• Users Online: 33
  • Print this page
  • Email this page

Table of Contents
Year : 2020  |  Volume : 3  |  Issue : 1  |  Page : 100-109

Basics of statistics – 2: Types of clinical studies

Department of Medical Oncology, Narayana Superspeciality Hospital, Gurugram, Haryana, India

Date of Submission07-Jan-2020
Date of Decision27-Jan-2020
Date of Acceptance01-Feb-2020
Date of Web Publication24-Feb-2020

Correspondence Address:
H S Darling
Narayana Superspeciality Hospital, Gurugram - 122 002, Haryana
Login to access the Email id

Source of Support: None, Conflict of Interest: None

DOI: 10.4103/CRST.CRST_15_20

Get Permissions


Introduction: Statistics is an unpleasant but essential part of an oncologist's academic life. This review is an attempt to simplify the concept of various types of clinical studies.
Methods: Various books, articles, and online resources were used to gather information on types, definitions, clinical role, and brief methodology of various retrospective and prospective studies.
Results: The collected information has been concisely put together along with relevant examples to make the manuscript reader-friendly.
Discussion: This article is a brief account of various types of epidemiological and clinical studies used in medical research. Although some modifications are made time to time to suit the study requirements, basic framework remains the same.

Keywords: Clinical research, clinical studies, clinical trials, epidemiological studies, prospective studies, retrospective studies

How to cite this article:
Darling H S. Basics of statistics – 2: Types of clinical studies. Cancer Res Stat Treat 2020;3:100-9

How to cite this URL:
Darling H S. Basics of statistics – 2: Types of clinical studies. Cancer Res Stat Treat [serial online] 2020 [cited 2021 Oct 24];3:100-9. Available from: https://www.crstonline.com/text.asp?2020/3/1/100/279095

  Introduction Top

Research is the prime need of social evolution. It was well narrated by Guru Nanak Dev Ji 500 years ago as “khojee upjai baadee binsai” (The seeker comes forth, and the debater dies down, or, the person who pursues research, flourishes, and the discursist perishes).[1] The most basic and reliable study available is human observation. There are various ways to put this observation into a clinical context to conceptualize it, interpret it, draw concrete conclusions, and make it available for the community and further research. Although medical science has made stunning advances in various fields, modern medical practice is still full of challenges, diagnostic/therapeutic dilemmas, and medicolegal issues. There is a constant focus on standardizing and streamlining the medical practices worldwide through evidence-based medicine (EBM), which necessitates the care of patients using the best available research data to guide clinical decision-making. This necessitates maximal participation of the medical fraternity as well as patients into the clinical studies to gather more and more robust, region-specific data.[2],[3] Various types of systematic studies are conducted to unearth the parameters of the disease etiology, biology, diagnosis, and treatment. The endeavor of clinical research is to reach as close to the truth as feasible within the ambit of ethical considerations. The evidence for EBM is derived from various studies which may include original research, systematic reviews, and meta-analyses. Original research may be in the form of data from individuals or a population as experimental or observational studies.

Clinico-epidemiological studies are carried out pertaining to human health such as to estimate the risk of some event, to find cause and effect relationships, or to study the clinical effects of a new drug or a modality. Depending on the desired endpoints, ethical concerns, and methodological feasibility, various types of studies are feasible. Population-level studies can be experimental or observational, retrospective or prospective. The population selection and the study design are the most important factors to draw reliable conclusions.[4],[5],[6],[7] In ascending sequence of robustness of evidence, the following is an account of various types of clinical studies.

Observational studies

These studies follow the natural course of the disease. There is no interference or modification from the study team. These can be of two types: descriptive and analytical. Most of the epidemiological studies are observational only.

Descriptive study

The researcher only describes what is observed in the study population. It observes the distribution of a given issue in a defined population and attempts to identify its association with time, place, person, and etiological agent. For example, the observation by Burkitt in Burkitt's lymphoma in Africa led to the discovery of Epstein–Barr virus as the etiologic agent. Another classical example is the description by John Snow of the cholera epidemic in 1854 in the Golden Square district of London, after which the administration could locate the Broad Street water pump as the source of infection.[8] A descriptive study is the first step of an epidemiological survey. It helps in formulation of a hypothesis, which is further verified by techniques of analytical epidemiology. Case reports[9],[10],[11] and case series[12],[13],[14],[15] are also types of observational studies where the unit of observation is a single patient.

Analytical study

An analytical study tries to establish a relationship between health status and prevailing factors. As compared to descriptive studies, which look at the entire population, in analytical studies, the subject of interest is individual. The objective is to test the hypothesis formulated by descriptive studies. The studies help to establish the strength of cause–effect relationship.

