Case Report and Case Series

A case report (singular) or case series (plural) is a description of one or several individuals with a disease or condition of interest. A case report can offer insights into diagnosis and management of a disease or condition by providing details about the patient’s clinical presentation, diagnostic work‐up, differential diagnoses, final diagnosis, management, and current disposition. Examples include descriptions of: orofacial manifestations of a patient with a systemic disease and strategies to manage the disease, unusually shaped teeth in a child or children with a genetic syndrome, or an adult presenting with orofacial pain from an unusual source such as a metastatic tumor and the diagnostic approach to determine the pain etiology. The description should be complete enough for use by another clinician who may evaluate a similar case. If the study is a case series, the same diagnostic criteria should be used to group the cases together for a report.

Case series can be very valuable in the description of new diseases or conditions. A good example is the large case series describing 63 cases of osteonecrosis of the jaw (ONJ) associated with the use of bisphosphonates.⁴ While the report of this emerging clinical condition suggested a relationship to use of bisphosphonate medication, an obvious limitation of this study design is the lack of a population of individuals without the disease or condition, or a “control” group. Other limits of a case series include the fact that most data are obtained via a retrospective review of existing clinical records. This introduces the potential for recall bias as the researchers are “looking back” at events and extracting record information, which often is a mixture of complete and incomplete facts. Also, the information is recorded for clinical care and not research purposes. Therefore, clinicians will use varying methods to evaluate patient outcomes, such as a non‐healing extraction site. If the patients were evaluated as part of a research study, the study team would use a predefined set of criteria to determine study inclusion and judge clinical outcomes and would collect a predefined set of information from the patients such as current and past medications.

Cross‐Sectional Studies

Cross‐sectional studies are employed frequently in clinical research. Research participants are evaluated at one time point and are not followed over time, creating a dataset that is a “snapshot” of the condition under study. Prevalence studies use cross‐sectional designs that describe the population under study, derive a representative sample of that population and define the characteristics under study to establish the prevalence of a disease or condition in a population.⁵ For example, the prevalence of oral human papillomavirus (HPV) infection in unvaccinated men and women has been estimated through the National Health and Nutrition Examination Survey (NHANES) 2009–2016.⁶ The NHANES study uses a statistically representative sample of the civilian non‐institutionalized US population. Many factors must be considered when designing a cross‐sectional prevalence study. First, it is not usually feasible to examine an entire population of individuals with a disease or condition. Therefore, the sample being examined should represent the entire population at risk and not only those most severely affected. In the example of ONJ, patients with small non‐healing affected sites that healed in two to three months without any intervention should be included as well as those with large lesions that persisted for months, to represent the entire spectrum of the disease. Second, all research participants should be evaluated using the same, standardized methods (read “Outcome Assessment” below). Prevalence studies require very large sample sizes and, therefore, may need to be conducted at more than one study site (multi‐centered) to achieve adequate enrollment. When the number of individuals with the disease of interest is very limited, it may not be possible to conduct a prevalence study.⁵

Cross‐sectional studies may also be utilized to assess relationships between an exposure or risk factor and the presence of disease. Because research participants are evaluated at one time point, causal inferences cannot be drawn between the risk factor and disease, representing a major limitation of this study design. Using the example of periodontal disease and cardiovascular disease, some cross‐sectional studies have suggested an association between the two conditions. It is important to recognize that the two conditions can occur together in a person because of a common underlying etiology, such as smoking, environmental exposures, and/or limited access to the health care system.⁷ Cross‐sectional studies linking these conditions do not prove that one causes the other. The temporal relationship, that is, which condition occurs first, cannot be determined from a cross‐sectional study. Nevertheless, such cross‐sectional designs have value in research, particularly to develop hypotheses for future studies. An initial relationship between a risk factor and presence of disease may be established in a cross‐sectional study before consideration of a more resource‐intensive study design in which risk factors for disease can be evaluated over time. When establishing initial relationships using a cross‐sectional design, the biologic plausibility between the risk factor(s) and disease and potential biases due to data collection at one time point should be described. For example, when exploring the relationship between periodontal disease and cardiovascular disease, study participants may report healthy behaviors such as frequent flossing if they have started to alter their oral health behaviors.

