Periodontal Research: Basics and beyond - Part II (Ethical issues, sampling, outcome measures and bias).

A good research beginning refers to formulating a well-defined research question, developing a hypothesis and choosing an appropriate study design. The first part of the review series has discussed these issues in depth and this paper intends to throw light on other issues pertaining to the implementation of research. These include the various ethical norms and standards in human experimentation, the eligibility criteria for the participants, sampling methods and sample size calculation, various outcome measures that need to be defined and the biases that can be introduced in research.

INTRODUCTION

In the first part of the review series, we have discussed on how to formulate a research question and the design of relevant studies. An important part of the research process is not only to formulate the question and choose the appropriate design but also decide how to assess and measure the phenomena you are aiming to research. Selecting the sample size and the outcome measures prior to the onset of the study is preferable to generate meaningful results. It is also important to adhere to certain ethical standards to protect the welfare of the participants. All these issues are discussed at length in the present paper to facilitate a researcher with an indispensable knowledge of the various complexities of research.

Ethical issues in research

The history of clinical research involving human subjects has witnessed several tragedies where many people were harmed mercilessly as a result of their unwitting participation in research. The brutal Nazi experiments on prisoners during the world war II, where human subjects were forcefully and ruthlessly tested to justify their suppositions was severely criticized and this inhuman attitude toward human participants prompted the evolution of various codes and reports. The Nuremberg code which was put forward in 1947 was a ten point statement which defined the conditions under which acceptable human experimentation should be carried out. The Tuskegee Syphilis Study, cited as “arguably the most infamous biomedical research study in U.S. history,”[1] where patients with syphilis were denied medical treatment, led to the formulation of the 1979 Belmont Report. This along, with the Helsinki declaration (originally adopted in June 1964 in Helsinki, Finland) of the World Medical Association, established certain ethical principles and guidelines to conduct research in human volunteers, employing good ethical standards. The Good Clinical Practice (GCP)[2] guidelines by the US Food and Drug Administration, 1996 regulates the proper conduct of trials in the United States.

In the era of evidence based medicine, there is an increasing focus on protecting the welfare of the subjects participating in clinical research. It is very appropriately stated by Silverman[3] in his article about ethical issues in clinical research that the conduct of research in human subjects does not connote just to the designing of the study and procuring the signature of the subject on the informed consent form. It also involves protecting the rights, interests, and safety of research subjects throughout the study duration. The United States regulation for the protection of human subjects provide baseline minimums with which everyone must comply in forming an Institutional Review Board (IRB), obtaining informed consent from research subjects, and conducting research in an ethical manner. Subject safety monitoring is the responsibility of several groups, including research ethics committees or Institutional Review Boards (IRBs), investigators and their research staffs, sponsors, and data monitoring committees, also called data and safety monitoring boards, especially in the United States.[4]

It is mandatory that all research involving human participants should be cleared by an appropriately constituted Institutional Ethics Committee, also referred to as Institutional Review Board (IRB), to safeguard the welfare and the rights of the participants. The IRB is not necessarily constituted by medical people alone but also has nonmedical people representing the community, ethicists, clergy or lawyers. Once the research protocol and informed consent forms are approved, the committee undertakes periodic reviews of the study with detailed reports, to ensure that the conditions are met in regard to the safety and wellbeing of participants.[4]

Informed consent should be obtained from every patient (or legal guardian, if necessary) and has to be included in the trial proposal. This implies that the patient is aware of and understands all the implications involved in the study which are known to the researchers, and is willing to accept these as a condition of his/her involvement in the study.[5] Participant confidentiality can be facilitated by using unique subject ID codes to identify participants, instead of their names. The coded list should be held in a separate spreadsheet or database from the study data. Any data stored on a computer should be password-protected and access limited to individuals involved in the data collection process.[6]

The GCP based on the international guidelines issued by World Health Organization and International Committee on Harmonization provide operative guidelines for ethical and scientific standards for the designing of a trial protocol including conduct, recording and reporting procedures and should be strictly adhered to while carrying out a trial.[7]

