Dr M N Parikh
Mumbai
Past President, FOGSI, Past Editor, JOGI Incharge
Email : parikh57@gmail.com
Research, Research, Research
Fundamental of Clinical Research by Dr. Mahendra N. Parikh
In charge – PICSEP Project of JOGI
Author of ABC of Research Methodology and Applied Biostatistics : A Primer for Clinicians and Researchers
AND Research Methodology Simplified : Every Clinician a Researcher
Introduction
When man evolved on this earth endowed with bipedalism his biggest asset was a larger brain capable of complex thinking. He was confronted by hostile creatures with great muscle power residing on the planet. He subdued and tamed them. Simultaneously he gradually changed his natural surroundings, developed civilization and manipulated nature to make his life more comfortable. All this was made possible by developing new ideas and improving the ways of doing things. Cultivating crops, producing fire, inventing the wheel are just a few examples. This process is never ending and man’s craze for finding out ways to make living more and more comfortable is eternal. Understanding disease process, preventing sickness, treating illness and promoting health are important aspects of human endeavour to improve his quality of life. This is made possible by research in medicine.
What is research?
According to Oxford dictionary research is careful search or inquiry; endeavuor to discover and collate new facts. Some other definitions of research are 1) careful or diligent search 2) studious inquiry or examination aimed at the discovery and interpretation of new knowledge 3) the collecting of information about a particular subject and 4) scientific and diligent study, investigation or experimentation in order to establish facts and analyze their significance. It is obvious that precise universally accepted definition of research is elusive.
Progress in every human activity is based on research. Health being the essence of good quality of life research in medicine is of vital importance. WHO defines clinical trial as any research that prospectively assigns human participants or groups of humans to one or more health related interventions to evaluate the effects on health outcomes.
Research question and clinical research
One who wants to do a research should at first define the aim or purpose of his research.What does he want to find out ? What new knowledge he desires to acquire? The research question can be as simple as what is the height of 10 year old boys to as complex as what is the molecular/genetic basis of primary carcinoma of the fallopian tube. This brings us to the fact that medical research can be clarified into basic or pure research and clinical research. Basic research aims at acquiring fundamental knowledge without focusing on its application in day-to-day activity. Clinical research or applied research generally deals with a problem relevant to clinical practice. The goal of clinical research is the application/utility of its results in patient care. Since the present write-ups are meant for FOGSI members who are clinicians in the first place I will deal with clinical research and that too in simple language consistent with space constraint which does not permit detailed description and discussion.
Types or classification of clinical research
There are many ways of classifying clinical research. I will describe one of the commonly used classification (Fig. I.1) Clinical research studies are grossly divided into observational studies and interventional studies.
Observational studies
In observational studies events occurring in a group of study subjects are observed and recorded. They may be compared with events occuring in another relevant group of study subjects. One observes the occurrence of congenital malformations in women having rubella infection in first trimester of pregnancy. In addition one may compare this occurrence with similar occurrence observed in another group of subjects who do not have rubella infection in first trimester of pregnancy. One merely observes events without making any effort to
Fig. I..1 – Types of Research Studies
Fig I.2 Cohort study
Fig I.3 Case Control Study
change the occurrence of events. Another example of observational study is to note the incidence of gastrointestinal infection in people using tap water provided by municipal authorities and compare the incidence with that in people who drink only mineral water. On the basis of the findings of observational studies one can formulate a hypothesis regarding the causation or etiology of a disease eg. causation of congenital malformation by rubella infection and of gastrointestinal infection by drinking suspicious tap water. Such hypothesis would have to be proved by other types of studies designed for the purpose eg. bacterial count studies on the tapwater and suitably treating tap water to rid it of bacteria. The two studies described above are called analytical studies because therein two groups of study subjects are studied and compared. When you study only one group of subjects (eg. pregnant women having rubella infection in the first trimester or people drinking tap water) to observe the occurrence of events the study is called descriptive study – it only describes the occurrence or prevalence/events without comparing them with those in another group of study subjects and without trying to find the etiology of the event. Incidentally case report and case series also constitute descriptive observational studies.