Cross-sectional study

This study is conducted by collecting and analyzing data across the entire group of random population at a specific time [Figure 1]. It is different from case–control studies because the latter takes two arms: one with a specific health-related issue and the other, a control arm from a comparable population group without that issue. Cross-sectional studies are easier to perform and are highly useful in describing some population characteristics, for example, prevalence of a disease or correlation of a cause and effect. The downside is that they cannot provide the proof of an etiological correlation or consequences of a factor or intervention. They may also be described as censuses. For example, one may collect the cases of gallbladder cancer simultaneously in Punjab, Kerala, and Uttar Pradesh populations to study the difference in disease prevalence in these three states.
Figure 1: Schematic of various analytical studies

Click here to view

Case–control study

It is a retrospective study which involves a comparison of two populations: cases and controls in regards to a disease or health issue. Both exposure and outcome have already happened in the past. It goes retrograde from effect to cause and uses a control arm to accept or refute a finding. The comparator arms must be balanced in terms of age, sex, occupation, social status, etc., to minimize the confounding effect. An excellent example is that of oral contraceptive pills (OCPs) and thromboembolism reported in 1969 by Vessey and Doll.[16] They interviewed the women who had been hospitalized with deep venous thrombosis (DVT) or pulmonary thromboembolism (PE) without any known procoagulant disorder and then compared them to age-, parity-, and marital status-matched women in the same hospital admitted with other diseases. They found that 50% of patients with DVT and PE had been using OCPs as compared to 14% of controls. This case–control study proved that OCP intake and DVT coexisted more frequently than the comparable population not taking OCPs. Hence, it was concluded that OCPs raised the risk of DVT six times compared to nonusers.

Cohort study

This is a type of observational study conducted to establish the causal association between a suspected cause and effect. The comparator groups are followed over a time period to determine the frequency of development of the outcome under evaluation. From research terminology, “cohort” means a group of people sharing a similar characteristic profile. This is of three types:

  Prospective Cohort Study Top

The outcome has not yet occurred at the time of study initiation. For example, the long-term consequences of uranium exposure were studied by recruiting a cohort of uranium miners and comparing them to participants unexposed to uranium.[17] The findings of this study showed an increased frequency of lung cancer in the uranium miners. Similarly, in 1951, Doll and Hill formed two cohorts (smokers and nonsmokers) from 40,701 physicians.[18] They were followed for 4.5 years. At the end of the study, it was inferred that the relative risk of smokers of death from lung cancer was 22.4 times higher than nonsmokers and the population attributable risk of death due to lung cancer in smokers was 86%.

  Retrospective Cohort Study Top

The outcome has already occurred. The investigator goes back in time and accesses preserved records of past employment, attendance, medical record registry, etc., Comparison groups are generated and then their details traced forward through time, up to the present. An example of such a successful study is of a cohort study undertaken in 1978, where a cohort of 17,080 babies born between 1969 and 1975 at a Boston hospital were studied.[19] Babies who underwent electronic fetal monitoring during labor were compared to those who did not. The conclusion was that electronic fetal monitoring reduced the neonatal death rate by a factor of 1.7 times.

  Combination of Retrospective and Prospective Cohort Studies Top

This is a study where the cohorts are identified from the past records and studied till present, and then, the observation is continued prospectively into the future for further outcome assessment. Such a study can be illustrated by the Brown and Doll study.[20] where they constituted a cohort of 13,352 patients treated for ankylosing spondylitis with high-dose radiation between 1934 and 1954. The outcome assessed was death from leukemia or aplastic anemia, which was higher in the selected cohort than in the general population. The study was further continued in a similar fashion to add a prospective component to further assess the same outcome.

Experimental studies

Experimental studies are interventional studies in which the investigator is intervening in some way which can alter the outcome. Experimental studies provide the most robust evidence for any hypothesis as these are meticulously planned to control for confounding factors that may affect the outcome, whereas observational studies just monitor the effect in relation to existing circumstances [Table 1].
Table 1: Difference between experimental and observational studies

Click here to view

Field trials

These are the trials conducted in healthy population about prophylactic measures for primary prevention, e.g. Community Intervention Trial for Smoking Cessation.[21]

Community trials

These trials have community, population, breed, sex, or community as a participating unit. For example, it might be worthwhile to compare the effectiveness of a community-based educational program in which some villages receive an educational program on the importance of not drinking and driving, avoiding tobacco, exercising, and child immunization while the comparator which would include other villages does not.

Clinical trials

These are the interventional studies conducted in human patients to find the “objective clinical value” of a health-related intervention to improve overall clinical outcomes. Clinical trial is a type of clinical study designed to evaluate and test newer preventive, diagnostic, or therapeutic interventions. This process is so stringent that on an average, only one out of every 5,000 molecules that enter drug discovery to the stage of preclinical development passes a Food and Drug Administration (FDA) approval.[22] After the discovery and development of a newer molecule, technology, procedure, diagnostic tool or intervention, it must undergo a few preclinical processes under the monitoring of the FDA to enter clinical trials in human beings. Preclinical studies encompass the testing of a newer drug molecule, equipment, procedure, or other diagnostic or therapeutic modality in animals before it can be tested in humans. During these studies, the drug is tested for its positive and negative effects on the health and metabolism. The main aim of preclinical studies is to find the safe dose to begin with trials in human beings and know the side effect profile of the new molecule. Clinical trials are generally conducted in five phases. Each phase has a specifically defined purpose in drug development and approval.