Apart from financing the study, the sponsors must demonstrate the safety of the intervention in pre-clinical studies, provide medical expertise, ensure that the trial design minimizes harms, participate in investigator selection, provide applications, notifications, and submissions to regulatory authorities, confirm IRB review, monitor to ensure Good Clinical Practice during the trial, supply and handle the investigational agent, maintain records and keep them available for inspection, perform safety evaluation and reporting, monitor and audit the data that have been collected during the study.[8]

In most epidemiological and community-based research (especially in India), it would be necessary to have the consent of the community, which can be done through the village leaders, the Panchayat head, the tribal leaders etc.[9] It is also important in an epidemiological study that the research is designed in such a manner that the privacy and the confidentiality of data are preserved and at the same time the community health needs are addressed.

Study setting and eligibility (inclusion/exclusion) criteria

The setting for a study refers to the population to which the results can be applied.[10] It usually establishes the boundaries of the study with respect to the sample, sampling area and the time period of the study. For example, the study setting may be limited to a rural area, an urban area or a private practice for a fixed period of time (1 month, 6 months, 10 years etc.) and it is generally understood that the investigator adheres to the study setting. Prior to the commencement of the study, the study population should be defined, along with the characteristics of the subjects that are to be included (inclusion criteria). For example, in most studies on periodontitis patients, there is a set criterion for the description of periodontitis. Although this is a matter of much variability and debate, the inclusion is based on standard approved criteria for the definition of periodontitis. Likewise, the characteristics of the subjects which are not to be included in the study (exclusion criteria) need to be established. For example some studies in periodontics exclude smokers, pregnant women, and patients with uncontrolled systemic disease etc.

SAMPLING

Sampling facilitates obtaining representative information about a population without having to investigate all the members of the population. A sampling frame, refers to a list of available population members, for example a list of patients, or the census data etc.[10]

A sample (n) is a “Finite” part of a statistical population whose properties are studied to gain information about the “WHOLE”. Sampling can be either probability sampling or non-probability sampling.

Sampling can be either purposive or non-purposive. A purposive (non-probability/deliberate) sample is one which is selected by the researcher subjectively. The researcher attempts to obtain the sample that appears to him/her to be representative of the population and will usually try to ensure that a range from one extreme to the other is included.

Convenience sampling

Is a type of purposive sampling where the population elements are selected for inclusion in the sample based on the ease of access.

Quota sampling

The population is first segmented into mutually exclusive sub-groups, just as in stratified sampling. Then judgment is used to select the subjects or units from each segment based on a specified proportion, which makes the technique one of non-probability sampling.

A non-purposive (probability) sample is one in which every unit in the population has a chance (greater than zero) of being selected in the sample, and this probability can be accurately determined. Types of probability sampling:

Simple Random Sampling

Stratified random Sampling

Systematic Random Sampling

Cluster Random sampling

Simple random sampling

This type of sampling is also known as chance sampling where each and every subject in the population has an equal chance of inclusion in the sample and each one of the possible samples has the same probability of being selected. Methods of simple random sampling include coin toss, random selection of numbers, lottery, and computer generation of numbers. For example, if we have to select a sample of 50 children to assess gingival pigmentation in a class of 100 children, then we can put the names or numbers of all the 100 on slips of paper and conduct a lottery or using the random number tables each child is assigned a number from 1 to 100. Then 50 random numbers are selected from the table.

Systematic sampling

This method of sampling uses a systematic method where the selection process starts by picking some random points in the list and then every nth element is selected until the desired number is secured. E.g. 5, 10, 15, 20 and so on.

Stratified sampling

Whenever the population from which a sample is to be drawn does not constitute a homogeneous group, then stratified sampling technique is applied so as to obtain a representative sample. In this technique, the population is stratified into a number of non-overlapping subpopulations or strata and sample items are selected from each stratum. If the item selection from each stratum is based on simple random sampling, then the entire procedure, first stratification followed by simple random sampling, is known as stratified random sampling. E.g. strata of age groups; gender (male/female) etc.