Fig. I. 4 – Cross sectional study
Fig. I.5 Analytical studies
Analytical studies
In analytical studies two groups of subjects are observed for occurrence of events. The occurrence in the two groups are recorded and compared or analyzed. Such studies constitute observational analytical studies. Basically there are three types of analytical studies viz. I) Cohort studies ii) Case control studies and iii) Cross sectional studies. It is possible to combine two of these studies in a hybridized study like case – cohort studies or nested case control studies (combine features of cohort and case control studies) which will not be discussed here for want of space.
One of the purposes of conducting research is to find out causation of disease. We can study subjects who are exposed (not by the investigator) to a suspected causative factor and observe them for development of a disease. Alternatively we can study subjects having the disease to find out about their past exposure to the suspected causative factor. Such exposure may occur due to nature eg. lack of iodine in the natural source of water leading to hypothyroidism or due to the lifestyle of the study subjects eg. smoking causing lung cancer or drinking alcohol causing cirrhosis of liver.
Cohort study (Prospective, forward looking or longitudinal study)
In Roman days soldiers would march in a group of 460-500 called a cohort. A group of study subjects is now called a cohort. Study of a group of subjects exposed to a suspected causative factor to look for development of the relevant disease constitutes a cohort study.
Since the investigator moves from exposure to disease the study is also called prospective or forward looking or longitudinal study. We select two groups of people eg. one smoking and one nonsmoking or one taking hormone replacement therapy and one not taking or one drinking and one nondrinking and follow them over years to look out for development of lung cancer or osteoporsis or liver cirrhosis respectively. Incidentally the two groups must be well matched to start with and the occurrence of the resultant disease compared statistically. Such studies can be designed to assess the etiology of a disease, natural history of the disease, incidence rates etc. These studies take a long time to complete and hence are costly and result in larger numbers dropping out from the study. In a cohort study the study outcomes in the two groups are statistically compared by risk ratio (RR) (Fig. I.2)
Case control study (retrospective or backward looking study)
This can be considered the reverse of cohort study. Here you select two groups of people one having and one not having the disease under study like lung cancer or liver cirrhosis or osteoposis. You now go backward in time to find out whether they were exposed to smoking, drinking or hormone replacement therapy respectively. This you find out by questioning the study subjects and rely on their memory recall and studying history recorded during previous hospital admissions of the subjects for any reason including road accidents. This study is done in opposite direction to that of cohort study and is called case control study or backward looking study or retrospective study. Both cohort and case control studies establish an association between exposure and outcome. But case control studies are of short duration, easy to conduct and less costly. Hence for diseases that take a long time to develop case control study is prefered over a cohort study. The drawback of case control study is recall bias of the subjects since one who has developed the disease will recall more accurately the details about the exposure. In a case control study the study outcomes in the two groups are statistically compared by odds ratio (OR) and not by risk ratio (RR)
Cross sectional study
In a cross sectional study the exposure to suspect causative factor and the outcome are studied concurrently. To find the relationship between maternal hypertension and low birth weight babies a number of deliveries are selected irrespective of maternal hypertension or low birth weight babies and one looks for both maternal hypertension and low birth weight in them. A cross sectional study is like a snap shot or a still film while cohort study and case control study are like a video or a movie film. Cross sectional studies cost less, can be completed in a short time, provide total requisite information and have no drop outs or attrition. It provides evidence of relationship between the two events (maternal hypertension and low birth weight) studied to generate a hypothesis. (Fig. I.4)
Fig. I. 5 Summarizes analytical studies
Experimental or Interventional studies
In observational studies the investigator merely observes the course of events. In interventional studies the investigator intervenes by doing something and observes the effect of such intervention on the course of events. In a observational study on anaemia during pregnancy the researcher starts with anaemic women in their first trimester of pregnancy and observes various events as they occur in her pregnancy, labour and postpartum period. He may decide to observe events like abortion, development of symptoms (marked weakness, oedema, breathlessness etc) due to anaemia, fetal growth, fetal birth weight, growth retardation, preterm labour, prolonged labour, postpartum hemorrhage, puerperal infection, lactational problems etc. In an interventional study he may intervene by treating anaemia in another group of anaemic women in their first trimester of pregnancy and observe the same events to study the effect of the intervention. There are various types of interventions that can be studied. Some of them are – administration of drugs (for anaemia, hypertension), giving prophylactic vaccination (rubella, hepatitis), conducting diagnostic tests (for blood sugar, thyroid function), carrying out a surgical procedure (resection of uterine septum, dilatation & curettage), counseling (diet, exercises, yoga), lifestyle changes (giving up smoking and drinking) etc. The purpose of medical research is to use the outcome of research to improve health care by developing preventive measures, achieving early quick precise diagnosis and finding out simpler less invasive patient friendly more effective and less costly treatment. This can be achieved by comparing the effects of new interventions with those of relevant current interventions. This is possible only by undertaking experimental or interventional research.
