Practical Biostatistics A Friendly Step by Step Approach for

Evidence based Medicine

Visit the link below to download the full version of this book:

[Link]
-approach-for-evidence-based-medicine/

Click Download Now

 Acknowledgments

My mother, for taking me by the hand up to the point I could go on with my

own feet.
My wife Ester and my son Renan, who accepted sacrificing our shared
time so that this project could become a reality.
Luiz Claudio S. Brillanti and Renan Suchmacher, for their contribution to
cover art and illustrations.
Joshua Dahlben (NYU) for language review of the manuscript.
Our patients, the very reason this book came into existence.

xii
 Part I

Introduction
The objective of Part I is to instruct the reader on how to choose the
study type that best fits his or her research goals and material resources
as well as how to determine the study type of a given paper.
 Chapter 1

Study Type Determination

A crucial stage in research planning is to determine the most suitable study

type, according to the investigator’s hypothesis (Chapter 4) and material
resources available. A proposed classification of study types for epidemio-
logical and clinical research is detailed in this chapter.

1.1 EPIDEMIOLOGICAL STUDIES

Epidemiological studies aim to establish the frequency of a condition in a
given population. In this type of study, establishing reliable risk or etiologi-
cal correlations is not possible because precise consistency between risk fac-
tors exposure or nonexposure, as well as between affected or nonaffected
groups, cannot be achieved. For this reason, epidemiological studies remain
mostly limited to measuring frequencies. Suspected risk correlations should
be clarified through analytical studies.
Epidemiological studies can be classified into three types: ecological
studies, cross-sectional studies, and longitudinal studies.

1.1.1 Ecological Studies

Ecological studies aim to determine the frequency of a given condition, sup-
posedly associated with some environmental factor, during parallel or subse-
quent past time spans. For example, during the past 3 years, students from
school A, which is located in an underprivileged district, have been present-
ing a higher frequency of upper respiratory virus infection compared to stu-
dents from school B, which is located in a middle-class district, whose
frequency is considered usual.

1.1.2 Cross-Sectional Studies

In cross-sectional studies, the frequency of a given condition in a naturally
evolving population under a suspected exposure factor is analyzed, like a
snapshot (Figure 1.1).

M. Suchmacher & M. Geller: Practical Biostatistics. DOI: 10.1016/B978-0-12-415794-1.00001-X

© 2012 Elsevier Inc. All rights reserved. 3
4 PART | I Introduction

CROSS-SECTION

TIME
group with the condition

TARGET
POPULATION
group without the condition

FIGURE 1.1 Schematic representation of cross-sectional study type.

1.1.3 Longitudinal Studies

In longitudinal studies, a cohort is followed for several years, or sometimes dec-
ades, in order to establish the frequencies of specific conditions and their correla-
tion with environmental or other biological factors. Comparisons can be
performed intrasubject or between different subjects, if possible. A classic exam-
ple is the Framingham Heart Study, which began in 1948 with 5209 subjects and
is currently in its third generation of participants. Knowledge on important envi-
ronmental factors currently associated with cardiovascular risk, such as lifestyle
(smoking, diet, and exercise) and aspirin use, has been derived from this study.

1.2 ANALYTICAL STUDIES

Analytical studies aim to establish correlation strength between a condition
and a factor putatively associated with its origin and/or natural history.
These types of studies are the main focus of this book and can be classified
either as observational or intervention studies.

1.2.1 Observational Studies

In observational studies, the frequency of a condition in a population is stud-
ied under so-called “natural” circumstances. As such, active intervention
from the investigator on these circumstances is not applicable. The main
objective of observational studies is to establish the degree of hazard for a
certain condition in relation to a considered exposure factor. Submitting the
observed population to “real world” situations is its advantage. Its limitation
is that it yields less accurate conclusions because uncontrolled variables and
potential confounders may generate bias. Observational studies can be classi-
fied as case control or cohort studies.