Case‐Control Studies

A case‐control study is an observational study in which the objective is to evaluate persons with the disease or outcome of interest (cases) and compare them with another group of persons without the disease or outcome (controls) to determine if certain exposures (such as being a current smoker or taking a specific medication) or characteristics are associated with the disease or lack of disease. If the exposure is found more frequently in the cases, it is termed a “risk factor” for having the disease. Sometimes the exposure is found more frequently in the control group, suggesting it might be a “protective factor” that helps protect against a disease. Because this design evaluates individuals with and without the disease or outcome of interest, it may be used to assess the presence or absence of disease, but not development of the disease. There are critical design issues that must be considered in a case‐control study.⁵ The exposure and disease in both cases and controls should be assessed in the same manner. Patients who have a severe disease may experience recall bias in that they remember more or over‐report past exposures or symptoms than generally healthy controls because they are seeking an explanation for why they have a disease. The cases need to represent the entire population of those with the disease, and the controls must be selected from the same population as the cases and should have the same prognostic characteristics. Finally, most experts recommend evaluating at least an equal number of controls as cases and matching cases and controls on variables that may differentially affect the exposure and disease. Selection of controls for a case‐control study can be difficult and can introduce bias into the study if not chosen carefully, as discussed at length in the literature.^8–10 A case control study performed across three dental practice‐based research networks assessed risk factors for ONJ.¹¹ ONJ cases were defined as having maxillary or mandibular exposed bone of any size that clinically appeared necrotic, without regard to duration or size. For each case, three controls with no current or previous history of bone necrosis were selected from the same dental practice where a case was diagnosed. Risk factors were ascertained in cases and controls, and the odds of having ONJ in patients who took bisphosphonates was compared to the odds of having ONJ in those who did not take bisphosphonate medication.

Case‐control studies are particularly beneficial when studying rare diseases. If the disease of interest is sufficiently rare, such as salivary gland cancers, it may be safe to assume that a sample of cases is representative of the entire population of those with the disease. Findings in case‐control studies are typically reported as the odds ratio of the exposure, whereas cohort study (see below) findings are expressed in terms of the relative risk of exposure. Case‐control study results cannot be reported as the relative risk, because the investigator determines, and can arbitrarily change, the disease prevalence within the study by setting the number of study participants with and without the disease or outcome of interest. When interpreting the results of case‐control studies, limitations in assessing temporality between the exposure and disease and the confidence interval of the association should be considered before making conclusions about the validity (a term used to describe how well the study measures that which it is intended to measure) of the results. A finding of a “dose‐response” (in which increasing levels of the exposure such as pack‐years of smoking are associated with increasing rates of the disease or condition) increases the strength of the evidence.

Because of criticisms about potential bias arising from control selection in case‐control studies, researchers may choose to utilize more than one control population when designing studies. In a classic example from the medical literature, the relation between estrogen use and endometrial cancer was established using a well‐designed case‐control study design and two control populations.¹² Cases of endometrial cancer admitted to hospitals (451 cases) were matched to two sets of control populations admitted to hospitals without endometrial cancer, one taken from those seeking gynecological services (442 cases) and another from those who did not seek gynecological services (446 controls). The choice for selection of two hospital control populations was to offset concerns about surveillance bias; that is, that estrogen users may be under greater surveillance for disease development.

Longitudinal Cohort Studies

Longitudinal cohort studies allow the opportunity to collect data over time. The purpose of this study design is to assess associations between an exposure or risk factor and subsequent development of disease or to determine outcomes of standard of care treatment. When performed prospectively to assess associations between an exposure and disease, a representative sample of the population of interest is assessed for an exposure at the beginning of the study, and then new cases of disease accrue during a period of follow‐up evaluation. At the end of the study, the differences in exposures between those with and without the disease are evaluated. In some cases, a single population is observed over a period of time to observe the natural incidence of a condition or the natural history of a disease. For example, a study of Swedish adolescents estimated the incidence of temporomandibular disorder (TMD) pain. All individuals aged 12 to 19 years in all Public Dental Service clinics in a Swedish county from 2000 to 2003 were followed over 3 years for development of TMD pain.¹³ Research participants with TMD were evaluated for differences that distinguished them from participants without TMD. In this study, TMD incidence was found to be greater in older children and in girls. More frequently, research participants may be selected for a particular exposure, along with a comparable group of controls without the exposure, and both groups are followed over time for development of disease.¹⁴ An example of a longitudinal cohort study examining outcomes of treatment is a study of 372 individuals with head and neck cancer who were enrolled prior to undergoing radiation treatment and followed to determine the rate of oral complications such as oral mucositis, oral pain, and oral health‐related quality of life.¹⁵

Cohort studies may also be retrospective, in which the exposure was captured in a standardized manner in the past, disease status is determined at a point in time before the outcomes of interest developed, and participants are followed over time. This study design assumes that the participant population (exposed and unexposed participants) is representative of the general population, and exposure history is collected accurately. Definitions of disease outcome should be reliable and reproducible and held constant during the study duration. Standard criteria for determining the disease outcome should be applied to exposed and non‐exposed participants to avoid bias.

