosing can be very meaningful as they may improve adherence to treatment recommendations.
But in the case of infections such as otitis media, community acquired pneumonia in young children, and acute sinusitis, the high rate of spontaneous resolution raises legitimate questions about the value of non-inferiority trial findings. Of note, in 2010 the FDA actually published nonbinding recommendations addressing this very subject. They called for increased use of superiority designs in these conditions and for better justification of the non-inferiority margins and efficacy endpoints used in trials involving infections that are not self-limited.
Pollyanna saw the best in every person, and found something positive in every situation. This is an admirable outlook on life, but one that doesn’t necessarily always translate well to the practice of medicine, especially when it comes to choosing the right antibiotic. We are all susceptible to the same biases and errors in logic, and not surprisingly there are similar issues in medical practice and research.
The Pollyanna phenomenon, an artifact of how we assess the success of our recommendations and study antibiotic effectiveness in largely self-limited bacterial infections, has likely led many physicians to make poor treatment choices. There is no better example of this than in the treatment of pediatric acute otitis media, where we have exposed many children to antibiotics that may only seem to be working but definitely increase the risk of developing resistance in the individual and the community. An understanding of pitfalls such as this, and awareness of the science behind expert treatment recommendations, can improve the care we provide our patients.