Case Control Studies

In case control studies, two groups are retrospectively compared, according
to the following model: (1) One group with the condition (case) is
Chapter | 1 Study Type Determination 5

subdivided into two subgroups one exposed and the other nonexposed to a
studied exposure factor; and (2) another group without the condition (con-
trol) is subdivided into two subgroups one exposed and the other nonex-
posed to the same factor (Figure 1.2).
Case control studies aim to determine the odds of acquiring a condition
under exposure to a considered factor. For example, the odds of miners pre-
senting asbestos-associated lung fibrosis relative to the general population
are 1.5:1. The advantages of case control studies are that they are less
expensive to perform compared to cohort studies and they can be performed
immediately because they are generally retrospective. Also due to this latter
aspect, their limitations are: poor control over the exposure factor, uncon-
trolled variables, and potential confounders. Given the fact they focus on the
outcome and “move back” toward the exposure factor, they are generally
retrospective. Their inferred association strength the odds ratio is calcu-
lated using a specific formula (Chapter 2).

Cohort Studies
In cohort studies, a cohort of healthy subjects is divided into two groups
according to exposure or nonexposure to a given factor exposed subjects
and nonexposed subjects in principle for prospective follow-up. At the
end of the study, the number of subjects with and without the condition is
measured (Figure 1.3).
Cohort studies aim to determine the risk of acquiring a condition under
exposure to a considered factor. For example, nuclear power plant workers have
a 2.5 times greater risk of presenting high-grade lymphoma relative to nonex-
posed subjects. The advantage of these studies is that they afford better control
over exposure level, covariates, and potential confounders because they are pro-
spective. Their limitations are the need to wait for exposure factors to exert their
effects and higher cost. Given the fact they focus on the exposure factor and
“move forward” to the outcome, they are generally prospective.
Their inferred association strength relative risk is calculated using a
specific formula (see Chapter 2). Note that the expression “cohort study”
refers to a specific type of observational study. In this sense, the term
“cohort” must be differentiated from its broader sense (see Glossary), which
can be applied more widely.

Retrospective Cohort Studies

In retrospective cohort studies, two groups are retrospectively identified and
“prospectively” compared according to the following model: A cohort of
healthy subjects is subdivided into two groups one exposed to a given fac-
tor and the other nonexposed to the same factor (Figure 1.4).
The advantages of retrospective cohort studies are that they are less expen-
sive to perform than cohort studies and they can be performed immediately
6 PART | I Introduction

Time

Investigation direction

Exposed subjects

Group with the condition

(case)

Non-exposed subjects
Initial
population
Exposed subjects

Group without the

condition (control)

Non-exposed subjects

FIGURE 1.2 Schematic representation of a case control study type.

Time

Investigation direction

Subjects with
the condition

Exposed subjects

Subjects without
the condition
Cohort of
healthy
subjects
Subjects with
the condition

non-exposed subjects

Subjects without
the condition

FIGURE 1.3 Schematic representation of a cohort study type.

Chapter | 1 Study Type Determination 7

Present time

Real time
Progression
Prospection

Investigation direction

Subjects with the

condition
Exposed
subjects
Subjects without
Cohort of the condition
healthy
subjects
Subjects with the
condition
Non-exposed
subjects
Subjects without
the condition

FIGURE 1.4 Graphic representation of a retrospective cohort study type.

because they are retrospective. Also due to this latter aspect, their limitation is:
poor control over the exposure factor, covariates, and potential confounders.

1.2.2 Intervention Studies

Intervention studies (sometimes referred to as “randomized studies” or “con-
trolled studies”) are prospective cohort studies generally performed using a
reference group (control group). Here, the investigator plans and actively
intervenes with the factors influencing his or her cohort, minimizing the
influence of uncontrolled variables and potential confounders. These studies
are classified according to the applied intervention model.

Reference Standard
A study’s reference standard changes according to its type: either noncom-
parative or comparative (controlled) studies.

Noncomparative Studies
In noncomparative studies, there is no reference standard to compare with
the tested group. Results interpretation is difficult because the investigator
8 PART | I Introduction

can never be certain whether or not the findings represent an actual influence
from the tested drug or vaccine.

Comparative Studies (Controlled Studies)

In comparative studies, there is a reference standard group (control) to com-
pare to the tested group. The reference group is represented either by an
active drug or vaccine known to be efficacious (active or positive control) or
by placebo (negative control). Results interpretation by the investigator is
more consistent because there is a comparison reference. In most situations,
control group subjects are contemporary to tested group subjects. However,
in some circumstances, historical (literature) controls can be used.

