Dentistry is a healing art that is well founded in science. Demands from within and outside the profession require that standards of practice be increasingly based on scientific evidence. Such evidence is acquired largely through published peer-reviewed research reports, commonly referred to as “the literature.” This chapter, which is adapted from the author’s postdoctoral course in research writing, discusses how scientific papers are structured according to the principle components of the research process.

One of the tenets of research is that the work be communicated publicly. A research report is the primary way to disseminate the findings of a study once it has been completed (after the fact). Similarly, a research proposal is written prior to beginning a study to obtain funding and/or permission to conduct the work (before the fact). Research reports and proposals have similar formats designed to explain to readers what the investigators did and found, or what the investigators will do and think that they will find, respectively.

Several types of research reports are found in the scientific literature. The predominant type is the original report, often referred to as a paper or article. Original reports typically are publications narrowly focused on a specific research question or idea that adds new knowledge, or confirms existing knowledge, in a particular scientific discipline. Most readers are familiar with the standard format used in original reports; namely, introduction, methods, results, and discussion, as described in detail below.

Another type of research report is the review, which summarizes sets of original reports in a scientific discipline. For the most part, reviews traditionally have been written by experts who often provide their own interpretation of the collected findings, and are therefore referred to as narrative reviews. Increasingly, however, a related type of review known as the structured review is being published, in part, to minimize the potential bias that may be associated with narrative reviews. Structured reviews follow stringent criteria regarding how articles are included and analyzed. These publications are particularly useful in documenting solid evidence in a field and also in identifying gaps or misperceptions that may exist.

Structured review articles can be both qualitative and quantitative in format. Qualitative structured reviews often use evidence tables to present summaries of articles that have been reviewed. Evidence tables are comprehensive listings of the salient aspects of each article that has been included in the review. These evidence tables allow the reader to judge the relative merit of the available evidence comprehensively and efficiently. Quantitative structured reviews additionally provide what is called a meta-analysis, which is a computational method for analyzing the data reported in a set of published papers. In essence, meta-analysis allows the reviewer to treat data compiled from individual studies as if such data were from one larger study. This is an extremely powerful technique to establish the strength of evidence, or lack thereof, in a scientific discipline.

Other examples of research reports include abstracts or proceedings of scientific conferences, which can be in the form of posters or oral presentations. A graduate dissertation or thesis is also a type of research report. Chapters in textbooks and monographs also can be categorized as research reports. Opinions of experts expressed in editorials or letters in journals can sometimes be considered as a type of scientific publication, although there are obvious limitations to how such information should be interpreted and used.

The predominant type of research proposal is the grant application. To be successful, grant applications must clearly describe the rationale, importance, and feasibility of a proposed research study. Grant applications also must clearly describe how the study will be performed, what results are anticipated, and how the results will be analyzed.

Increasingly, research proposals in the biomedical sciences have multiple levels of investigation: human, animal, and in vitro. By using humans as research subjects, investigators can identify clinical evidence of a particular disease or condition and can test new interventions or treatments. Animal models allow specific ideas or interventions to be tested that would not be feasible to do in humans. This is especially true in research activities that have little or no known benefit and relatively high risk for subjects. In vitro cellular, molecular, genetic, biochemical, and biophysical analyses are useful in identifying or confirming biological mechanisms that explain a particular condition, risk factor, or treatment outcome. As part of the grant application process, investigators often must apply for institutional permission to use humans, animals, and certain hazardous materials such as radioisotopes, recombinant nucleic acids, or toxic chemicals. Such applications themselves are proposals that must justify the particular permission being sought.

Original Reports Reflect the Basic Format of the Research Process

Research can be defined as any focused, systematic inquiry or activity designed to contribute to generalizable knowledge and enhanced understanding of a particular subject (Centers for Disease Control and Prevention 1999). It is critical that the activity be systematic; that is, it must follow set rules or patterns designed to produce valid conclusions and allow repetition and confirmation. Often called the scientific method, this approach ensures that the knowledge in a scientific discipline is based on objective facts rather than on unsupported opinions. The criterion of generalizability further ensures that such knowledge can be applied to populations or conditions outside the sample group or condition being studied within a particular research investigation. The research process follows a logical sequence. Typically, research begins with an idea or observation that forms the basis of a hypothesis that can be tested and subsequently evaluated. The process is cyclical in that subsequent systematic investigations modify the accumulated knowledge base as new evidence is obtained (Appendix Figure 15.1).

