Converting Science into Art: The Challenge of the Translationists

Translating Evidence into Practice 

A common justification for the advent of evidence-based dentistry and medicine is timeliness. The Institute of Medicine reported that it took 17 years before research results got incorporated into practice.1 Examples of delayed implementation of life-saving treatments are aplenty. Before 1970, there was a 5% to 10% 4-year survival rate for Hodgkin disease. In 1970, a study reported that 80% of the Hodgkin's patients could be brought into complete remission, yet it took 11 years before results were disseminated.2 Similarly, for testicular cancer a treatment was identified that had a cure rate of 80%, yet it took 3 years before results were disseminated.2

In part to avoid such delays, evidence-based medicine came to the forefront. A new term was coined for people who translate research into clinical practice: translationists. The main challenge for clinical translationists is to decide when to take action, when to translate. Consider the following example on levels of evidence on substance X and translation.3 Among people at low risk for coronary heart disease (CHD), the level of substance X is approximately 0.5 nmol per liter, among high-risk people the level is approximately 0.4 nmol per liter. Evidence on substance X is so convincing to some that individuals with low levels of substance X are labeled as diseased. Without evidence that increasing substance X will lower disease risk, substance X becomes approved by the Food and Drug Administration (FDA). Is this product ready for translation?

Better evidence comes along. There is now a cohort study on approximately 50,000 individuals with a 10-year follow-up. The study controlled for risk factors and reported that “the overall relative risk of major coronary disease with taking substance X was 0.56 (95 percent confidence interval, 0.40 to 0.80).”4 Longer follow-up confirmed this effect.5 Is substance X ready for translation now?

Even better evidence appears. Three meta-analyses are reported on multiple epidemiological studies reporting that there now exists “extensive and consistent observational evidence that substance X reduced the risk for CHD about 35%.”6 Furthermore, it is reported that substance X for individuals with heart disease leads to a larger increase in life expectancy than treatment for mild or moderate hypertension. Numerous organizations are encouraging translating the research findings. The American Heart Association and the American College of Physicians both recommend substance X and managed care organizations considered prescription of substance X a criterion for good medical practice.6 The translation was working. An increasing large number of individuals started talking substance X.7

Let us unmask now what substance X is. Substance X is estrogen-progestin. The Women's Health Initiative suggests estrogen-progestin increased the risk for CHD by 29%.8 Premature translation may have caused tens of thousands of premature deaths among postmenopausal women.9 This case report illustrates that potential danger of labeling healthy people as diseased and to translate evidence that is not based on rigorously designed randomized trials into clinical practice. This is not an isolated example. Many randomized controlled trials have refuted results of widely held beliefs that were translated into clinical practice long before unequivocal evidence became available. Supplemental dietary fiber was advocated for decades to healthy individuals to prevent colon cancer. For instance, a World Health Organization publication reported in 1989: “There is very strong support that dietary fiber protects against colorectal cancer…The greatest effect is seen with wheat bran, and this has been documented by many groups in many countries.”10 This recommendation remains up to this day unsupported by randomized controlled trials. A systematic review reported by the Cochrane organization reported in 2002 that 5 randomized controlled trials on a total of over 4000 individuals failed to identify a reduced risk of colon cancer (Asano T, McLead RS (2002). Dietary fibre for the prevention of colorectal adenomas and carcinomas. Cochrane Database Syst Rev 2:CD003430.). Antioxidants and vitamin supplements have become embedded in popular culture and are widely marketed to healthy individuals. Evidence in support of vitamins is based on the same suspicious epidemiological studies that advocated estrogen replacement therapy. Randomized controlled trials have now shown that these antioxidants, just like hormone replacement therapy, increases mortality risk.11

These experiences suggest that translationists should only act cautiously, and that maybe “un-translationists” should work alongside translationists. Un-translationists should be given the task of undoing the damage done by translationists. Give the stickiness of ideas and beliefs their job may be particularly challenging.

When it comes to giving advice to healthy people, translationists should stick with unequivocal evidence of randomized controlled trial based on true end points. Suggestions for translation involving agents with suspected harm are a lot more complex. Unequivocal evidence on suspected harm by means of definitive randomized controlled trials is unobtainable. The approach under such circumstances can range from a precautionary approach to a risk-based approach.