Phase 0 trials

Phase 0 trial was a new concept introduced by the FDA in 2006 to hasten the clearance of the clogged drug pipeline. Researchers administer <1% of the therapeutic dose of an investigational drug to 10–15 healthy volunteers. This is a short, quick trial of less than a week, with hardly any adverse effect because of the “microdose.” The purpose is just to see whether the newer agent will work as well in humans as it did in animals. The nonfunctional drugs will immediately be stopped for further testing, providing a quick answer to pharmaceutical companies, in a way modernizing the clinical research and saving the unnecessary consumption of money and resources. Most molecules fail at this juncture still many more fail late in development.

Phase I trials

Researchers test an experimental drug or treatment in 20–100 human patients for the first time. This is also called a “dose-finding study” as the objective is to select a recommended Phase II dose, the so-called “maximum tolerated dose (MTD)” in a given schedule and route of administration. The primary endpoint is toxicity. The grade and types of dose-limiting toxicities (DLTs) help to determine dose escalation or de-escalation. Conventionally, it is assumed that both efficacy and toxicity increase with dose, so the highest tolerable dose of a particular agent or a combination of agents is selected while expecting it to have the highest efficacy.

In adults, the starting dose is based on animal toxicology studies and generally is one-tenth the dose lethal to 10% of a cohort of mice, expressed in mg/m2. Often, a “modified Fibonacci scheme” is used to determine the dose levels for successive cohorts. The starting dose is increased by 100% in the second level, and subsequent levels are increased by adding 67%, 50%, 40%, and 33% of the dose established in the preceding cohort, reflecting increasing caution with the higher doses. The oldest and still most commonly used Phase I design is the so-called 3 + 3. In this design, patients are enrolled in cohorts of three, beginning at the lowest dose level, and are then observed for acute toxicity. If none of the three evaluable patients experiences a DLT, the dose level is escalated. At any dose level, when DLT is observed in one patient, three more patients are added. The MTD is defined as the dose level immediately below the dose causing a DLT in the second patient. This methodology depicts Phase IA trials where a single ascending dose is used at a time [Figure 2]. On the contrary, in Phase IB trials, multiple escalating dose arms run in parallel to study the pharmacokinetics and pharmacodynamics of multiple doses of the drug simultaneously, through blood and body fluid sampling, evaluating the safety and tolerability.
Figure 2: “3 + 3” design of Phase 1a trial, D0 (Starting dose)=1/10th of dose lethal to 10% mice

Click here to view

With the changing pace and methodology of medical technology, certain modifications are in vogue. Phase I trials of “targeted therapies” may not require dose escalation to DLT. Hence, the study endpoint here is the “optimal dose” producing the best desired response rather than the MTD.[23] Continual reassessment method (CRM) aims at moving to the MTD more quickly and thereby increasing both the efficiency of this early phase of development and the likelihood that patients treated in Phase I will receive potentially beneficial treatment. Delayed toxicities are common in oncology. Time-to-event CRM considers toxicity over a longer period of time while still allowing relatively quick escalation decisions.

In children, Phase I trials usually start after some adult data on the agent of interest are already available. One efficient method is to start children's trials at 80% of the adult Phase II dose. Starting doses derived from the adult MTD or recommended Phase II dose are presumably close to the childhood MTD, and escalation should proceed cautiously, using approximately 30% increases over the preceding dose level. Dose escalation often continues in children beyond the Phase II dose established for adults because children often display greater tolerance to chemotherapy.

Phase II trials

The experimental drug or intervention is given to a larger group of people to test the efficacy and to further evaluate its safety. The drug is often tested among patients with a specific type of cancer. The primary purpose is to determine whether the new agent is sufficiently promising to warrant further study, usually by comparing the new treatment with a prespecified standard or historical control. If an agent shows activity, it can be tested in a Phase III clinical trial. Although the gold standard for evaluation of clinical benefit in oncology is improvement in overall survival, this is rarely a feasible outcome in Phase II trials. It takes too long, and study agent effects are likely to be confounded, with effects of subsequent therapies rendering the survival results uninterpretable. The most common Phase II endpoint is objective response. Other commonly used endpoints include time to tumor progression, progression-free survival, overall survival, quality of life, change in molecular biomarkers, and change in functional imaging. Both Phase I and Phase II studies ideally should be carried out in treatment-naive patients to avoid the problems of cumulative toxicity from prior therapy and acquired tumor drug resistance; however, this is hardly ever possible because of the obvious reasons. One strategy to increase the generalizability of results of Phase II studies is to perform a brief Phase II study in patients before standard therapy begins. This design is called a “Phase II window” or “upfront window.” In general, the sample size is 100–300 participants. Pediatric Phase II trials are relatively small, with sample sizes generally in the 15–30 participants per arm range. Phase IIA studies are generally pilot projects trying to prove clinical activity (“proof of concept” studies), whereas Phase IIB studies attempt to establish the ideal dose at which the drug shows maximal biological activity with minimal adverse effects (“definite dose-finding” studies).