Research writing, like the research that it describes, also is a process having a logical structure. Scientific reports and proposals both follow a formulaic pattern designed to communicate the various components of a particular research endeavor. As indicated in Appendix Figure 15.2, research writing is designed to answer the following questions in original reports (or grant proposals).

What did (or will) you do?

Why did (or will) you do it?

How did (or will) you do it?

What did (or will) you find?

What does (or will) it mean?

What might you do next (or do if you hit a roadblock)?

Who helped (or will help) you, including paying for it?

These questions form the basis of the standard format used in original reports or grant proposals. The essential components of a scientific report or proposal often are presented as separate sections of the written work, but also can be collapsed and combined as deemed appropriate. The remainder of this appendix discusses these various components.

The introduction section of original reports generally states the purpose of the study, and also describes the background information that supports why the study was conducted. In performing the study, the researchers had to establish a balance between how interesting or important the research question was and how feasible it was to test, in terms of time, resources, and available methodologies. As suggested in Appendix Figure 15.3, a research question that is easily answered oftentimes may lack significance. On the other hand, an important research question may not be testable due to scientific or logistical constraints.

Any interesting research idea needs to be developed and refined so that it can be tested; otherwise, it remains merely an interesting idea with little practical merit. The process has analogy to the popular riddle: Which comes first, the chicken or the egg? In other words, one needs to know at least something about an area to come up with an interesting research question in the first place. The research question may be too broad or ill-defined at first, which then requires one to review the literature in the field to focus the question. This becomes a repetitive, iterative process until the nascent idea is transformed into a precise hypothesis worthy of investigation; that is:

The key point to be made is that the steps taken to develop a research question into a testable hypothesis are the same as the steps taken to conduct a focused review of the existing knowledge base for that same research question (Appendix Figure 15.4).

The Hypothesis

The hypothesis is a clear and concise statement of the research idea that is to be tested. Often, the hypothesis is presented as the purpose of a particular study with specific aims or objectives that list how the purpose was (or will be) achieved. Hypotheses are built according to relatively simple logical patterns: if/then, cause/effect, and intervention/outcome. In simplest terms, a hypothesis of any clinical study has the following structure:

Such a structure points out four key factors to be considered in focusing the research question into a testable hypothesis: (1) Problem and patients, (2) Intervention, treatment, or risk factor, (3) Comparison intervention or group, and (4) Outcomes or effect. The acronym PICO describes this well-accepted format for framing hypotheses (Richards 2007). Each of these four factors addresses the following questions, respectively:

1. What is the problem of interest and the patients or subjects it concerns?
2. What is the main intervention, treatment, or risk factor being considered?
3. What is the main alternative to which the intervention, treatment, or risk factor will be compared? To what group will patients or subjects of interest be compared?
4. What do the researchers hope to accomplish? What do they realistically expect to see? To what will the risk or exposure lead? What outcome would be particularly worrisome?

Some examples of PICO-formatted research questions are listed in Appendix Table 15.1. It is important to note that by following this format each example makes clear not only what will be tested but also what general methods or measurements most likely will be used.

Publication Databases and Search Strategies

There is little doubt that the biomedical literature has burgeoned during the past several decades. Increasing numbers of original reports are being published in increasing numbers of journals. This presents a considerable challenge for researchers interested in focusing a research question; that is, how to identify relevant publications efficiently and effectively without missing important ones and without obtaining ones that are not directly relevant. An important first step is to recognize that the literature is organized in several electronic databases that can be searched.

