• Users Online: 946
  • Home
  • Print this page
  • Email this page
Home About us Editorial board Ahead of print Current issue Search Archives Submit article Instructions Subscribe Contacts Login 

 Table of Contents  
Year : 2022  |  Volume : 22  |  Issue : 4  |  Page : 312-313

Controlling bias in research

Department of Prosthodontics, Faculty of Dental Sciences, Sri Ramachandra Institute of Higher Education and Research, Chennai, Tamil Nadu, India

Date of Submission29-Aug-2022
Date of Decision15-Sep-2022
Date of Acceptance22-Sep-2022
Date of Web Publication03-Oct-2022

Correspondence Address:
Anand Kumar Vaidyanathan
Department of Prosthodontics, Faculty of Dental Sciences, Sri Ramachandra Institute of Higher Education and Research, Chennai, Tamil Nadu
Login to access the Email id

Source of Support: None, Conflict of Interest: None

DOI: 10.4103/jips.jips_405_22

Rights and Permissions

How to cite this article:
Vaidyanathan AK. Controlling bias in research. J Indian Prosthodont Soc 2022;22:312-3

How to cite this URL:
Vaidyanathan AK. Controlling bias in research. J Indian Prosthodont Soc [serial online] 2022 [cited 2022 Nov 26];22:312-3. Available from: https://www.j-ips.org/text.asp?2022/22/4/312/357798

Bias in clinical research depart the results systematically from the true values due to lack in standardization of protocol.[1] The high demand for publishing articles in the field of academics and acceptance of research that has positive results by the journals has made the researcher to hasten their research, and focus on desirable study outcome. As a consequence, errors in research are becoming inevitable that could be either systematic or random. A random error could be prevented by increasing the sample size, but Bias, a form of systematic error, is difficult to control as multiple factors are involved.[2] Bias can occur in the planning, data collection, analysis, and publication phases of research. Understanding research bias will help in controlling error in research and avoid suboptimal or potentially harmful treatments rendered to the patients/participants.

Bias differs according to the study design and the error increases when an inappropriate study design is selected for the research hypothesis. The research question that leads to a prospective study design is always better than a retrospective study, unless the research involves a rare disease or condition that requires a proof of concept retrospectively so that a long term research could be planned. Similarly, a randomized controlled trial with a standardized protocol has reduced bias than an observational study design. However, large volume of observational research are being conducted especially when a research is required in short-term for completion of an academic target especially with the graduates.

Among the observational study design, a Cohort conducted in prospective manner has better control of bias than the other observational study. Prospective study design begins from an exposure, or a disease or treatment followed up for a specific period; while a cohort can be retrospective when the disease or failure event in a specified cluster is followed back for the presence or absence of exposure. Another method of retrospective study that is undertaken to identify the cause or exposure to a disease is considered as a Case control study. Researcher often confuse with a retrospective cohort and a case-control study. To differentiate between a retrospective cohort and case-control; A retrospective cohort study identifies groups based on the intervention while a case-control study identifies the groups based on their outcome,[3] e.g., occurrence of failure with xenograft around the implant will be a Case control, whereas the effect of xenograft in implant treatment is Cohort. A retrospective design has the probability of high bias due to missing data collected from the patient.[4] Similar to case-control, a cross-sectional observational study also has the disadvantage of missing data. This type of study design is often used to evaluate the prevalence of an event through questionnaires and/or analyse the treatment and outcome at a single point of time. The researcher does not go back or follows an event, but defines the state of event during the specified time.

In contrast to the observational studies that only observe an event, experimental studies (clinical trials) tests a hypothesis. The occurrence of error is possible even with the clinical trial, but comparatively less due to equal distribution of compromising factor in both the control and the test groups due to randomization of population.[5] With the increased submission of research article to the journals, the authors should understand that clinical trial gets more weightage than an observational study design.

At preliminary phase, in search of the cause of a disease, the researcher recruits more exposed (test) than the unexposed (control) leading to an incorrect Measure of association. The selection bias due to the missing data of the patient related information can occur, especially in a retrospective study when data is collected from registries. Inappropriate definition of the eligible population, uneven diagnostic procedure, inaccurate sampling frame are few other reasons for selection bias. Selection bias occurs with the knowledge of the researcher and hence, blinding of patient recruitment is very essential to prevent bias. Allocation concealment is an essential aspect of randomized controlled trials that can avoid selection bias.[6]

Confounding bias can happen without the knowledge of the researcher; a hidden factor that is not considered when including a participant in a group,[7] e.g., to identify smoking as a cause of implant failure, participants included based on smokers and nonsmokers, and we often fail to consider the other risk factors like osteoporosis, diabetes etc., in inclusion criteria. There can be more chance of these confounding factors present in the test participants than the controls. In contrast, the clinical trial has an equal chance of these confounding factors distributed in both the test and control groups due to randomization of the sample, thus preventing incorrect association. A stringent inclusion and exclusion criteria's will help in choosing a homogenous sample and the right comparison group. The collection of data of all possible exposure or risk factors for the occurrence of a disease, a prolonged follow up is essential in preventing the occurrence of selection bias. Hence, a prospective study is always better when compared to retrospective study design to avoid missing data. Though selection bias cannot be avoided with observational study; the possible shortcomings should be mentioned as limitations, that would enable the readers to formulate a new research question to test a hypothesis.

