Project design

Next to choosing the subject for study, time spent carefully designing a project is never wasted. There are many different types of scientific study. The design used depends totally on the study. Beloved of present scientists, the randomised controlled trial is regarded as the best method of scientific research. It must not be forgotten that much surgical practice has been advanced by other different types of study such as those listed in Table 72.2. For example, testing a new type of operation often requires a pilot study to assess feasibility followed by a formal randomised controlled trial.

Research can be qualitative or quantitative. Quantitative research uses hard data to speak for themselves. A medical condition is analysed systematically using hard objective end­points such as death or amputation. In qualitative research data often come from patient narratives, and the psychosocial impact of the disease and its treatment are analysed, for example narratives of breast cancer. This sort of data is often collected using quality of life measurements. A variety of different quality of life questionnaires exists to suit several different clinical situations. Much of the best research is both quantitative and qualitative.

As finances for healthcare are always stretched, it is also important to include a cost—benefit analysis in any major area of research so that the value of the proposed intervention or change in treatment can be assessed.

Sample size

Calculating the number of patients required to perform a satisfactory investigation is a very important prerequisite to the study. An incorrect sample size is probably the most frequent reason for research to be invalid. So often surgical trials are marred by the possibility of error caused by the inadequate number of patients investigated.

Type I error — benefit is perceived when really there is none (false positive).

Type II error — benefit is missed because study has small numbers (false negative).

Calculating the number of patients required in the study can overcome this bias. Unfortunately, it very often reveals that a larger number of patients is needed for the study than can possibly be obtained from available resources. This often means expanding enrolment by using a multicentric study. There is no point in embarking on a trial when it will never be possible to recruit an adequate sample size. Never forget that more patients will need to be randomised than the final sample size to take into account patients who die, drop out or are lost to follow-up.

The following is an example calculation for a study to recruit patients into two groups. In order to calculate a sam­ple size it is common practice to set the level of power for the study at 80 per cent with a 5 per cent significance level. This means that if there is a difference between study groups, there is an 80 per cent chance of detecting it. Based on previous studies, realistic expectations of differences between groups should be used to calculate sample size. The formula below uses the figures of a reduction in event rate from 30 to 10 per cent (e.g. new treatment expected to reduce complication rate such as wound infection from 30 to 10 per cent):

8x r(100—r)+s(100—s)

(r~s)2

8x    30(100—30) + 10(100—10) = 60 needed in each group

                            (30~10)2

Eliminating bias

It is important to imagine how a study could be invalidated by thinking of things that could go wrong. One way to eliminate any bias inherent in the data collection is to have observers or recorders who do not know which treatment has been used (single blind). In the best randomised studies neither patient nor researcher is aware of which therapy has been used until after the study has finished (double blind). Randomised trials are essential for testing new drugs. In practice, however, in some surgical trials randomisation may not be possible or ethical.

Study protocol

Now that the question to pose has been decided and it has been checked that sufficient patients will be available to enroll into the study, it is time to prepare the detail of the trial. At this stage a study protocol should be constructed to define the research strategy. It should contain a paragraph on the background of the proposed study, the aim and objectives, a clear methodology, definitions of population and sample sizes, and methods of proposed analysis. It should include the patient numbers, inclusion and exclusion criteria, and the time scale for the work. At this stage it is helpful to construct a flow diagram giving a clear summary of the research protocol and its requirements (Fig. 72.1). It is helpful to imagine the paper that will be written about the study, before it is performed. This may prevent errors of data collection.

When a study is planned, sufficient time should be reserved at the beginning for fund-raising and obtaining ethical approval if required, and afterwards for collecting the data and writing it up. A data collection form should be designed or a computer collection package developed. Do not forget that if data are collected on computer, appropriate safeguards for privacy and confidentiality will be necessary. It is important to ensure the co-operation of any other specialities or clinicians who will be involved in the study and to agree on the sharing of responsibility for the trial. This will also help to prevent disagreement about who takes the credit once a study is ready for presentation and publication.

Ethics

Common sense is the best guide to whether a study is ethical or not. If there is any doubt consult a local ethics committee, available in all hospitals. Whenever a patient has treatment chosen by chance, ethical approval is required. All multicentric trials need approval. Even for nonrandomised trials it is important to try and obtain written consent for any new or extraordinary procedure or therapy. All studies involving animals require approval from statutory licensing authorities.

Leave plenty of time for this to be obtained and do not embark on the study until approval has been granted. Ethics committees prefer to see fully developed trial protocols but it is often possible to get some preliminary advice from the committee chairman. Ethics committee forms are often long and detailed, and it is important that these are filled in correctly. All dealings with ethics committees should be intelligent and courteous.

Most ethics committees are briefed to ensure that a project does not incur hidden expenses to a hospital. The cost of nonroutine investigations and extra treatments should be covered by a grant application, or at least underwritten by a hospital finance department.