One significant advantage of well‐conducted prospective cohort studies over other study designs is that the exposure is collected in a standardized manner, and cases are incident (new cases). This design provides more information about the natural history of the disease, as well as direct estimates of incidence and relative risk.⁵ Longitudinal cohort studies have the potential to initially or further establish the temporal relationship between exposure and disease and a dose‐response relationship, both of which increase the strength of the study conclusions and may provide evidence about the association (or causality) between an exposure and disease. An important factor in prospective longitudinal cohort studies is the ability to retain the cohort over time. Participants who drop out of research studies may differ from those who remain and may introduce attrition biases into the population sample.

Longitudinal studies by their nature are resource intensive. Large sample sizes for rare diseases and long durations for chronic diseases may be required. Maintaining the use of consistent study methods, such as standardized collection of the exposure, and retaining research participants in the study are continual challenges.

Clinical Trials

The purpose of a clinical trial is to determine whether a particular intervention is associated with a change in a prespecified health‐related outcome. A clinical trial involves prospective assignment into one or more intervention groups and assessment of an outcome measure.

Clinical trials can be classified into four phases (Phase I, II, III, or IV) or stages.¹⁶ This step‐wise approach reduces the potential for harm from a previously untested intervention and allows investigators to assess safety and determine potential efficacy of a new treatment while minimizing time and costs. A Phase I trial often is the “first‐time in human” study, meaning trial participants are the first humans to receive the new drug or treatment. These studies are not randomized or blinded and are typically performed with a small number of participants. The primary goal is to evaluate the safety of the agent and determine a safe dose range for subsequent studies. A Phase II trial tests the new drug in individuals who are randomized to different treatments, with goals of determining potential efficacy and establishing a more complete safety profile. Feasibility of using the treatment also can be determined. The Phase III trial enrolls hundreds or thousands of participants and is sometimes called a “pivotal study.” These trials are designed to test efficacy in a much larger segment of the population with the disease or condition, and results are used to gain drug approval from government agencies. Phase III trials should generate generalizable results. Phase IV trials are post‐marketing studies to determine how well a treatment found effective in a Phase III trial works in the community and to assess any side effects associated with its long‐term use in the overall population.

A randomized controlled trial (RCT) compares participants receiving the intervention under study to a control group, such as participants receiving another treatment, usual care, placebo treatment, or no treatment. Potential study participants from a well‐defined study population are assigned at random to receive or not receive the intervention(s) under study, and then well‐defined endpoints are measured at a specific time point. Intervention efficacy is assessed by comparing the outcome measure between the intervention group(s) and the control. RCTs provide the strongest evidence for the causal nature of a modifiable factor (such as inflammation in a periodontal pocket), and the effect that modifying the factor has on disease outcomes (such as reduction in pocket depth).

A key component of RCTs is that research participants are assigned to one of the study arms at random to eliminate the potential for bias in treatment assignment. Random, concealed, or “blinded” allocation of treatment helps ensure that any baseline differences in the treatment groups arise by chance alone. The random allocation process involves generating an unpredictable random sequence and then implementing the sequence in a way that conceals the interventions until participants have been formally assigned to their groups. Both randomization and concealment are necessary to maximize validity in RCTs, and reproducibility of the allocation order and the concealment process are necessary to maintain integrity of the research study. Other important features of high‐quality RCTs include independent or “blind” assessment of research endpoints and data analysis based upon the treatment assignment, also known as analysis by “intention to treat.” Intention to treat analysis removes artifacts from the study that are caused by unequal attrition in the two study arms, or by treatment crossover.

There are three levels of concealing treatment (blinding) in an RCT: (1) participants are unaware of their study treatment group; (2) the investigators are unaware of the participant’s study treatment group; and (3) the statistical analyses are conducted without knowledge of the groups’ study treatment. Multiple levels of blinding can occur in an RCT and should be considered when feasible. Recent oral health RCTs that followed the strict principles of clinical trials were two Phase III studies testing periodontal therapy as a treatment to prevent preterm birth¹⁷ and to improve glycemic control.¹⁸

A limitation of the RCT study design is the concern about external validity, or the extent to which RCT results are applicable beyond the research study. In addition, RCTs are expensive because of the logistics involved in sampling, blinding, treating, and following hundreds of participants; in addition, extremely large sample sizes are required to study rare outcomes. Consequently, some research questions may be more appropriately addressed using other study designs.