Dentistry may be a unique among all prevention businesses in that preventive dental treatments are recommended almost worldwide, are commonly used almost from cradle to grave, and yet remain based on the weakest of evidence. This problem is further complicated by the tendency in dentistry to label most people as diseased. Just like it was dangerous to label menopause as a disease, it may be dangerous to label dental conditions as disease. Examples of labeling everybody as dentally diseased are aplenty. A recent survey suggested that over 95% of subjects showed some symptoms of gingivitis,12 a national survey suggested that 65% of US residents had noticeable incisor irregularity,13 92% had dental calculus,14 and approximately 30% had intraoral mucosal lesions.15 Does there exist a person whose oral health is well? Does there exist a person who is not in need of some dental intervention? This situation is reminiscent of an editorial written in the New England Journal of Medicine on “The last well person.”16 Maybe a similar editorial should be written on the “last dentally well person.” Such a high prevalence of nonwell dental people justifies many treatments, preventive and others, and yet unequivocal evidence on the effectiveness of such treatments remains commonly lacking.

Consider the following case history. Substance X is recommended to prevent caries based on biological plausibility. The FDA suggests that substance X can be used to reduce tooth decay (Food and Drug Administration, 21CFR872.6390). Yet, there is minimal epidemiological evidence of effectiveness. Neither is there supportive evidence from randomized controlled trials in adults.17 The one clinical trial that is most reflective of typical use concluded that substance X was not associated with reduced caries risk (relative risk, 1.01; 95% confidence interval, 0.85-1.20; P value, .93). Is this substance X ready for translation?

Substance X is dental floss, which has become almost synonymous with western civilization and is recommended to everybody with teeth. There may be little reason to suspect a string of inert material may have any adverse general health consequences. But then again, there was little reason to believe that vitamin supplements could significantly increase mortality risk. How convinced are we about the benefits and the harms of other widely recommended dental interventions, preventive or other? The “presumption of preventive medicine,” as referred to by the Canadian epidemiologist David Sacket [Taubes G (2007). Do We Really Know What Makes Us Healthy? New York Times. New York, NY] and dentistry needs to be skeptically evaluated as long as no randomized controlled trials are available.

Translation from clinical trials to clinical practice may be less challenging when dealing with individuals who are in pain or are suffering. In particular, the more serious or life-threatening the disease, the less unequivocal evidence is required on promising treatments. The 100,000 lives campaign18 spearheaded by Don Berwick provides an illustration of how translation of interventions, some of them with less than optimal evidence, in sick populations can have a significant impact on mortality. In 2003, 100 patients died each day in hospitals in the United States because of care they received, not the disease from which they were suffering.19 In December 2004, Don Berwick started a campaign to save 100,000 lives over 1.5 years and in June 2005 the campaign achieved its goal: 122,300 lives were saved.18

Is a similar translation campaign possible in the dental arena? Some dental interventions can have adverse side effects. For instance, consider patients with recent coronary stenting who are about to undergo an invasive dental procedure. Taking such patients off anticoagulant therapy during the first year after coronary stenting increases mortality risk 10-fold. If anticoagulant therapy is stopped, the mortality risk is 7.5%. If it is not stopped, the mortality risk is 0.7%.20 Not knowing this information may kill a patient. Such efforts should and are translated by organizations such as the American Dental Association but little is known how well this translation works. Some translations do not work well. For instance, the American Dental Association recommended E-speed in 1988, yet the compliance with this guideline in the ensuing decades was low.

For a variety of practical reasons, D-film technology remained the standard of care in 73% of US dental offices up until 1999.21 Not translating information on ionizing radiation may adversely impact patient health. With a US population of approximately 300,000,000 individuals and a high prevalence of dental care, it may not be unreasonable to assume that approximately half may receive the equivalent of 1 full-mouth D-film dose. An outdated, but commonly used, statistic to reflect the fatality risk of this dose is that 1 in 100,000 patients will die as the result of such an exposure. If 150,000,000 are exposed, this could mean 1500 deaths. Would a translation campaign to reduce the dental radiation exposure by 10% (150 lives saved) be worthwhile? Similar questions can be raised regarding ionizing radiation exposures associated with orthodontic practice.22 Translation is important.

Translating Basic Science Evidence into Clinical Trials 

The challenges for basic science translationists are serious; the future is inherently unpredictable, and the significance of most discoveries is not recognized until years, if not decades later.23 In 1971, President Nixon promised a cure for cancer within a decade. In 1997, President Clinton promised an HIV vaccine within a decade. Medical breakthroughs cannot be predicted, as they commonly occur accidentally, rather than on command. Flemmig identified penicillin accidentally; Viagra's surprising side effect was not intended. Maybe more surprisingly, most breakthroughs are not recognized as significant at the time of discovery. The importance of hand washing in the medical setting did not become firmly accepted until the germ theory came around.