Phase III trials

Phase III trials compare a new intervention to the standard-of-care practice. Patients are randomized to comparator arms to confirm the effectiveness and monitor side effects of the intervention arm as compared to the control arm. Randomization is done to match the baseline characteristics. The number of arms in the Phase III trial, the need for blinding and the ratio of randomization depends on the design of the Phase III trial. There are two or more treatment arms in Phase III trials. The control group gets the standard-of-care treatment, which can be an active treatment or a placebo or even best supportive care. The other group gets a new treatment. Blinding is done to mask the patient allocation to treatment arms, to avoid any kind of bias. The study may be stopped early if the toxicity of the new agent is too high or if one group fares exceedingly better. Phase III clinical trials are often needed before the FDA will approve the use of a new drug for clinical use. The ideal endpoint for evaluation of clinical benefit in oncology is improvement in overall survival although this may not be practical when deaths tend to occur long after treatment. Consequently, alternative endpoints are often chosen that are presumed to be surrogate markers of long-term survival, such as the disappearance of detectable tumor or the absence of recurrence or metastases at 3 years. As a result, an endpoint that is widely used in trials of childhood malignancies is “event-free survival.” The sample sizes of Phase III trials vary widely but can range from as few as several hundred to thousands or even tens of thousands (generally 300–3000), as in adult prostate or breast cancer prevention trials.

  Factorial Designs Top

Phase III trials can compare more than two treatments at once. In a factorial design, two or more questions are addressed simultaneously in the same cohort of patients. For example, in the National Wilms' Tumor Study III,[24] early-stage patients were randomized between postoperative radiotherapy and no postoperative radiotherapy and also between two different chemotherapy regimens.

  Noninferiority Trial Top

Most trials are designed or intended to prove superiority (better efficacy) of the intervention arm. If this is not feasible, then the design chosen is a noninferiority trial. This is a Phase III trial whose purpose is to demonstrate that a new treatment is no less efficacious than a standard treatment. This is important if the new treatment is clearly more desirable than the standard in some other way (e.g. less toxic, less invasive, less expensive, or more convenient).

  Bioequivalence Trial Top

Bioequivalence trials are used to show that a new treatment is identical (within an acceptable range) to a current treatment. This is used in the registration and approval of generic drugs (or biosimilars in case of monoclonal antibodies) that are shown to be bioequivalent to their branded reference drugs. Sample sizes for equivalence trials are usually larger than those in difference-seeking trials because physicians are very reluctant to adopt a new treatment that is less effective than current treatments.

Biomarker-targeted trials

Molecularly targeted therapies are an increasingly important focus of clinical oncology in adults and in pediatrics. The proper design and interpretation of these trials is complex and is an area of active methodologic research. There are several key questions to be asked in considering such a trial. First, how certain are we of the relevance of the putative target or biomarker? Second, how good is the assay for the biomarker? Third, how common is the target? For example, in breast cancer, tamoxifen and the estrogen receptor, which is commonly positive, and trastuzumab and HER2, which is much less commonly positive. The benefit of tamoxifen was observed despite including estrogen receptor-negative and estrogen receptor-positive patients, while trastuzumab was studied only in HER2-positive participants and might not have been approved if it had been studied in a broader population less likely to respond. Three major modern methods of biomarker-driven trials are as below [Figure 3].
Figure 3: Various models of biomarker-driven trials; BDRx – biomarker-driven treatment

Click here to view

  Basket Trials Top

Traditionally, oncology pharmaceutical research is pursued to test one or two drugs in a single disease. A basket trial evaluates the efficacy of a single investigational agent or a treatment across multiple cancer types defined by disease stage, histology, previous treatment history, molecular biomarkers, or demography. Usual primary endpoint is overall response rate to estimate the drug efficacy in a single-arm study. A strong response may allow further recruitment in that subgroup to yield more data sufficient to obtain a marketing approval. Vemurafenib is the first FDA-approved drug on the basis of a basket trial. This approval was based on data from the Phase II VE-BASKET study,[25] a nonrandomized, histology-agnostic evaluation of the efficacy of vemurafenib, a BRAF V600 kinase inhibitor.

  Umbrella Trials Top

An umbrella trial evaluates multiple new agents or treatments in a single disease population. After completion of safe dose-finding part, the researchers move to efficacy testing component. The Lung-MAP study[26] is an umbrella trial design which contains a screening part and a clinical trial part. The screening part enrolls patients while on prior palliative therapy or after progression on prior therapy for Stage IV or recurrent non-small cell lung cancer (NSCLC). The clinical trial part enrolls patients on biomarker-based study arms or nonbiomarker-based arms evaluating agents in participants without any positive study biomarker.