Relation Between Samples and Within Samples

The linkage between and within samples can change according to the study.

Nonpaired (Independent) Samples Studies

In nonpaired samples studies, individuals from a group can be freely matched
for comparison with any individuals from the other group. Comparability
is provided by the study design. For example, a sample of elderly patients
admitted for cholelithiasis is divided into a group directed for emergency
cholecystectomy and another group directed for conservative cholelithiasis
management aiming for further elective surgery, and mortality and morbidity
rates are compared.

Paired (Dependent) Samples Studies

In paired samples studies, individuals from a group can be matched for com-
parison exclusively with specific individuals from the other group. Three
types of pairing are possible:
G Self-pairing: An individual is matched with him- or herself. For example,
a sample of male patients with chronic neuropathic pain due to lumbar
disk herniation complicated with radiculoneuropathy is treated with acu-
puncture for 3 months and with anticonvulsants for 3 months thereafter
for optimal pain management comparison.
G Natural pairing: Natural pairing is performed between two different but
extremely well linked individuals, such as monozygotic twins.
G Artificial pairing: Artificial pairing is performed between independent
individuals, with pairing done according to a specific studied variable.
For example, a sample of elderly patients admitted for cholelithiasis is
divided into a group directed for emergency cholecystectomy and another
group directed for conservative cholelithiasis management aiming for
further elective surgery, and mortality and morbidity rates are compared.
Nevertheless, according to study design, individuals from a group can
Chapter | 1 Study Type Determination 9

only be paired with individuals from the other group who share the same
serum bilirubin level ranges.

Awareness of Tested Drug, Vaccine, or Exam

Open studies, single-blind studies and double-blind studies have different
levels of awareness, either from the investigator or the study subject, or both.

Open Studies
In open studies, both the investigator and the study subject are aware of the
nature of the tested drug or vaccine. It is applied whenever it is neither possi-
ble nor desirable to conceal them from both. An example is a comparison
study between PUVA (psoralen1 UVA exposure) and subcutaneous etaner-
cept for skin psoriasis control. The limitation of open studies is bias genera-
tion both from the investigator and from the study subject.

Single-Blind Studies
In single-blind studies, the investigator, but not the study subject, is aware of
the tested drug or vaccine. It is applied whenever it is neither possible nor
advisable for the investigator to not be aware of them. An example is a com-
parison study between dobutamine and milrinone efficacy for pulmonary
capillary wedge pressure control during cardiogenic shock. The limitation of
single-blinded studies is bias generation due to self-suggestion by the
investigator.

Double-Blind Studies
In double-blind studies, neither the investigator nor the study subject are
aware of the tested drug or vaccine. It is applied to avoid bias generation due
to self-suggestion from both parties. An example is a comparison study
between placebo and a type 5 phosphodiesterase inhibitor for erectile dys-
function management.

Study Subject Allocation Method

The allocation method for the study can be either nonrandomized or
randomized.

Nonrandomized Studies
In nonrandomized studies, the investigator selects study subjects to be allo-
cated to one of the study groups according to pre-established criteria. It is
applied whenever study design demands selective allocation. An example is
an influenza vaccine efficacy study of two groups of subjects of different
age ranges (40 50 and 60 70 years). Nonrandomized studies are suscepti-
ble to bias due to misinterpretation of data.
10 PART | I Introduction

Randomized Studies
In randomized studies, the investigator randomly allocates study subjects
into study groups. The advantages of these studies are:
G Random error minimization through equal distribution of individual
characteristics between or among study groups.
G Systematic error minimization through avoidance of study subject selec-
tivity by the investigator.
G Minimization of the influence of uncontrolled variables through equal
distribution between or among study groups.

Follow-Up Methods
Follow-up methods for a study can include parallel or crossover studies.

Parallel Studies
In parallel studies, study groups progress in parallel during the investigation
until its end.