The responsibility of translationists is substantial, as decades of research, hundreds of millions of dollars, and tremendous opportunity costs are at stake with wrong decisions. MRFIT in 1982 was conducted at an alleged cost of $115 million, the Women's Health Initiative at a cost of more than $800 million, and the National Lung Screening Trial at a cost $350 million. Conduct of such trials often reflects decades of work. To avoid negative results, it has been suggested that large, definitive randomized controlled trials should be initiated— when possible—when available sound epidemiology is supportive, when mechanistic studies have been conducted, and after both Phase I and Phase II studies have been completed.

The need for sound epidemiology can be illustrated with the story on Chlamydiae and myocardial infarction. A systematic review of the epidemiological evidence in 2000 created “considerable doubt about the existence of any independent association between Chlamydia pneumoniae and coronary heart disease.”24 Should one proceed with the conduct of definitive trials in light of such evidence? Randomized controlled trials were initiated and a meta-analysis of trials concluded that antibiotics against Chlamydia pneumoniae provided no benefits,25 and that macrolides increased mortality risk, mostly from coronary heart disease.26

For low birth weight, which was similarly hypothesized to be the result of periodontal or other “infections,” randomized trials similarly indicated that antibiotics such as metronidazole increased the risk of preterm birth.27 In periodontal disease, randomized controlled trials were initiated even though the available epidemiological was unsupportive of the hypothesis (see Figure 1). The first randomized controlled trial results published in the New England Journal of Medicine appeared to confirm the epidemiological summary of no association (Figure 1).28 Three additional randomized controlled trials are anticipated to be completed at the time of publication of this report. The MOTOR study had an estimated study completion date of May 2007. The Periodontal Infection and Prematurity Study had a reported anticipated completion date of May 2008, and the SMILE study had an estimated study completion date of June 2008. Time will tell whether the advice to only initiate randomized controlled trials when sound epidemiology is available holds for the periodontal disease and low birth weight trials.

View Large Image Download to PowerPoint

Figure 1. Periodontal Disease and LBW.

Mechanistic studies have also been suggested as necessary before the initiation of definitive randomized controlled trials. With β-carotene there were few mechanistic studies available “Deciding whether beta-carotene or retinol/retinoids should be the primary candidate for prevention trials and at what doses was speculative.”29 Given that there are 272 isomers of β-carotene, that retinols and retinoids and other carotenoid antioxidants such as lycopene and lycopeins could have been selected, such a speculative attitude in selecting an intervention has been considered “hard to forgive.”30 Similarly, the apparently haphazard dose determination without consideration whether supranormal doses may be harmful was, in hindsight, a mistake. It has since been discovered that, as opposed to retinol/retinoids, there are no nuclear receptors for β -carotene, that there was little evidence that β -carotene suppresses the progression of neoplasia, and that there was no evidence that β -carotene inhibits the oxidation of low-density lipoproteins.31

Mechanistic studies on the periodontal disease–low birth weight association have similarly been lacking. How does scaling reduce low birth weight risk? Does it achieve its goal by reducing an alleged periodontal infection? If so, what alleged infection and should such an infection be diagnosed before treatment? How long does an infection need to be suppressed if it is to affect low birth weight, and what are the critical periods with respect to fetal development? Or, does scaling achieve its goal by suppressing periodontal inflammation? Or, does scaling achieve its benefits through other mechanisms? Is periodontal scaling the “right” periodontal treatment? Answers to such questions may be desirable to ensure a higher success rate of definitive randomized trials.

Finally, before the initiation of definitive trials, there should be a full set of exploratory studies. The absence of such small clinical trials for β-carotene was, in hindsight, a blunder, as it could have altered the design of the large-scale studies.32, 33 Is the absence of such small trials a similar mistake when considering the ongoing definitive trials on low birth weight and periodontal disease? Time will tell.

Conclusions 

Translating evidence into clinical practice is a necessary aspect of evidence-based practice. Concerns may arise about translating evidence on interventions that are assumed to be beneficial, yet for which no rigorously randomized controlled trials are available. Experiences in medicine suggest that such translation may be dangerous, especially when it comes to prevention in healthy populations. Translating evidence on suspected harmful agents is more complex, as it is impossible to conduct randomized controlled trials under this situation. Under such circumstances, risk-based and precautionary philosophies often clash. Translating basic science evidence into clinical trials is particularly challenging. It has been suggested not to invest in definitive randomized controlled trials unless sound epidemiology is available, mechanistic studies provide understanding of the mode of action, and a complete series of Phase I and II trials have been conducted.