It is interesting to observe that the Saudi Dental Journal has not only a biostatistician editorial board member but also a biostatistician on the advisory board. This far surpasses many of the United States dental journals where the formal need for such experts has yet to be recognized and implemented. In this regard, the wisdom of the Saudi Dental Journal editorial board is to be commended. 

Of particular importance to the work of statistical editors is the recognition of the basic differences between the approaches of the clinician and the biostatistician to the solution of dental problems. In fact, this paper is the result of long-standing conversations and writings of two people, one of whom is a clinician and the other a biostatistician. Dr. Barry Mollenhauer, the clinician, who is a renowned Australian orthodontist, has written that statistics reduces knowledge to generalizations. The biostatistical view is that statistics often report means and standard deviations in order to represent or summarize an entire set of data. Subsequently, tests of significance are carried out and results presented, again in terms of the computed statistics. From the outcome of these tests, generalisations are made despite the agreement of Dr. Mollenhauer  with Nietzsche that "generalisations are to some extent...falsifications". Yet it must be emphasized that good statistics provide the most accurate way to synthesize information. Too often, a mean is interpreted as an absolute, with no consideration of the variability of the data, which is provided by the standard deviation. Statistics is also the best tool to show what outcome is most probable, since statistics are probabilistic in nature. On reflection, "this makes statistics appear more mature and clinically relevant to the experienced clinician, since experience shows that few things are absolute or black-and-white".    

The clinician observes that "it must be recognized that mathematics is an abstraction, therefore copious instantiation should at least be on an equal footing with substantiation". Instantiation means presenting instances or cases, whereas substantiation means putting in concrete form based on mathematical evidence such as the biostatistician does. Years of clinical practice entails seeing and managing many patients, some of whom may represent outliers or subpopulations. Yet these terms are seldom seen in evidence-based papers. If a study does indeed find outliers, they are often discarded since they are extreme observations, despite the importance of these observations in answering the research questions. They may hold the key to the truth. Again, correct interpretation is often compromised if the variability of the data is not considered. Outliers produce large variability, as they are distant from the average value. When they are discarded, variability is artificially reduced. It therefore appears as if statistics is indeed reductionist whilst the experienced clinician uses every piece of data. That is, it seems as if statistics invariably reduces and therefore loses data, in part due to lack of proper interpretation. So, the clinician favours instantiation. At the far opposite end of the spectrum, the biostatistician clearly favours substantiation. And philosophically, this is a major difference between the clinician and biostatistician and their attempts to resolve a problem.         

It must be acknowledged that many clinicians dislike statistics. Dr. Mollenhauer indicates in his writing that their dislike is usually on three grounds. One dislike arises from ignorance because clinicians often do not understand that statistics is based on probability. This means that "clinicians are susceptible to the commonly implied naive idea that statistics can prove something in absolute terms". Yet the biostatistical view is that nothing is "proven" by statistics. Formal rejection of a hypothesis means that the anticipated outcome was likely or probable at a certain level of significance. Because of the probability involved, the outcome is not an absolute certainty. And if a hypothesis is not rejected, results are often taken as confirming that no difference exists. In fact it only means that under the circumstances of the research, we were unable to detect a difference at a given level of probability. The clinicians' second dislike "comes from too many papers using statistics reaching conclusions they perceive to be erroneous". This may be due to poor research design, but it may also be due to the clinician's incorrect bias. Power analysis enables us to estimate the sample size needed to find statistically significant results, which lead to rejection of the null hypothesis. Power analysis is seldom carried out, which leads to the observation that sample size may often be inadequate to find a presumed or hypothesized outcome. Appropriate sample size is an integral part of research design. By using a larger sample, we can do a better job of reaching a good decision about the hypotheses under question. The third dislike is that "clinical success is based on details" (instantiation), so that methods that reduce data to mathematical evidence (substantiation) are generally unacceptable for the clinician.           

From the statistical point of view, in order to better evaluate complex clinical data, more sophisticated statistics are available such as multivariate analyses. For clinic patient data, often many variables are measured, and the advantage of the multivariate methods is that the intercorrelations or relationships among the many variables are taken into account by the analysis. Since the variables are interconnected in nature, over-simplistic statistics may not adequately reflect the clinical situation, thus making preferable analyses which consider the relationships as opposed to the utilization of many independent tests. In addition, data should be routinely evaluated for normality prior to using statistical tests which depend upon normality. There should be greater utilization of non-parametric tests when normality is not demonstrated.

Is there a place for clinical research and statistics for experienced clinicians? The answer is yes, and the place is in the dental clinics in private or government sectors, as proposed in the December 1999, Saudi Dental Journal editorial. According to the clinician, "each clinic has its own set of variables. Therefore, each clinic has to establish what is the best set of variables for itself". Furthermore, agreement is reached by the author and the clinician that a study done in one clinic or teaching department may or may not be applicable to another clinic. Dr. Mollenhauer observes that "as with all good experimental designs, key features are full documentation and standardisation of procedures to control many of the remaining variables". Interestingly, most of the published clinical researches are carried out in teaching departments, usually with novice student operators as opposed to the experienced clinicians found in the dental clinics. There are two apparent problems with this. One is that very likely, there are considerable differences between the work of students who are in the process of learning clinical skills and that of experienced clinicians who have presumably perfected the skills. Thus, generalisation of the results based on research in the teaching departments to the private or government sector does not seem reasonable. And secondly, each of many students often contributes only a few subjects to the study making assessment of operator differences difficult if not impossible. However, research in the private sector requires an infrastructure that includes funding, research advice, biostatistical consultation, and editorial support as proposed in the earlier editorial. It is crucial for the advancement of dental research in the Kingdom of Saudi Arabia and subsequent improvement of the quality of dental care delivery that the private and government sectors be involved. Only research done in these sectors adequately reflects the contributions of practitioners to the dental care delivery system.

Ellen A. BeGole, BS, MS, PhD
Associate Professor of Biostatistics,
Departments of Orthodontics and Endodontics,
College of Dentistry, University of Illinois at Chicago,
Chicago, Illinois, USA;
Member of the Advisory Board of the Saudi Dental Journal.