Systematic Reviews

A systematic review is a structured process of comprehensively reviewing published research studies focused on a research question in which inclusion and exclusion criteria for study selection are established a priori. The purpose of the systematic review is to determine the “state of the science” by objectively identifying, appraising, selecting, and synthesizing high‐quality research evidence. Such reviews may also elucidate a paucity of high‐quality evidence and, therefore, identify research questions to be addressed in future studies. For more information about systematic reviews, see Chapter 29: How to identify, interpret and apply the scientific literature to practice.

Study Design

Investigators should employ a study design that is suitable and most appropriate for answering the clinical research question of interest. In general, investigators should review the literature on the topic of interest, define the purpose of the study and hypothesis to be tested, and then use the strongest research design that is acceptable and feasible to address the research question.

Sample Size

When designing clinical research studies, an important consideration is the required number of study participants to draw meaningful statistical conclusions. The sample size depends upon the variability of the data and the effect size, or the difference between values.²⁵ Analyses should be conducted to determine the sample size needed to accomplish the goals of the study. For longitudinal studies, sample size calculations should take into account study participant attrition over time. Many clinical studies suffer from small sample sizes, and this issue is often a reason why studies are excluded from evidence‐based reviews.²⁶

Selection of Disease and Control Groups

Another critical factor to consider when designing a study is the definition and selection of the disease and control groups. The disease or case group should be carefully defined to include those individuals with the disease or condition of interest, but without other conditions or variables that may affect the validity of the study results.

Many clinical studies include a control group to compare study outcomes between the case or disease population and those who do not have the outcome of interest or do not receive the intervention being tested. For these studies, the control group should be clearly defined to include those individuals who are similar to the case or test group in many aspects so that there are minimal differing variables between the comparison groups.^8–10 Careful selection of control populations was described in the Case‐Control Study section above. To guide control group selection, studies may match cases and controls based upon variables that may contribute to the exposure or disease presentation or severity such as gender, age, current/past smoking history, and setting where participants may be recruited for study participation, such as a tertiary care clinic. Controls should be chosen in the same manner as cases to ensure they have the same likelihood of having the exposure of interest.^8–10 For example, if cases are patients with head and neck cancer being recruited from a radiation oncology clinic, controls should be patients recruited from the same clinic, who are undergoing treatment for another condition.

Potential for Bias

When designing and conducting a research study, there are numerous types of bias that must be considered to maintain the validity of study results. Bias can occur at any phase of research, including decisions related to the study design, implementation of study procedures, methods of data collection, the process of data analysis, and publication of study results.²⁷ Some bias is inevitable and inherent in certain study designs, and the investigator must show that efforts have been undertaken to lessen the impact of study bias.

Methods to avoid selection bias in observational studies include enrolling consecutive individuals reporting to a clinic who meet inclusion criteria, or recruiting individuals from an existing large population using consistent recruitment criteria for all individuals. RCTs use randomization and allocation procedures for treatment arm assignment to avoid bias.

Other biases can occur during data collection. The use of objective, validated measures for outcome assessment and independent data collectors will reduce these biases (see section Outcome Assessment below). Some biases can be addressed through data analyses, such as accounting for potential confounders and effect modifiers.

Outcome Assessment

Study outcomes or endpoints used in a clinical study or trial must be measurable, reliable (consistent and repeatable), and valid to document disease prevalence and/or progression or determine the efficacy of the intervention being tested. The outcome must be reproducible, and there should be published evidence of its validity. For example, if the goal of a study is to quantify oral cancer pain, the investigator should use a validated instrument to collect pain measures appropriate for the population being studied. In this example, an appropriate instrument would be a pain scale that had been tested previously in a population for whom cancer pain had been assessed and for whom the cultural values related to expression of pain had been similar.

Methods for ascertainment of study outcomes also need to be standardized. Examiners should be calibrated, by having them each examine the same group of patients to measure outcome assessment agreement with each other (inter‐rater reliability) and having them examine a set of patients repeatedly to measure their outcome assessment agreement with themselves (intra‐rater reliability).²⁸ Studies that assess caries and periodontal disease over time usually conduct calibration sessions annually or prior to a wave of study visits, during which examiners are calibrated to a gold standard examiner and compared numerically using percentage agreement or kappa scores.^17,29

Loss to Follow‐up and Retention