As with most things, the answer is more complicated than it may at first seem.

Up to now the official answer was simple – always finish prescribed courses of antibiotics. Stopping early, it was argued, could increase the risk of antibiotic resistance. The idea is that bacteria which are a little resistant to the antibiotic will survive longer, and if they are not wiped out by completing the course, the partially resistant bugs will then reproduce, with random mutations leading to greater resistance. The antibiotics are essentially a selective pressure leading to resistance.

This makes sense, and is almost certainly true with certain chronic infections, such as tuberculosis.

A new review, however, suggests that we cannot apply this rule simplistically to all infections. It depends on the type of bacteria involved.

Some bacteria exist in the body as part of our normal flora. Mucous membranes are colonized with bacteria, for example. If, however, those same friendly bacteria get into the wrong place, like the lungs or the blood, they can cause an infection. In such cases, the longer the course of antibiotics, the more likely bacteria living in the body will develop resistance. If they ever cause an infection, that infection is therefore more likely to be resistant to the antibiotics previously used.

So in essence there are two factors at cross purposes – completing a course of antibiotics to prevent partially resistant bacteria from surviving, and minimizing the course of antibiotics in order to minimize overall selective pressures on evolving resistance (for the target bacteria and other bacteria in the body).

We therefore need to balance these two concerns.

What the authors of the review argue is that the evidence for completing a course of antibiotics in order to reduce the risk of resistance is lacking. Since, they argue, that we do not know that completing a course is a risk, but we do know that longer antibiotic use is a risk for resistance, the recommendation to complete antibiotic courses in all cases is not optimal.

In fact, they argue that the concept of an “antibiotic course” is outdated and not evidence-based. An “antibiotic course” is essentially a predetermined length of time for which specific antibiotics should be taken to treat a specific infection. For many infections there is evidence that certain durations are important for effectiveness, but not necessarily to minimize resistance.

The alternative approach is to individualize antibiotic treatment duration. This could mean substituting another factor for duration, such as resolution of symptoms. In some situations it could also mean resolution of abnormal findings on laboratory tests or imaging studies. These other factors are already incorporated into evaluating patients with infections and monitoring their response to treatment. The question is – do we rely more on these symptoms and laboratory test factors, or adhere to predetermined lengths of treatment?

Where does this leave us?

What is the bottom line take-home message from this review? Well, the CDC has  already changed the wording of its recommendations to the public from “complete the course” to “take exactly as prescribed.” That is a more general recommendation that essentially means, just do what your doctor tells you to do.

That is good advice. Essentially what this review is telling us is that antibiotic treatment duration is complicated, and determining the optimal duration needs to be more individualized and account for more information. Every type of infection might have its own parameters – some may require a specific course duration, others may rely on specific symptoms or tests. The number of variables is huge.

What this review does not mean is that patients should stop all antibiotic courses when they feel better. That is also an overly simplistic algorithm that is likely to be suboptimal, and may result in a partially treated infection (a potentially serious outcome).

Instead, doctor’s recommendations will have to become more nuanced. We also need more research to better inform these decisions. This is also another area where expert systems could be highly useful – crunching all the variables to make optimal prescribing recommendations.