Adaptive designs

An adaptive clinical trial evaluates a medical treatment or modality by efficacy and side effects on the prescribed schedule and fine-tunes the design of the trial protocol in accordance with these outcomes. This process is called adaptation, which may continue throughout the study period. The investigators may modify dose, inclusion and exclusion criteria, etc. However, the whole adaptation process is conducted within the purview of a prespecified protocol. The FDA has issued special guidelines on the design of such trials. The premise of an adaptive trial is to hasten the selection process of useful therapies in exact doses and formulations and to determine the ideal patient populations. As compared to conventional fixed design trials, such trials are often more fast, efficient, informative, and ethical. There is also more judicious utilization of resources such as time, money, and required sample size. The FDA permits an adaptive Bayesian clinical trial to have interim evaluation to stop or to (i) adjust patient accrual; (ii) assess stopping the trial early either for success, futility, or harm; (iii) reversing the hypothesis of noninferiority to superiority or vice versa; and (iv) dropping arms or doses or adjusting doses.[27]

Cluster-randomized trial

In this trial design, individuals are randomized in groups. The randomization unit is group as compared to individuals. For instance, “A national cluster-randomized controlled trial to examine the effect of enhanced reminders on the socioeconomic gradient in uptake in bowel cancer screening”[28] is an excellent and simple example of cluster randomization. The researchers enhanced the usual reminder by including a heading “A reminder to you” and a short paragraph restating the offer of screening in simple language. Out of 1,68,480 individuals, receiving a reminder over 20 days, the enhanced reminder was received by 78,067 individuals and 90,413 received the usual reminder. In this trial, the groups were randomized as per day of invitation. As a result, there was a small but statistically significant (P = 0.001) increase in participation with the enhanced reminder (25.8% vs. 25.1%).

Feasibility studies and pilot studies

Feasibility studies are small bits of research conducted before starting the main study to discover the practical issues expected to be faced. These may be helpful in sample size calculation, willingness of participants and clinicians, compliance rates of participants to study protocols, and time needed to complete the study. Pilot studies are “mini” versions of the complete study to further assess the practicality of the protocol and possible challenges likely to be confronted. Akin to the main study, the primary outcome is also assessed. In internal pilot studies, these data contribute to the final analysis too, which is not the case in external pilot studies.

Phase IV trials

Out of 548 new chemical entities approved by the FDA between 1975 and 1999, 45 (8.2%) received black box warnings and 16 (2.9%) were withdrawn from the market over more than 7 years. Hence, the FDA mandated postmarketing studies from 2007 onward, which are carried out after the approval by the FDA and are meant to provide detailed real-life information about drug's risks, benefits, and best use. The drug is tested in large patient populations across the world. The additional information obtained may include some rare side effects which may only be found in large groups of people, or with other concomitant medications and long-term toxicities, for example, secondary malignancies. Troglitazone and rofecoxib are two recent examples of drug withdrawals[29] from the market on the basis of Phase IV trials. The minimum time period mandated for Phase IV clinical trials is 2 years.

Reviews and analyses

The framing of clinical guidelines by various international societies takes into account the comprehensive assessment of the highest forms of research-based evidence available worldwide in any field of medicine. A single clinical trial, even a large well-conducted one, is seldom sufficient to provide a dogmatic answer to a clinical question. The “truth” needs to be understood by examining all sources of data as critically and objectively as possible. Systematic review and meta-analysis are the higher forms of medical literature devised to condense the available evidence with an explicit, unbiased approach weighing the strengths and fallacies of the individual studies, the populations and interventions, and specific outcomes that were assessed.

Systematic review

A systematic review is the most robust proof of available scientific evidence. It is a comprehensive critical summary of “all existing studies” that meet selective eligibility criteria to address a specific clinical situation. It involves a stringent process that mandates systematic recruitment of studies that have evaluated the same specific research question, critical analysis of the studies that may involve meta-analysis, presentation of salient findings, and explicit discussion of the limitations of the evidence and the review. For example, a recent systematic review, studying correlation of “Red and Processed Meat Intake and Cancer Mortality and Incidence,”[30] evaluated 118 articles (56 cohorts) with more than 6 million participants. Out of these, 73 articles were eligible as per study criteria. The study concluded that possible actual impact of red and processed meat consumption on cancer mortality and incidence is very small, and the certainty of evidence is low to very low. This is in contrast to various studies with varying results on the association of these two entities. Systematic reviews are superior to traditional “narrative” reviews and textbook chapters. The latter generally do not dive deep into the literature, lack transparency in the study recruitments, may lack dependable supporting evidence, may not be completely unbiased, and fail to provide quantitative comparisons.