Cluster sampling and area sampling

Cluster sampling involves grouping the population and then selecting the groups or the clusters rather than individual elements for inclusion in the sample. When all the units within a cluster are selected, the technique is referred to as one-stage cluster sampling. If a subset of units is selected randomly from each selected cluster, it is called two-stage cluster sampling. The clusters form the primary sampling units (PSU's) and the units within the clusters are the secondary sampling units (SSU's).

Multi-stage sampling

This is an extension to the idea of cluster sampling. This technique is meant for larger data extending to a considerably large geographical area like an entire country. In multi-stage sampling the first stage may be to select large primary sampling units such as states, districts, towns, and finally certain families within towns.

Sample size estimation

The calculation of the size of the sample to be surveyed in a research project is pivotal as it would facilitate the researcher with some information on the timeframe, logistics of the project and also the future generalizability and relevance of the results. Estimation of sample size is not as simple as the turn of a wand by a magician (statistician) and the numbers are generated! The clinician first needs to understand certain important issues before approaching a statistician for sample size estimation. First of all what is the result of other studies which used a similar intervention? Or what is the range of the present level of or rate of occurrence of the event that we are interested in? Secondly, what is the level of α or β errors that we are willing to accept?[11] That is, to what extent our result could be due to chance? The description of α and β errors will be elaborated in the next part of this review series. Conventionally, 5% for type 1 and 20% for type II error is accepted.[11]

Third, what is the result we are expecting? That is, what is the clinically meaningful difference?

Based on these questions, the sample size is calculated using various formulae. In cases of clinical trials, an additional 10-15% can be recruited, taking into consideration the dropout rate of patients.

The observational unit (subject vs. site)

The unit of observation in an experiment or observational study is the smallest unit with a unique set of important characteristics which is independent of other similar units in that its response cannot be affected by these other units, and which can be assigned to each of the treatments in an experimental study. The choice of observational unit is a crucial player in periodontal research and there is a general confusion prevailing regarding the same.[12] This is primarily because the data obtained in periodontal research usually have variables measured at site level (pocket depth, furcation), tooth level (mobility) and subject level (systemic health, smoking status). In clinical trials, usually the subject or the mouth is selected as an observational unit. The study should be designed and analyzed in such a way that the randomization process should randomize the experimental units (the mouths) rather than sub-units (the teeth) to the different treatments. Previous studies in periodontics relied on site-based analyses but it is not appropriate to treat sites and teeth in the same patient as independent units as they share the same environmental conditions of immune response and systemic health.[13] Newer studies have addressed these issues with multi-level analyses to avoid clustering of data within patients.[13]

Examiner training and standardization/calibration

It is prudent that the clinical examiners who are assessing patients are well-trained and calibrated in order to achieve good reproducibility of data with minimal error and bias. Clinical examiners in a clinical trial should practice various calibration exercises to maintain the uniformity of assessment of the clinical outcome measures like various gingival and periodontal indices, pocket depths, and clinical attachment levels etc. Examiners are assessed for their reproducibility of measurements over various time points and then assessed statistically for their accuracy and consistency. Hefti and Preshaw[14] introduced the term “Examiner alignment and assessment” as a more descriptive term for the process of training examiners who will participate in a clinical trial.[14]

Examiner alignment aims to develop a mutual agreement among the examiners regarding the understanding of various clinical parameters. This is facilitated by various bench top and clinical demonstrations and practice sessions. Examiner assessment is aimed at intra- and inter-examiner reproducibility.[14]

BIAS AND CONFOUNDING

Bias is an “opinion or feeling that favors one side in an argument or one item in a group or series; predisposition; prejudice.”[15] In an epidemiological perspective, bias is present when the results from the study are systematically distorted and so are consistently above (or below) what they should be. Sackett[16] in 1979 identified 24 biases and many more have now been identified. A few major biases are listed as follows:

Selection bias

When the study participants are not representive of the population of interest. This bias usually happens during the selection of subjects for control and test groups where the investigator assigns subjects such that they differ with respect to extraneous factors.