The most popular data base is MEDLINE, which is produced by the National Library of Medicine (NLM) in Bethesda, Maryland. MEDLINE is a comprehensive bibliographic database of citations to published journal articles in the biomedical sciences. It covers all aspects of healthcare: dentistry, medicine, nursing, allied health fields, biomedical and preclinical sciences, pharmacy, psychiatry, etc. MEDLINE indexes approximately 4,800 journals containing more than 15 million citations, dating from 1950 to the present. MEDLINE is free, open to the public, and available 24 hours/day, 7 days/week on the Web from any computer worldwide. This database is dynamic, with citations updated weekly. It can be readily accessed online through the PubMed portal (PubMed): http://www.ncbi.nlm.nih.gov/entrez.

MEDLINE is not the only database available to biomedical and social scientists. Others include Biosis; CINAHL (Cumulative Index to Nursing and Allied Health Literature); Cochrane Central Register of Controlled Trials; Cochrane Database of Systematic Reviews, Genetics, Genomics and Proteomics Databases; PsycINFO; TOXLINE; TOXNET; and the Web of Science databases known as Science Citation Index and Social Sciences Citation Index. These databases would also be searched depending upon the particular research question being asked.

After an appropriate database has been identified, it next becomes necessary to execute an effective strategy to search for relevant publications. In this regard, it is useful to recognize that databases are compiled by library science personnel who read and catalogue the articles according to set criteria. For MEDLINE, these criteria constitute a controlled vocabulary thesaurus known as medical subject headings (MeSH), which represent the subject content of each article. MEDLINE uses more than 50,000 MeSH terms (also referred to as subjects) with more than 30 subheadings that are attached to the MeSH to further describe a particular subject. MeSH subjects also are grouped in hierarchies called trees that organize the relationships among diverse subject headings. A tree progresses from the most general (broad) term to the most specific (narrow) term. The indexers at the National Library of Medicine assign 8–20 MeSH for each article, which can then be used by researchers in their literature searches.

The literature search process is generally organized into successive steps that result in the retrieval of a manageable number of relevant articles. The first step is to use various search terms (or fields) as needed. The MeSH terms are frequently used, as are specific text words from an article’s title or abstract. The key words listed in articles are also useful terms. Searches can also be conducted according to the names of known authors or specific titles of journals. Once a first pass through the available literature is completed, the search is then either widened or narrowed, depending on the initial results. This is referred to as exploding or focusing, respectively. So-called limits can be applied to large sets of articles to reduce the number of articles identified and refine the search. These limits can be according to subject age or gender, type of publication, years of publication, language, dental specialty, or other criteria. The search results can be combined to include only articles that contain more than one of the search elements. Once these are done, the most efficient next step is to review the title and abstract of each individual article, without reading the entire article, to determine if the retrieved article is relevant to the topic. This entire process then repeats until the researcher is satisfied with the search results.

It should be noted that universities appoint professional searchers, typically library staff, to assist researchers in developing their research agenda. These individuals are experts in information recovery and generally provide ongoing training for researchers in need of assistance. Such training is very useful in improving the efficiency and effectiveness of one’s literature searches.

Summarizing the Literature

The net result of the literature search is to establish a set of articles with the most relevance to the intended research investigation. These articles must then be read and summarized so that they can form the basis for justifying why the intended investigation is necessary, and what is hoped to be accomplished. This constitutes the introduction section in an original report (or the background/significance section in grant proposals). How these articles are actually summarized is usually a matter of preference of the individual researcher. However, to be useful, each article retrieved from the search process should be summarized with the following questions in mind:

1. What was the authors’ research question?
2. How did the authors attempt to answer that question (i.e., research design and methods)?
3. What were the results?
4. What is the article’s relevance to my own research question or project?
   1. Does it help indicate the current state of knowledge?
   2. Does it help argue the case for my own research question?

Once reviewed and summarized, the information reported in selected literature is compiled into a narrative that provides general background information and the details of the current state of knowledge in the relevant area. Arguments are then presented defending the need to answer the specific research question, which is formulated as a testable hypothesis. As indicated in Appendix Figure 15.4, these summaries are organized into a hopefully convincing story that provides the rationale for the intended research investigation.