Randomised controlled blinded trials
Randomised controlled trials (RCTs) are considered the gold standard in clinical research. Blinding is highly desirable during RCTs.
Importance of controls
The researchers are eternally struggling to find better ways of preventing, diagnosing and treating diseases. When one says better it is always in comparison with something. If you see a beautiful girl pass by you say ‘Oh! She is so very beautiful’. When she is followed by a more beautiful girl you say ‘Woh! This girl is more beautiful’. The designation ‘more beautiful’ is the result of comparing her beauty with that of the previous girl who acts as a comparator. In the same way to find out whether a new intervention being studied is better than the existing one the two must be compared. The existing intervention which acts as a comparator is designated as a control. During research two similar groups of subjects are selected. One group is administered the new intervention under study and is designated as study group. The other group is administered the current best intervention (usually called the standard of care) and acts as a comparator group designated as control group. The effects of the intervention in the two groups are compared to find out whether the new intervention being studied is better or superior to the currently practiced intervention. Any research study aimed at finding a better preventive measure (live vaccine vs. toxin/toxide, monovalent vs polyvalent vaccine), better diagnostic or screening test and better treatment modality, whether medical or surgical, must have a group of subjects as a control group. Incidentally as will be shown later while studying a new diagnostic test the same study group is subjected to both the new test as well as the existing gold standard test. There are many diseases (like rabies, AIDS, progressive muscular dystrophy) for which there is no curative treatment. When a researcher is evaluating a new treatment claiming cure for these does he need a control group? If so where does he find it ? Yes, he does need a control group. Why does he need it? Some conditions (eg. viral infections) are self limited and many patients affected by others show spontaneous cure in due course of time (eg. inflammations or injuries). Any treatment, even if ineffective may show some cure rates. In fact a patient receiving new treatment, although ineffective, may show improvement by virtue of the psychological effect of receiving some treatment or a new treatment. This is particularly true of chronic conditions having exacerbations and remissions. In situations like these, any cure or benefit is likely to be attributed to the treatment under study even if the treatment is ineffective. To overcome this situation the researcher selects a control group to whom he administers an ineffective medication having no effect on the disease under study. Such ineffective medication is called ‘Placebo’ which must be exactly similar to and indistinguishable from the drug under trial in all respects like colour, taste, size, weight route of administration etc. Placebo is an inactive substance masquerading as a drug in the form of tablet, capsule, suppository, injection etc. There are few exceptional situations where we do not need a control group. If a new treatment giving hopes of cure is to be tried for conditions with extremely poor prognosis (eg. rabies & AIDs) it is immoral and unethical to deprive a group of patients a possibly useful treatment in order to use them as controls. Similarly, trial of a new treatment promising dramatic success in diseases having extremely high mortality can do without a control group. When penicillin was tried for pneumococeal pnenmonia in the 1940s control group had no justification in view of the then very high mortality of pneumoccal pneumonia. However, control group must be used in trials of new interventions for short duration diseases having current effective treatment and for diseases with spontaneous remissions. Sometimes instead of studying a control group, a group of patients previously treated by current modality can be used as a matching or similar control group by retrieving necessary data from hospital records. This is designated as historical control. This can be used only if there are no suitable changes in factors like diagnostic or assessment criteria over the period of time elapsing after the data of the controls were recorded.