Observer or measurement bias

When an examiner consistently over/under reports a variable (a characteristic). This must be resolved in training and calibration sessions.

Recall bias

An information bias in which the subjects with disease (cases) tend to recall past exposures better than controls.

Attrition bias

It happens when subjects quit the study before its completion. These drop outs in the clinical trial cause bias in the results due to the decreased sample size in one group or may be both, and also the decreased follow up time period.

Migration bias

Another kind occurs either when individuals drop out of study or move from one group to another.

Publication bias

Studies which show significant results are more likely to be published in journals than those with insignificant or negative results.

Allocation bias

When treatment groups in an experimental study are not comparable with respect to the variables influencing the response of interest.

Confounding is a mixing of the effect of an exposure with the effect of another variable that is associated with the exposure and is an independent risk factor for the disease.[17] Confounding factors are variables, which compete with the hypothesized risk factor as explanations for the observed response. For example, smoking and alcohol are together confounding factors in causing periodontitis. Smoking is a known risk factor for periodontitis. If a study is undertaken to associate between alcohol consumption and periodontitis and if the patients smoking status is not assessed in data analysis, it will appear as if alcohol is a strong risk factor for periodontitis when in fact it is not.

Contamination occurs when an intervention administered to an intervention group of an experimental study percolates into the control group.[18] This is usually seen in a split mouth study or in a crossover study where the intervention from one site filters into the control treatment. It could also be seen with respect to information which is given from one group by word of mouth to the members in the other group. For example, if a trial compares two brushing methods and when the participants are allowed to converse, one of the member leaks information about his brushing technique to a member from the other group and this recipient of information would unknowingly master the brushing technique of the other group. This might dull or mask the true effect of the intervention.

Hawthorne effect

It was first described by Roethlisberger and Dickson in 1939 in an industrial setting. The rationale of this is that in a research project, individual behaviors may be altered by the study itself, rather than the effects the study is researching. Subjects who are singled out to participate in a mouthwash trial consciously tend to improve their oral hygiene and may show lower plaque scores.

Rosenthal effect (Experimenter effect)

Research also demonstrates that the expectations and biases of an experimenter can be communicated to experimental subjects in unintentional ways and that these cues may significantly affect the outcome of the experiment.[19] For example, differences in oral hygiene instructions given to patients can affect the response of the subjects. Experimenter effects can be avoided by using double blinded studies where the experimenters’ expectations and biases are not communicated to the study subjects.

EVALUATION OF DIAGNOSTIC AND SCREENING TESTS

Reliability is the extent to which an experiment, test, or any measuring procedure yields the same result on repeated trials. To simplify, it measures the same thing in the same way on two separate occasions. An index is reliable when it yields the same score if a patient is examined by two different examiners (inter-examiner reliability) or by the same examiner on two different occasions (intra-examiner reliability).

Validity refers to whether a study is able to scientifically answer the questions, it is intended to answer. A test is valid if it measures what it is supposed to measure. A study that readily allows its findings to generalize to the population at large has high external validity. To the degree that we are successful in eliminating confounding variables within the study itself is referred to as internal validity. Internal validity has to do with the accuracy of the results whereas external validity has to do with the generalizability of the findings to the population.

A diagnostic test for a given condition would ideally be positive for every person in whom the condition was present and negative for those in whom it was absent. But some tests will report a false positive result on some individuals who are free of the condition and others may also miss some of those who actually have the disease giving a false negative result.

Sensitivity is the probability (usually expressed as a percentage) that a subject with the disease will have a positive test result.[20]

Specificity is the probability that a subject who is free of the disease will have a negative test result.[20]

The positive predictive value (PPV) is the probability (usually expressed as a percentage) that an individual who has a positive test result actually has the disease.[20]

It is the probability (usually expressed as a percentage) that an individual who has a positive test result actually has the disease.

