Many JADA readers have commented on the sometimes complicated statistical analyses used in some of our research articles. It is unrealistic to expect all readers to comprehend every difficult statistical calculation—I certainly don’t—yet these analyses are necessary to validate and demonstrate the importance of a study. There are certain statistical analyses that all of us should strive to understand.

As oral health care providers, we are always looking for the best treatment modalities for our patients. Recent great strides in biomedical technology have made vaccines against periodontal diseases and caries, salivary tests for oral and other types of cancers, and many other treatments and screening modalities imminent realities. While many new technological advances sound wonderful, we need to appreciate their advantages and disadvantages before implementing changes in patient care. And we need to ask ourselves whether we always understand what we are being told by well-meaning sales representatives. Are we truly able to evaluate claims of benefit and superiority? In seeking answers to such questions, it may help to review a couple of examples.

Suppose a new periodontal vaccine has been introduced for sale to general dentists. The manufacturer is understandably enthusiastic about its success and actually reports that the use of this innovative vaccine can reduce the risk of developing periodontal disease by 40 percent. But what does this claim really mean? Does it mean that a patient using this vaccine will reduce his or her risk of developing periodontal disease by 40 percent, or does it mean that if we use this vaccine with all our patients we will be able to reduce periodontal disease by 40 percent within our patient population? How can we evaluate the true benefit of this vaccine?

Let’s presume that 2,000 volunteers participated in the clinical study that determined the efficacy of the vaccine. One-half of the volunteers were given the vaccine; the other one-half were given a placebo. If 50 volunteers of the 1,000 who took the placebo developed periodontal disease and 30 volunteers of the 1,000 who received the vaccine developed periodontal disease, the vaccine actually reduced the relative risk of periodontal disease by 40 percent. A 40 percent reduction of 50 equals 30.

Now let’s look at this result a bit differently—by absolute benefit, or absolute risk reduction. The vaccine benefited an additional 20 volunteers—50 less 30—for every 1,000 patients treated, or 2 percent. Thus, for every 1,000 patients in an oral health care professional’s patient population (if comparable to the sample used in the study) receiving the vaccine, 2 percent of patients will benefit. Two percent is quite different from the 40 percent reduction noted by the manufacturer.

From a clinical perspective, it may be even more worthwhile to know how many patients need to receive this vaccine before one patient benefits from it. This estimation often is referred to as "number needed to treat" (NNT). As an additional 20 patients of 1,000 would benefit (the absolute benefit), it would require the treatment of 50 patients for one single patient to benefit from this vaccine. Thus, 49 patients would have to receive the vaccine before one patient would benefit.

Which of the numbers above is the most important? That depends on a number of different variables. If the side effects of this vaccine were substantial—or even minimal—the oral health care professional would need to make a decision about the importance of the benefit to one patient versus the potential side effects to 49 others. A high NNT is acceptable if the disease we are treating is severe and the side effects minimal, though a balance will always be struck. As clinicians we have the responsibility to communicate these different scenarios to our patients.

In another scenario, let’s assume that 1 percent of male patients older than 45 years with specific red-and-white lesions will develop oral cancer. Mr. Smith, a 50-year-old, is told by his dentist that a novel method for detecting the risk of developing oral cancer is available for patients with his particular red-and-white lesion. This new test, he is told, is hailed by its manufacturer as predicting the development of oral cancer in 90 percent of patients with this red-and-white lesion who truly have cancer—a 90 percent sensitivity. Among those who will not develop oral cancer, 1 percent still will get a positive test result, or a false-positive result—99 percent specificity. If this test is administered to Mr. Smith and the result is positive, what is Mr. Smith’s actual risk of developing oral cancer? What should we tell Mr. Smith? This situation is one that all oral health care professionals face when they recommend certain tests that screen for, monitor or establish a disease diagnosis.

We need to create training opportunities for oral health care professionals regarding how to analyze risk assessment and how to translate that assessment into clinical care for individual patients.

There is a fairly straightforward way of determining Mr. Smith’s level of risk. Among 1,000 men older than 45 years and who have a specific red-and-white lesion, 1 percent, or 10 men, will develop cancer. Ninety percent of these men will have a positive test—nine men. Of the 990 men who will not develop oral cancer, 1 percent still will have a positive test—approximately 10 men. Thus, 19 men of 1,000 will have a positive test result, but only 10 will develop oral cancer. A positive test, therefore, would indicate an actual cancer risk for 10 people of 19, or approximately 50 percent of this patient population. This is the positive predictive value (PPV), a finding more useful to the clinician than are sensitivity or specificity. The PPV will help the clinician interpret the value of a positive test result, whereas sensitivity and specificity evaluate a test compared to a gold standard. In the example above the PPV was 50 percent, the same as if we were to flip a coin.

Health care professionals often are faced with statistical calculations and jargon that need to be translated into clinical reality. We need to create training opportunities for oral health care professionals regarding how to analyze risk assessment and how to translate that assessment into clinical care for individual patients. Manufacturers use different methods to show the benefit of their products, and health care professionals are faced with the sometimes daunting task of having to translate these claims for patients. We need to appreciate the scientific rationale behind research messages—but not necessarily believe everything we are told.

	Editor’s note:

TOP Editor's note:

 
This issue of JADA contains an article on the public health implications of the use of dental-protective chewing gums (page 553). It is a companion piece to the supplement mailed with this issue, titled Saliva and Oral Health. We hope readers will find both article and supplement useful.