Citations and Citation Management

The format of citations in original reports is specified by the scientific journal in which they are published. In the biomedical literature, these tend to follow the Uniform Requirements for Manuscripts Submitted to Biomedical Journals (International Committee of Medical Journal Editors 2007). Typically, one of two general formats is used: the numbering method for citing articles or the name and year method for citing articles. As the name implies, the numbering method cites references according to the order in which they appear in the publication. These are then listed in ascending numerical order at the end of the paper. In contrast, the name and year method cites references by indicating the name of the author(s) and the year in which the reference was published. These are then listed alphabetically at the end of the paper. Although each journal has its own rules, typically only up to three authors’ names are listed in the body of the text, whereas all authors are listed at the end of the paper. An excellent resource that describes citation management can be found at The Writing Center, University of Wisconsin-Madison (The Writing Center).

It also should be noted that several software programs to facilitate the management of references are commercially available. These help organize one’s personal electronic database of articles to which future articles can be added. These software programs allow citations to be easily inserted into the text of manuscripts that are being prepared for publication. A useful advantage is that the citations can then be listed at the end of the paper according to the varying formats specified by different journals; these formats come preset within the software program.

In discussing the methods used in research studies, it is helpful to distinguish between the general design of a particular investigation and the specific protocols used to generate data.

Study Designs

The design of a study refers to the general format of how the investigation is conducted. Several formats are commonly used in clinical studies, as reviewed in Callas (2008). These formats are based on several defining characteristics: the timing of data acquisition, the extent of influence or direct action by the researchers, and the amount of involvement and risk for the study subjects.

In the broadest sense, research study designs are either descriptive or explanatory. Descriptive studies are those that identify and report various characteristics of interest such as age, gender, race, geographic location, and incidence or prevalence of a particular disease, as examples, without testing a specific hypothesis. Descriptive studies are useful in generating information that can be subsequently used to develop hypotheses that can be tested. Clinical case reports, specifically those reported as case series, are examples of descriptive studies.

Explanatory studies (also referred to as analytical studies) are designed to answer and explain specific questions; that is, to actually test research hypotheses. These studies can be either prospective or retrospective: prospective studies collect and analyze data going forward from the start of an investigation, whereas retrospective studies collect data after an outcome has occurred or they analyze existing data that have been collected previously. Explanatory studies can be further classified as observational or experimental (also referred to as interventional). As the name implies, observational studies are those in which “natural” changes or differences in one characteristic (variable) are studied in relation to changes or differences in another variable(s), without any direct intervention by the investigator. In contrast, experimental studies are those in which the investigator actively intervenes by changing a particular variable and then measures what happens to other variables.

Observational studies can be further classified as cohort studies (also referred to as longitudinal), case-control studies, and cross-sectional studies. In cohort studies, groups selected by the presence or absence of a risk factor or other characteristic suspected of being a precursor for an outcome of interest are followed prospectively over time and the outcome is subsequently measured. In case-control studies, two groups are analyzed retrospectively to determine possible causes or risk factors for a particular outcome of interest. The two groups are defined by the presence (case) or absence (control) of the relevant outcome. In cross-sectional studies, data are collected at one point in time and then analyzed for the concurrent presence or absence of a factor suspected to be associated with a particular outcome characteristic. If data are compared between groups of subjects with and without the outcome characteristic, then the cross-sectional study is considered to be explanatory. If data for only one group of subjects is reported, then the cross-sectional study is more aptly considered to be descriptive.

Experimental or interventional studies are collectively referred to as clinical trials, which are designed to produce cause-and-effect relationships among variables of interest. In clinical trials, subjects are assigned into either experimental (test) or control groups. The experimental group is actively subjected to a suspected causal variable or intervention, while the control group is not, and predetermined outcome variables are then measured prospectively. There are several types of clinical trials that are characterized according to how subjects are assigned into the study groups and the nature of the control group.

The randomized clinical trial (RCT) has long been considered the gold standard in clinical research design. In the RCT, subjects are randomly assigned into either the experimental or the control group. Randomization is very important from a design standpoint because the process ensures that the two comparison groups are as similar as possible in multiple characteristics (for example, age, gender, health status), except for the suspected causal variable or intervention.

