Shining a Light on Numbers Needed to Treat (NNT)

I see my cardiologist once a year about my high blood pressure and we also have conversations about the current state of healthcare, both the quality and cost. These conversations started when he asked me (knowing I am a psychologist) why antidepressants are always among one of the top four selling drugs.

We had discussions about this and they continued after I gave him a copy of my book What’s Wrong With Much of U.S. Healthcare and How to Fix It. He obviously read it because he had pertinent comments and questions about much of the material, particularly the great number of psychiatric and non-psychiatric FDA approved drugs that have only marginal clinical effectiveness.

For example, researchers were able to analyze the data for all randomized placebo-controlled depression drug clinical trials submitted to the FDA. They found that almost half (49%) of the submitted trials found no significant difference in improvement between those on the drug and those on inert placebo! As a psychologist trained in graduate school to conduct sound experiments and research, and with the expertise needed to critically examine the research of others, this one statistic is enough evidence that antidepressant drugs are not effective or justified in treating clinical depression.

In addition to allowing the elimination of mildly depressed patients from these trials, the FDA incorrectly allows introduction of a placebo washout period to remove additional subjects whose depression might improve on placebo alone! This policy is another violation of sound science, since in the real world the drug will be dispensed to a larger group of depressed patients than those in the biased clinical trial!

Despite this attempt to remove those depressed patients who would improve on inert placebo alone (about one-third of all depressed patients), Irving Kirsch found that 75% of the improvement was due to the placebo effect and only 25% was the drug effect! Again, another finding that seriously questions the advisability of using drugs to treat depression.

At one time, I believed that only FDA-approved psychiatric drugs had only marginal effectiveness and not the vast majority of non-psychiatric drugs for the treatment of “medical” problems. However, when I expanded my review to all health research, I found I had been wrong. I discovered in the literature that many popular drugs for “medical” problems also have only marginal clinical effectiveness.

My cardiologist and I discussed this, and he offered the opinion that physicians prescribe these drugs for fear of being charged with malpractice if they do not prescribe them since some patients do get better on many of these drugs! Let’s look at my cardiologist’s justification of this defense of prescribing FDA-approved drugs with very limited effectiveness. However, before doing so the reader needs some background information to understand a statistic among others that describes a treatment’s effectiveness.

The Number Needed to Treat (NNT)

There is an important statistic in healthcare called the Number Needed to Treat (NNT). This number is the number of patients with the health problem that must be treated in order to have an impact on one person/patient. NNT can vary from an NNT of 1 (one person treated, one person helped) to an NNT of 100 and more.

According to a Cochrane review of antidepressants as a treatment for depression, the NNT for SSRIs is 7, and 9 for tricyclics. That is, one would have to give seven depressed patients an SSRI to ensure that one of the seven would improve more than if all seven had been treated with inert placebos. In other words, one of the seven would benefit from the treatment and the other six would be needlessly exposed to the adverse side effects of the antidepressant.

With an NNT of seven, depressed patients need to make this decision: Is the one-in-seven chance of a benefit beyond placebo worth the risk associated with taking the drug?

Common side effects of antidepressants are nausea, weight gain, lack of sexual desire/impaired orgasm, fatigue, insomnia, dry mouth, blurred vision, and constipation. More serious and rarer side effects include suicidal thoughts, allergic reactions, mania, seizures, and tardive dyskinesia (TD). In figuring the cost of drug treatment of depression, one must also include the cost of drugs (and physician charges) to treat some of the adverse side effects of antidepressants.

Based on my review of the “medical” literature, it is not uncommon to find that for treatments of “medical” problems, the NNT is often 10 to 15 and even considerably higher. The higher the NNT, the worse the efficacy. What about statins to prevent heart attacks and strokes? Statins have an NNT of 60 for preventing a heart attack and an NNT of 268 for stroke! Those are the numbers of healthy people needed to take statins for five years for those projected outcomes to be achieved!

With NNTs like these, how does the medical profession continue to recommend statins? Or other drugs with similar NNTs?

I have never heard a physician recommending a statin say anything about side effects. For statins, the Mayo Clinic reports that serious side effects include muscle pain, liver damage, and some cases of memory loss or confusion. With such outcomes and possible side effects, I am glad I looked at the research on statins and stopped taking statins 10 years ago against my physician’s advice (with life style improvements, my cholesterol is now in the normal range).

Psychiatrists might say, shouldn’t antidepressants be prescribed even though only one in seven patients—about 14% of all patients—experience a reduction in symptoms beyond what would be achieved with placebo?

So let’s do the math. Imagine you prescribe antidepressants to 10,000 depressed patients. That would mean 1,400 patients would benefit. Prescribers must be concluding that this provides a good risk-benefit ratio for all 10,000 patients.

But isn’t it malpractice to expose 8,600 patients to the side effects of antidepressants without their experiencing any benefit in reduction of symptoms beyond placebo? And then, of course, many of these patients will be exposed to other drugs to treat the antidepressant’s side effects. I believe that is the greater malpractice!

Moreover, in the above case, the 14% are those who achieved a reduction in symptoms at the end of a six- to eight-week trial. In assessing the effectiveness of any health treatment, one must factor in that many health conditions such as the flu, upper respiratory infections, depression, and others are also self-limiting: They often get better with or without treatment, particularly over longer periods of time.

In many health problems, the NNT will vary considerably for individual patients depending on the severity of their illness, and whether they are at high or low risk for having the bad outcome. The provider should discuss all of these factors so that patients are informed about the benefit/risk ratio when offered a prescription by their provider.

What are some other measures of effectiveness for cardiovascular therapies?

Warfarin and aspirin are both blood thinners to present strokes. In atrial fibrillation, warfarin has an NNT of 80.  Fewer than two patients in this group will perform better than those on aspirin.

In studies of aspirin to prevent a first-time heart attack or stroke, no deaths were prevented.

In studies of coronary bypass surgery to prevent heart attacks, one death was prevented in every 25 treated patients; one in every 10 to 14 treated were helped (ie, the drug prevented a non-fatal heart attack).

In studies of coronary stenting for non-acute coronary disease, compared to drug treatment, there were no significant differences in outcome: 145 deaths in surgically-treated patients vs. 144 deaths in medication-treated patients.

I will leave it to the reader to decide if any of the above therapies are justified and which treatments should or should not be authorized for payment in any health plan. I heartily concur with the need for respecting patients’ rights in receiving health services. But unless they are informed of NNT risks and benefits, their right to make an informed choice is being denied.

One plan to reshape the current healthcare system

I don’t think we can rely on physicians to provide patients with NNT data so they can make informed decisions. This information will have to come from other sources.

One tool for improving healthcare in this way is to have the payer for the proposed treatment do this NNT assessment, and then decide whether the potential harms outweigh the potential benefits to the patient population as a whole. If the harms outweigh the benefit, then the payer can present other treatment options with a better NNT.

For instance, a payer could determine that an NNT of 7 for treating depressed patients with drugs does more harm than good, and thus promote alternative treatments. During my days managing behavioral health care for a management company, 65% the callers complained of depression and or anxiety symptoms. Yet, my firm never authorized drug treatment for these symptom disorders. The callers were instead referred to Ph.D. psychologists and clinical social workers and the average treatment took just under six sessions.

And what was the result for patients? Ninety percent or more expressed satisfaction with this care, versus a 76% satisfaction rate for the behavioral health industry as a whole.

My reading of the health literature leads me to believe that improving the quality of healthcare, and lowering its costs, will require major changes to the present system. Nothing short of revolutionary change is required to improve the current healthcare system.