During the course of the study, there are possibilities of patient not reporting back or does not prefer to answer a specific question. This nonresponse bias could be converted into an information by obtaining the demographic details of nonrespondent. In coherence with nonresponse bias, attrition bias occurs because of loss of participant due to complication of outcome. Comparing the demographic details of both the respondents and nonrespondents/lost participants, we can convert these biases into a specific demographic information that identifies the reason for nonresponse or loss of participant in the study.

Also, during the course of the study, an information bias can occur with open-ended or ill-defined pre and postoperative questionnaires and the conveying capacity of the investigator with interview based questionnaires to achieve their required outcome. Secondly, the collected information from a register or records can lead to information bias. Moreover, it is essential to use standardized questionnaires to avoid information bias. With the pandemic, the researchers started to use self-made questionnaire that is circulated through google forms. These questionnaires need to be validated using a sub-population of the main group, and later can be circulated for a research purpose. However, we commonly fail to validate the questionnaire and an ethical clearance is mandatory for a questionnaire study. When the patient's reported data are used, the trial design should mask the intent of the question in the structured interview and should use the validated scales for data acquisition. Similarly, a self-administered questionnaire with clear instruction is better than an interview in reducing the information bias. But, the possibility of low response rate is greater and can be managed as discussed in nonresponse bias. Other than the questionnaire study, the information bias also occurs in other observational studies and clinical trials due to nonstandardized equipment. This could be managed by use of standard measurement devices.

In an observer bias, the clinician sees only positive aspect of the test group because he is already aware of the participant group. Also, the clinician may perform the clinical procedure for the test group better than the control group. Blinding, either single, double or triple depending on the research question can reduce most of the biases. Participants also need to be blinded of the group to which they belong, to reduce the performance bias.

Detection bias, measurement bias and instrument bias due to a nonstandardized equipment and lack of training to investigator are few other types that occur in a quantitative analysis.[8] Recall bias occurs when both experiment and disease status are known at the time of study, and the clinician recalls the test (experiment) group more than the control group. A standardized protocol, training of the investigators, blinding, standardized instruments and a control that reveals standard measurement can reduce bias during analysis stage.

The study guidelines Strengthening the Reporting of Observational studies in Epidemiology for observational, Consolidated Standards of Reporting Trials for clinical trial and Consensus-based Clinical Case Reporting Guideline Development) for case report need to be used as a guide as the researcher formulates the research question. This would minimize errors that could occur due to investigators negligence.

Finally, the author should avoid publishing only selective results and hiding the negative aspects during the course of design to avoid publication bias. Also, the editor should avoid selecting publications based on affiliation of the authors causing publication bias.

  References Top

Sica GT. Bias in research studies. Radiology 2006;238:780-9.  Back to cited text no. 1
Systematic versus Random Error – Differences and Examples this Entry was Posted on 26 June 2021 by Anne Helmenstine Science notes. Available from: Available from: https://sciencenotes.org/systematic-vs-random-error-differences-and-examples/. [Last updated on 2021 Jun 28].  Back to cited text no. 2
Dettori JR, Norvell DC, Chapman JR. Grab control! Choosing the right comparison group in an observational study. Global Spine J 2019;9:456-8.  Back to cited text no. 3
Hennekens CH, Buring JE. Epidemiology in Medicine. Boston: Little, Brown, and Company; 1987.  Back to cited text no. 4
Suresh K. An overview of randomization techniques: An unbiased assessment of outcome in clinical research. J Hum Reprod Sci 2011;4:8-11.  Back to cited text no. 5
[PUBMED]  [Full text]  
Moher D, Fortin P, Jadad AR, Jüni P, Klassen T, Le Lorier J, et al. Completeness of reporting of trials published in languages other than English: Implications for conduct and reporting of systematic reviews. Lancet 1996;347:363-6.  Back to cited text no. 6
Skelly AC, Dettori JR, Brodt ED. Assessing bias: The importance of considering confounding. Evid Based Spine Care J 2012;3:9-12.  Back to cited text no. 7
Hammer GP, du Prel JB, Blettner M. Avoiding bias in observational studies: Part 8 in a series of articles on evaluation of scientific publications. Dtsch Arztebl Int 2009;106:664-8.  Back to cited text no. 8


Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
Access Statistics
Email Alert *
Add to My List *
* Registration required (free)

  In this article

 Article Access Statistics
    PDF Downloaded128    
    Comments [Add]    

Recommend this journal