Nonrandomized clinical trials also can be found in the literature, but these studies are not considered as strong as the RCTs. Studies that compare two different groups (i.e., experimental and control) within the same study are considered stronger than those in which each research subject serves as its own control (called self-controlled trials, in which subjects participate in both the experimental and control groups at different times during the study). Along the same lines of reasoning, studies that compare two different groups within the same study are considered much stronger than studies in which the experimental group is compared with an external control group, either the general population itself or different groups studied in previous research investigations (called historical controls).

As described above, a systematic review is a type of literature review that attempts to identify, pool, and interpret available evidence on a specific research question, usually from RCTs, so that the strengths and weaknesses of the evidence can be made clear.

The various study designs can be summarized as follows:

• Case Series Description: Did you see something interesting? These are a grouping of anecdotal observations about a particular outcome that helps to generate initial research ideas.
• Case-Control Studies: What happened? In this type of study design one group of subjects already has a particular outcome (cases), as compared to another group that does not (controls).
• Cross-Sectional Studies: What is happening? Groups are examined at one point in time for the presence or absence of a particular outcome.
• Cohort (Longitudinal) Studies: What will happen “naturally”? Groups are followed over time for the occurrence or non-occurrence of a particular outcome.
• Clinical Trials: What will happen “experimentally”? One group (experimental) is subjected to a specific manipulation while another group (control) is not; both groups are examined at a future point in time for the presence or absence of a particular outcome.
• Systematic Reviews/Meta-Analyses: How can existing data in separate studies be critically summarized so that the “real” answer to a research question is identified? A systematic review uses rigorous, objective criteria to retrieve, evaluate, and summarize published scientific papers that are relevant to a topic. A meta-analysis is a statistical method used in certain systematic reviews that allows quantitative data published in different scientific papers to be evaluated and combined as if they were all from one large study.

Hierarchy of Evidence

One of the fundamental goals of biomedical research is to discover and advance knowledge for alleviating human abnormalities and diseases and improving overall quality of life. Thus, research study designs have been ranked according to how directly applicable their respective findings may be to the human population and how well potential sources of study bias or error have been reduced. This ranking is often referred to as the hierarchy of evidence and includes in vitro, animal, and human clinical studies. This hierarchy can be illustrated by an evidence pyramid, as shown in Appendix Figure 15.5. The figure is available from the online Evidence Based Medicine Course at the Medical Research Library of Brooklyn, State University of New York Downstate Medical Center. This is an excellent tutorial on evidence-based medicine that can be accessed freely (Markinson).

The different types of evidence are labeled in the figure, with the least clinically relevant at the bottom and the most clinically relevant at the top. The top five layers indicate evidence generally considered strong enough to be clinically relevant (i.e., directly influencing clinical decisions). The bottom layers have considerable merit in terms of providing scientific information, but do not provide sufficient strength of scientific evidence to warrant direct relevance to humans. As illustrated, systematic reviews and meta-analyses are considered to be the highest level of evidence in biomedical research. These serve to increase the credibility and power (discussed below) of individual studies, and are extremely useful in helping busy practitioners distinguish between reality and hype, so to speak, regarding diagnostic and treatment interventions.

Study-Specific Protocols

These are intuitively understood as constituting the methods of a study. In a periodontal clinical study, for example, measurements of attachment loss, probing pocket depth, recession depth, gingival index, plaque index, or bleeding on probing would each be done according to specific protocols. Researchers determine a priori what measurements would be made, how many teeth or sites would be evaluated (for example, Ramfjord teeth) (Fleiss et al. 1987), by which instruments (for example, manual vs. electronic periodontal probe), and how often (for example, number of posttreatment recall visits). Oftentimes, these protocols are well established and commonly used across many different studies. In such instances the protocol is merely indicated by a published citation without much description of the details of the technique. A classic example of this is the highly cited Gingival Index (GI) of Loe and Silness (1963). On the other hand, if a well-accepted protocol has been modified, or if a new method has been developed for the particular research study, then it is expected that a more detailed description will be provided. Thus, a balance is established between clarity and brevity in published reports. Appendix Figure 15.6 provides some examples of study-specific methods. The list is for illustrative purposes only and is not meant to be comprehensive. Moreover, many of these are applicable to human subjects and animal models, depending on the objectives of the research study.