Randomisation – why and how?
When a new intervention is being studied the study group and the control group must be as similar as possible. In other words in a controlled study all study subjects must be similar irrespective of their belonging to study group or control group. The two groups should be well matched. They should differ only in the type of intervention they receive. This is possible only if uniform criteria for recruitment and for inclusion and exclusion in the study are defined and followed very meticulously with no exception. This can take care of all known factors influencing the outcome of the study. Yet there could be some unknown factors which could affect the outcome. Being unknown they would be ignored during matching. This possibility is nullified by giving each and every participant an equal chance to get allotted to either study group or control group. This process by ensuring that a participant enters one of the two groups by chance and not by investigators choice is called randomisation.
Randomisation can be done by tossing a coin. This requires uniformity in the process every time the coin is tossed. Besides it is tedious to do so when the number of participants is large. Lastly, as the allotment is to be done just before the administration of the intervention, coin tossing in the operation thereafter after anaesthetizing the patient is not feasible. It is better to use random number tables which are freely available. As an alternative one can use computer generated numbers. Randomisation can be done at the stage of planning the study but random allocation of a study participant must be concealed till the start of administration of the intervention. It is mandatory that those sponsoring the study, those recruiting the study subjects and those conducting the study are not involved in the process of randomization and allocation .
All short cuts to randomization and use of simple methods like odd and even number of age in years, birth date, recruitment number, outpatient case paper number or indoor admission bed number are illegitimate and invalid since allocation to the group based on these is compliant to the wishes of any member of the investigating team.
Blinding or Masking
When a new treatment is under trial everybody involved in the study including the sponsors, investigators, study subjects, those administering treatment, evaluators of the treatment, statisticians and ancillary staff knowingly or unknowingly feel and consciously or unconsciously believe that the new treatment is better than the older one that is being compared with. In other words they are biased in favour of the new treatment or intervention. This bias is likely to affect the outcome of the study and hence needs to be nullified. This is achieved by hiding from the people concerned in various stages of the study the nature of the treatment or intervention received by individual participants. Such hiding is called blinding or masking. When only the participant is so blinded it is called single blind study. When the investigator/clinician is also blinded along with the participant the study is called double blind study. When in addition those doing follow up evaluation and the statistician are also blinded it is called a triple blind study. Though unblinded trials are simple in design blinded trials carry greater credibility and hence are far more desirable.
Diagnostic tests
On the basis of clinical evaluation and his past experience the clinician short lists a few conditions as probable diagnosis. This we call differential diagnosis. Now he submits his patient to diagnostic tests to arrive at the diagnosis. When a new diagnostic test is developed it is evaluated by conducting a study. The study involves subjecting the patients suspected of having the disease simultaneously to the new diagnostic test as well as the current gold standard diagnostic test. The results of the two tests are compared by statistical evaluation to assess the accuracy and utility of the new test. Two groups of patients are not used but instead the gold standard test acts as a control against which the new test is evaluated. The statistical methods used for this evaluation are very simple and easy to understand although at first sight they appear complex.
Multicentric trials
Many trials need a large sample size and some need to be conducted quickly to benefit needy patients as is the case when a promising effective remedy is under study for a condition having no current treatment. In both these situations conducting a trial concurrently at multiple centers spread over nationally or even internationally becomes handy. We are familiar with eclampsia trial and hormone replacement studies. It is necessary that all the participating centers strictly follow the protocol of the study designed by the sponsoring agency which has the responsibility of collecting, pooling and analyzing the data received from the various centers.
Other types of studies
Space does not permit describing other research studies like large simple trials, research by questionnaires, epidemiological studies, surveillance studies and phase 0-IV studies.