Meta-analysis, which is frequently included in systematic reviews, is a statistical study that quantitatively clubs the results from different studies. It generally provides an overall estimate of the net benefit or harm of an intervention, despite individual studies showing unequivocal results. Meta-analysis can also provide an overall quantitative estimate of other parameters such as diagnostic accuracy, incidence, or prevalence. In contrast to the results of systematic review quoted above, a meta-analysis of 10 cohort studies reported a statistically significant dose–response relationship between meat consumption and colorectal cancer risk[31] with a 17% increased risk (95% confidence interval [CI], 1.05–1.31) per 100 g/day of red meat and an 18% increase in risk (95% CI, 1.10–1.28) per 50 g/day of processed meat. Similarly, an updated meta-analysis of 18 randomized trials on adjuvant doxorubicin-based chemotherapy in soft-tissue sarcoma[32] showed statistically significant overall survival benefit (odds ratio for death: 0.56; 95% CI, 0.36–0.85). In contrast to this result, a pooled analysis of individual patient data from the two largest adjuvant EORTC trials of doxorubicin- and ifosfamide-based chemotherapy, totaling 819 patients, was negative for any survival benefit.[33] This is the fallacy of a meta-analysis, which is circumvented by systematic analysis.


Different clinical trials of the same disease and drug may use different designs, endpoints, and dosages. Regression analysis of primary studies may be vital to adjust for potential confounders or explain differences in results among participants. This meta-analytic technique is commonly known as meta-regression. This enables a better systematic cross-trial comparison. Instead of individual patients serving as the units of analysis, each individual study is considered to be one observation. Meta-regression tests the statistical interaction between the subgroup variable (e.g., dose) and the treatment effect (e.g., relative risk of death). The meta-regression successfully explains the heterogeneity across studies, showing an association between treatment duration and the effect of treatment on death that was not apparent within the individual trials. For example, a meta-regression analysis on early-life energy restriction (ER) and cancer risk in humans[34] found that early-life transient severe ER seems to be associated with increased cancer risk in the breast and prostate cancer.

Subgroup analyses

A simpler alternative to meta-regression is to categorize studies into subgroups based on clinically relevant variables. Separate meta-analyses of these subgroups can then be graphically compared. Graphical comparisons and meta-regression can be useful together since the former provides an overview while the latter permits multivariable analysis and evaluation of interactions. In a systematic review of treatment of obstructive sleep apnea on depressive symptoms,[35] a subgroup analysis showed details about the few studies of patients with baseline depression and the more numerous studies of patients with no baseline depression. A meta-regression from the same systematic review allows statistical comparisons between the effects in the different subgroups in this case highlighting a statistically significant difference in the comparison of effects in studies of people with baseline depression versus studies of people without baseline depression.

Network meta-analysis

Systematic reviews commonly compare the relative effects of several drugs (e.g., different CDK4 inhibitors) or modalities related to one class, not just the comparison of two specific interventions, as done usually in individual trials. When multiple specific interventions are compared across trials, a network of studies can be established where all the studied interventions are linked to each other by individual trials. Network meta-analysis or multiple treatment comparison evaluates all studies and all interventions (may be from different classes) simultaneously to produce multiple pairwise estimates of relative effects of each intervention compared with every other intervention. For example, an interesting network meta-analysis of the frontline immune-checkpoint inhibitor (IO)-based regimens in advanced NSCLC suggested superiority of pembrolizumab + platinum doublet as compared to other IO-based regimens.[36]

Let us test what we have learned

  • Q1: Choose the correct sequence considering the robustness of evidence

    1. Analytical studies < descriptive studies < systematic review < meta-analysis
    2. Observational studies < experimental studies < meta-analyses < systematic review
    3. Preclinical trials < analytical studies < meta-analyses < Phase IV trials
    4. Pilot studies < observational studies < meta-analyses < experimental studies

  • Q2: A dose-finding study is

    1. Preclinical trial
    2. Feasibility study
    3. Phase 0 trial
    4. Phase 1A trial

  • Q3: Choose the purely retrospective study

    1. Meta-regression
    2. Cohort study
    3. Case–control study
    4. Cross-sectional study

  • Q4: Multiple cancers are tested for a single intervention in

    1. Basket trial
    2. Umbrella trial
    3. Adaptive design
    4. Placebo-controlled trial

  • Q5: The main focus on drug toxicity is in

    1. Postmarketing study
    2. Dose-finding study
    3. Both of above
    4. None of above

    5. Q6: Generally, the sample size in a Phase II trial is

    1. 10–15
    2. 20–100
    3. 100–300
    4. 300–3000

  • Q7: A study comparing different drugs from different trials is

    1. Network meta-analysis
    2. Adaptive design
    3. Subgroup analysis
    4. Postmarketing surveillance

  • Q8: A trial conducted on healthy volunteers uses

    1. MTD
    2. “3 + 3” design
    3. 10% of dose lethal to 10% mice
    4. 1% of therapeutic dose

  • Q9: The methodology used to adjust the potential confounders while comparing various studies is

    1. Meta-regression
    2. Meta-analysis
    3. Cluster randomization
    4. Subgroup analysis

  • Q10: An adaptive Bayesian trial design can involve all except

    1. Dropping an arm
    2. Introducing a new arm
    3. Reversing the hypothesis of superiority to noninferiority
    4. Changing the dose

Answers: 1 (b), 2 (d), 3 (c), 4 (a), 5 (c), 6 (c), 7 (a), 8 (d), 9 (a), 10 (b)

Financial support and sponsorship


Conflicts of interest

There are no conflicts of interest.