The negative predictive value (NPV) is the probability that an individual who has a negative test result does not have the disease.[20]

The best tests are the ones that are good at detecting most of the people with the condition (high sensitivity) and at excluding people who do not have the condition (high specificity).[21]

OUTCOME MEASURES AND ENDPOINTS IN RESEARCH

Outcome measures can be divided into three broad categories:[22]

Laboratory measures

Clinical measures

Patient-based measures.

Laboratory tests include biological tests on cell cultures or mechanical tests on clinical materials. It is essential that any material to be tested clinically goes through a laboratory preclinical phase. A laboratory measure would include a study which investigates the role of platelet derived growth factor on periodontal fibroblasts.

Clinical outcome measures would include various indices used to measure gingival or periodontal parameters like modified gingival index, CPITN or assessment of bleeding on probing, measurement of pocket depths (PD) and clinical attachment level (CAL).

Patient based outcome measures highlight patient perceptions of the delivery and outcome of various healthcare systems. This measure will usually take the form of a questionnaire. A patient-based outcome measure would include the assessment of Oral Health Related Quality of Life (OHRQoL) in a group of periodontitis patients.

All RCTs assess response variables, or outcomes (end points), for which the groups are compared. Outcome measures can also be classified as primary or secondary. The primary outcome measure is the primary question that the study is designed to answer and the trial is powered to answer such a question. The primary outcome measure is the pre-specified outcome considered to be of greatest importance to the relevant research question and is usually the one used in the sample size calculation.[23] Other outcomes of interest are secondary outcomes (additional outcomes) which often include unanticipated or unintended effects of the intervention.

It is possible to specify more than one primary outcome in the design of the trial but inflated sample size may occur as a consequence of this.[24] A “composite” primary outcome can be used sometimes by combining several individual outcomes but they must be carefully developed and reported in a detailed manner.[25]

End points in periodontal clinical trials

A periodontal endpoint is any measurement that is thought to be related to the periodontal disease process and that is used to assess periodontal treatment efficacy.[26] Researchers are usually faced with the challenge of selecting an endpoint that provides clear, unequivocal evidence of a benefit to the patient.[27]

True endpoints are tangible outcomes that directly measure how a patient feels, functions or survives. True endpoints are sometimes referred to as clinically relevant endpoints, clinically meaningful endpoints, terminal endpoints, or ultimate endpoints.[26] True end points are usually subjective and examples include pain, bleeding on brushing, and oral health related quality of life etc.

Surrogate endpoints (intermediate endpoints, biological markers, biomarkers) are physiological or biochemical markers that are easier to measure than true endpoints. Temple[28] defined it as “a laboratory measurement or a physical sign used as a substitute for a clinically meaningful endpoint that measures directly how a patient feels, functions, or survives.” These end points are usually objective and intangible to the patients mind. Typical surrogate endpoints in periodontal research include anatomic measures (e.g. probing depth), measures of inflammation, microbiological measures, and immunologic measures.[26]

However, there is no evidence to suggest that any single outcome measure (e.g. change in clinical attachment level or radiographic bone fill) is more important or predictive of long-term success in periodontal therapy. To assess the cumulative beneficial effects of treatment modalities the use of composite endpoints have been put forward[27] where they are traditionally expressed as binary outcomes classified as success or failure at the patient level. Composite endpoints combine the clinical parameters CAL with radiographic parameters (percent bone fill [%BF] or linear bone gain), offering a potential benefit over single-outcome measures for determining clinically meaningful effects on regenerative procedures.

CONCLUSION

The present paper introduces to the readers, various pivotal issues in the implementation of sound and ethical research. It is prudent to take into consideration, various issues in the design of research studies, especially ones involving human experimentation such that the results are more valid, ethical, and meaningful.