Setting Up a Research Project

Selecting a Research Topic

Finding the solution to a problem is often relatively simple compared with finding the right question. Since resources are limited, it is important to find the topic best suited to one's capabilities and, most importantly, which addresses a worthwhile problem. One recommendation is to consider the value of a given research question using the mnemonic FINER which indicates it should be judged on whether it is Feasible, Interesting, Novel, Ethical and Relevant ( ).

What areas are most suited to research by the clinician? In general, those questions that relate closely to everyday practice are more easily approached than basic scientific questions with no obvious clinical relevance. These can arise from the practitioner's own experience and are often prompted by questions and problems arising in the course of normal clinical practice.

• ▪

  Does product A, for example, give better results than product B?

• ▪

  Is a given category of patient more likely to exhibit a certain characteristic? The category might relate to age, sex or ametropia, while the characteristic might be a sign, symptom, type of disorder or response to a given treatment.

• ▪

  Does a certain new technique have advantages over a more conventional method?

• ▪

  What are the success rates and limitations of a given product?

Defining the Objectives

It is fruitless and wasteful simply to take an aimless ‘look’ at a problem. A clear statement of the reasons for undertaking the study and of the goals the study hopes to achieve will help direct the work at every stage. It will help in designing the experiment and in deciding how the results of the experiment should be analysed. The main conclusions of the study will answer the questions posed by the objectives.

Defining the aims also helps in assessing the feasibility and usefulness of a study. Are the objectives, for instance, too broad and too varied? For example, where the purpose is ‘to determine whether contact lens A is better than contact lens B’, it may be unrealistic to attempt a broad assessment; therefore it may prove necessary to narrow the objectives to selected aspects of lens performance, such as visual performance. Where there is more than one objective, a potential conflict may become apparent. It may not be feasible, for example, to consider contact lens deposit resistance independently from physiological effects, as the two could be related.

With some studies, it is appropriate to restate the research question as a hypothesis. In other words, the question is phrased as a simple but specific statement summarising the theory under test, e.g. contact lens A provides better high-contrast visual acuity (VA) than contact lens B. For the purposes of statistical testing, the hypothesis is often restated as a null hypothesis, e.g. high-contrast VA is no better with contact lens A than with contact lens B. By assuming no difference between data sets, the statistical test estimates the probability of the result happening by chance. If the statistics fail to confirm the null hypothesis then the alternative hypothesis is accepted: in this case, that there is a difference in VA between the two lens types. Remember that scientific theories are not provable, only falsifiable.

Researching the Literature

By not properly understanding the background to a problem, there is a risk of any research being wasted through poor methodology or inappropriate study design. A survey of previous work helps to confirm that the proposed study is worthwhile as well as to reveal gaps in current knowledge. If the subject has already been researched, a review of the literature may suggest alternative ways of approaching the subject or may point to additional areas of research. It is usually helpful in developing the experimental method and in selecting the techniques to use. If the subject turns out already to have been researched, this does not necessarily invalidate any further work, as there may have been deficiencies in the previous work. A study replicating the original results serves to add weight to the original findings, but of course there is the possibility that the new study may produce entirely different results from the original.

An internet literature search is an obvious good starting point. The medical database PubMed ( ) allows online searches of refereed journals based on keywords, authors, etc. Sometimes the downloaded abstract provides enough useful information, but it is usually necessary to see the full paper to have a good understanding of current knowledge and to derive the full benefit of previous work. Many refereed and non-refereed journals now allow open access to their archives through their websites. Other sites that are useful for obtaining papers include , , Web of Science and Google Scholar. For more obscure papers, copies can be requested from one of the authors or from a university, institution or national library.

Designing the Study

Having decided the objectives of the study and reviewed how other people have tackled similar projects, the researcher will have a rough plan of how the study should proceed. At an early stage in the project, it is invaluable to draft a summary of the key aspects of the study:

• ▪

  objectives

• ▪

  study design

• ▪

  study products

• ▪

  number of subjects

• ▪

  entry criteria

• ▪

  key variables, etc.

• ▪

  visit schedule.

From this, a more detailed protocol can be prepared.

A wide range of ingenious study designs can be used in clinical research ( ). Cohort studies, for instance, are relatively expensive but avoid selection and survival bias; in other words, the comparison groups are likely to be similar in composition, and any extraneous factors affecting completion of the study are similar for the various groups. The selection of study design, therefore, has to be matched with aims of the study and the questions posed. Most studies follow one of three fundamental designs ( Table 32.1 ) and these have varying strengths and weaknesses:

• 1.

  A case-control study examines a given problem by comparing those who exhibit the problem with those who do not. This type of study helps determine the risk factors associated with that problem – for instance, the habits and characteristics that predispose some patients to certain problems.

• 2.

  A cross-sectional study involves measuring a wide range of variables at a given time and is particularly useful for measuring prevalence. The prevalence of a given problem is the proportion of patients who exhibit the problem at a given point in time, whereas the incidence is the proportion who will exhibit the problem over a period of time.

• 3.

  A cohort study, the largest and most complex type, involves assigning a given treatment (e.g. contact lens type) to a group of patients and then monitoring the outcomes over a period of time. This type of study may or may not include a control group and may be undertaken prospectively or retrospectively.

Taking corneal desiccation staining as an example, a retrospective cross-sectional study might be used to estimate the prevalence of the problem. This could be an evaluation of the records of all soft lens patients attending a first follow-up visit. Causative factors could be determined using a case-control design comparing a group of patients showing desiccation staining with a control group who show no corneal staining. If previous work has suggested a method for avoiding desiccation staining, a cohort study might be undertaken to test the hypothesis that this results in a lower incidence of staining than other methods.

A further categorisation of clinical studies is whether they are:

• ▪

  Observational or non-interventional – a study which does not modify the treatment that patients would normally receive.

• ▪

  Interventional or experimental – a study that modifies some part of the treatment. These studies usually involve a control against which the treatment under test can be compared.

In some ways, the use of controls is easier in contact lens research than in other areas of medical research.

• ▪

  Cross-over designs, in which test and control products are used in succession, are less likely to result in carry-over effects than with pharmaceutical trials.

• ▪

  Contralateral designs, in which different products are used simultaneously in opposite eyes, may be the most efficient method of comparing two contact lens designs or care products. However, there is some evidence of a sympathetic contralateral physiological response ( ), and a contralateral design may not be appropriate with some studies comparing physiology.

  A component of study design that requires careful consideration is the number of subjects to be included. The main risk from using an inadequate sample size is failing to detect a significant difference where one exists; this is known as a Type II error (false-negative). Formulae exist for estimating the sample size required to demonstrate a statistically significant difference ( ). These require, firstly, a determination of what constitutes a clinically significant size difference and, secondly, an estimation of the variance of the final results. Because contact lens studies often examine a large range of variables and estimates of the outcomes are often unavailable, sample sizes are frequently based on judgement rather than sample size calculations.