  References Top

PAGE 1255 – Gurmukhi to English Translation and Phonetic Transliteration of Siri Guru Granth Sahib. Available from: http://www.srigurugranth.org/1255.html. [Last accessed on 2020 Jan 07].  Back to cited text no. 1
Noronha V. Making a case for cancer research in India. Cancer Res Stat Treat 2018;1:71-4.  Back to cited text no. 2
  [Full text]  
Malik PS, Sahoo RK, Khurana S. Challenges for cancer research in India: What's the way out? Cancer Res Stat Treat 2018;1:179-80.  Back to cited text no. 3
  [Full text]  
Park's Textbook of Preventive and Social Medicine – Community Health. Available from: http://www.communityhealth.in/~commun26/wiki/index.php?title=Park%27s_Textbook_of_Preventive_and_Social_Medicine. [Last accessed on 2020 Jan 07].  Back to cited text no. 4
Ioannidis JP. Why most published research findings are false: Author's reply to Goodman and Greenland. PLoS Med 2007;4:e215.  Back to cited text no. 5
LeLorier J, Grégoire G, Benhaddad A, Lapierre J, Derderian F. Discrepancies between meta-analyses and subsequent large randomized, controlled trials. N Engl J Med 1997;337:536-42.  Back to cited text no. 6
Strzebonska K, Waligora M. Umbrella and basket trials in oncology: Ethical challenges. BMC Med Ethics 2019;20:58.  Back to cited text no. 7
John Snow and the 1854 Cholera Outbreak – Past Medical History. Available from: https://www.pastmedicalhistory.co.uk/john-snow-and-the-1854-cholera-outbreak/. [Last accessed on 2020 Jan 29].  Back to cited text no. 8
Talreja VT, Noronha V, Patil VM, Joshi A, Prabhash K. Desperate times, desperate measures: Low-dose nivolumab-induced remission in relapsed NSCLC. Cancer Res Stat Treat 2019;2:266-7.  Back to cited text no. 9
  [Full text]  
Abraham G, Menon N, Patil VM, Joshi AP, Prabhash K. The efficacy of low-dose immunotherapy in head-and-neck cancer. Cancer Res Stat Treat 2019;2:268-9.  Back to cited text no. 10
  [Full text]  
Dhumal S, Sable N, Patil VM, Vallathol DH, Prabhash K. Induction chemotherapy in recurrent, unresectable esthesioneuroblastoma. Cancer Res Stat Treat 2018;1:171-2.  Back to cited text no. 11
  [Full text]  
Kashyap L, Noronha V, Patil V, Joshi A, Mahajan A, Mittal N, et al. Dermatofibrosarcoma protuberans of head and neck: Clinical outcome of nine cases treated with imatinib. Cancer Res Stat Treat 2019;2:112-8.  Back to cited text no. 12
  [Full text]  
Noronha V, Enting D, Thippeswamy R, Joshi A, Patil VM, Prabhash K. Hypersensitivity reactions to paclitaxel with a modified dexamethasone intravenous premedication regimen. Cancer Res Stat Treat 2018;1:78-83.  Back to cited text no. 13
  [Full text]  
Ankalkoti B, Pareekutty NM, Kattepur AK, Kumbakara R, Vazhayil R, Balasubramanian S. Life and limb after limb salvage surgery for osteosarcoma in an Indian tertiary cancer center. Cancer Res Stat Treat 2019;2:145-51.  Back to cited text no. 14
  [Full text]  
Pareekutty NM, Kattepur AK, Alapatt JJ, Satheesan B. Unplanned excision in soft-tissue sarcomas – Is it the unkindest cut of all? Cancer Res Stat Treat 2019;2:152-7.  Back to cited text no. 15
  [Full text]  
Vessey MP, Doll R. Investigation of relation between use of oral contraceptives and thromboembolic disease. A further report. Br Med J 1969;2:651-7.  Back to cited text no. 16
Tomásek L, Swerdlow AJ, Darby SC, Placek V, Kunz E. Mortality in uranium miners in West Bohemia: A long-term cohort study. Occup Environ Med 1994;51:308-15.  Back to cited text no. 17
Doll R, Hill AB. Mortality in relation to smoking: Ten years' observations of British doctors. Br Med J 1964;1:1460-7.  Back to cited text no. 18
Jenkins HM. Thirty years of electronic intrapartum fetal heart rate monitoring: Discussion paper. J R Soc Med 1989;82:210-4.  Back to cited text no. 19
Smith PG, Doll R. Mortality among patients with ankylosing spondylitis after a single treatment course withxrays. Br Med J (Clin Res Ed) 1982;284:449-60.  Back to cited text no. 20
Community Intervention Trial for Smoking Cessation: I. Cohort results from a four-year community intervention. Am J Public Health 1995;85:183-92.  Back to cited text no. 21
Drug Approvals – From Invention to Market 12 Years! Available from: https://www.medicinenet.com/script/main/art.asp?articlekey=9877. [Last accessed on 2020 Jan 29].  Back to cited text no. 22
Patil VM, Noronha V, Joshi A, Abhyankar A, Menon N, Banavali S, et al. Low doses in immunotherapy: Are they effective. Cancer Res Stat Treat 2019;2:54-60.  Back to cited text no. 23
  [Full text]  
D'Angio GJ. The National Wilms tumor study: A 40 year perspective. Lifetime Data Anal 2007;13:463-70.  Back to cited text no. 24
Hyman DM, Blay JY, Chau I, Raje NS, Fernandez ME, Wolf J, et al. VE-BASKET, a first-in-kind, phase II, histology-independent “basket” study of vemurafenib (VEM) in nonmelanoma solid tumors harboring BRAF V600 mutations (V600m). JCO 2014;32 Suppl 15:2533.  Back to cited text no. 25
Herbst RS, Gandara DR, Hirsch FR, Redman MW, LeBlanc M, Mack PC, et al. Lung master protocol (Lung-MAP)-A biomarker-driven protocol for accelerating development of therapies for squamous cell lung cancer: SWOG S1400. Clin Cancer Res 2015;21:1514-24.  Back to cited text no. 26
Bhattacharjee A, Vishwakarma GK, Banerjee S. A Bayesian approach for dynamic treatment regimes in the presence of competing risk analysis. Cancer Res Stat Treat 2018;1:51-7.  Back to cited text no. 27
  [Full text]  
Raine R, Moss SM, von Wagner C, Atkin W, Hans IK, Howe R, et al. A national cluster-randomised controlled trial to examine the effect of enhanced reminders on the socioeconomic gradient in uptake in bowel cancer screening. Br J Cancer 2016;115:1479-86.  Back to cited text no. 28
Onakpoya IJ, Heneghan CJ, Aronson JK. Delays in the post-marketing withdrawal of drugs to which deaths have been attributed: A systematic investigation and analysis. BMC Med 2015;13:26.  Back to cited text no. 29
Han MA, Zeraatkar D, Guyatt GH, Vernooij RWM, El Dib R, Zhang Y, et al. Reduction of red and processed meat intake and cancer mortality and incidence: A systematic review and meta-analysis of cohort studies. Ann Intern Med 2019. doi: 10.7326/M19-0699. [Epub ahead of print].  Back to cited text no. 30
Norat T, Lukanova A, Ferrari P, Riboli E. Meat consumption and colorectal cancer risk: Dose-response meta-analysis of epidemiological studies. Int J Cancer 2002;98:241-56.  Back to cited text no. 31
Pervaiz N, Colterjohn N, Farrokhyar F, Tozer R, Figueredo A, Ghert M. A systematic meta-analysis of randomized controlled trials of adjuvant chemotherapy for localized resectable soft-tissue sarcoma. Cancer 2008;113:573-81.  Back to cited text no. 32
Le Cesne A, Ouali M, Leahy MG, Santoro A, Hoekstra HJ, Hohenberger P, et al. Doxorubicin-based adjuvant chemotherapy in soft tissue sarcoma: Pooled analysis of two STBSG-EORTC phase III clinical trials. Ann Oncol 2014;25:2425-32.  Back to cited text no. 33
Elands RJ, Simons CC, Dongen Mv, Schouten LJ, Verhage BA, van den Brandt PA, et al. A systematic literature review and meta-regression analysis on early-life energy restriction and cancer risk in humans. PLoS One 2016;11:e0158003.  Back to cited text no. 34
Povitz M, Bolo CE, Heitman SJ, Tsai WH, Wang J, James MT. Effect of treatment of obstructive sleep apnea on depressive symptoms: Systematic review and meta-analysis. PLoS Med 2014;11:e1001762.  Back to cited text no. 35
Frederickson AM, Arndorfer S, Zhang I, Lorenzi M, Insinga R, Arunachalam A, et al. Pembrolizumab plus chemotherapy for first-line treatment of metastatic nonsquamous non-small-cell lung cancer: A network meta-analysis. Immunotherapy 2019;11:407-28.  Back to cited text no. 36


  [Figure 1], [Figure 2], [Figure 3]

  [Table 1]


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

  In this article
Prospective Coho...
Retrospective Co...
Combination of R...
Factorial Designs
Noninferiority Trial
Bioequivalence Trial
Basket Trials
Umbrella Trials
Article Figures
Article Tables

 Article Access Statistics
    PDF Downloaded155    
    Comments [Add]    

Recommend this journal