Crossover Studies
In crossover studies, the groups change their respective arms at a specific
point during the investigation. For example, a new ergot alkaloid derivative1
acetaminophen combination showed a significant decrease in migraine
symptoms in group A subjects compared to group B subjects, who took only
acetaminophen. After regimen swap, group B subjects showed the same
symptom decrease first associated with group A subjects, and the latter
presented the same symptoms level that group B subjects did in the first
phase of the study. Advantages of crossover studies are:
G They corroborate the findings of the first phase of the study by reprodu-
cing them in the second phase, reinforcing the final conclusion of the
study.
G They afford comparison between groups in a self-paired manner; that is,
groups A and B can be compared against themselves at the end of the
study, and this allows for greater biological homogeneity.
The limitation of crossover studies is the need for a washout period
between study phases.
Study design is the combination of the awareness level of the tested drug
or vaccine, the chosen comparative reference, study subject allocation and
follow-up methods, the planned duration of the study, the number of groups
and subgroups, and the exams involved all adjusted to provide an answer
to the investigator’s hypothesis. A randomized, double-blind, crossover,
placebo-controlled trial is an example of a desirable study design.
Chapter | 1 Study Type Determination 11

1.3 SUMMARY OF STUDY TYPES

The following classification is not definitive. Different study types not listed
and study types in which mixed characteristics are adopted can be implemen-
ted depending on the investigator’s objectives and resources.
G Epidemiological studies
G Ecological studies
G Cross-sectional studies
G Longitudinal studies
G Analytical studies
G Observational studies
- Case control studies
- Cohort studies
- Retrospective cohort studies
G Intervention studies
- Reference standard
Noncomparative studies
Comparative studies
- Relation between samples and within samples
Nonpaired sample studies
Paired sample studies
Self-pairing
Natural pairing
Artificial pairing
- Awareness of tested drug or vaccine
Open studies
Single-blind studies
Double-blind studies
- Study subjects allocation method
Nonrandomized studies
Randomized studies
- Follow-up method
Parallel studies
Crossover studies
Study type choice will aid in the decision-making process to determine
the most appropriate biostatistical model for answering the investigator’s
hypothesis.
Part | I Reader Resources 13

Part I Reader Resources

SELF-EVALUATION
1. Find the correct correspondence:
1. Comparative studies
2. Artificial pairing
3. Randomization
4. Crossover
5. Cross-sectional studies
a. Pattern of relation of a studied sample with another sample
b. A form of analyzing a condition frequency, used in epidemiologi-
cal studies
c. A type of reference standard used in intervention studies
d. A follow-up method used in intervention studies
e. A form of subject allocation in a study
A. 1 b, 2 e, 3 c, 4 d, 5 a
B. 1 c, 2 e, 3 a, 4 d, 5 b
C. 1 c, 2 a, 3 e, 4 d, 5 b
D. 1 a, 2 e, 3 d, 4 c, 5 b
2. Mark the incorrect affirmative:
A. Cohort studies can only be prospective.
B. Double-blind studies provide more reliable results than single-blind
studies.
C. Relative risk is an index that provides more robust information than
odds ratio.
D. In parallel studies, subjects from study groups can never be swapped.
3. Read the following study abstract and then answer the question:

With the aim of assessing the relation between C-section, vaginal delivery, and
Apgar scoring in the obstetrics ward of a county hospital during a 10-year
period, 2052 patient records were retrospectively evaluated. An Apgar #6 was
considered as the cutoff score, and Apgar scores .6 were considered as the
control. Premature infants and infants from mothers with reports of severe
abnormalities during prenatal care were not included.

How would you classify this study?

A. Randomized intervention study
B. Double-blind comparative study
C. Cohort study
D. Case control study
14 PART | I Introduction

ANNOTATED ANSWERS
1. C.
2. A. In retrospective cohort studies, two groups are retrospectively identified
and “prospectively” compared in a similar manner as in usual cohort studies.
3. D. The authors’ statement implies that establishing some degree of rela-
tionship between C-section and vaginal delivery procedures and Apgar
scoring is intended, and this excludes epidemiological studies. This study
is not prospective, and this clearly excludes intervention studies as well.
Observe the following diagram:

Time

Investigation direction

C-section

Group with Apgar score ≤6

(case)
Vaginal delivery

Initial
C-section Population

Group with Apgar score >6

(control)

Vaginal delivery

It can be seen that C-section and vaginal delivery groups were retrospec-
tively followed as case (Apgar #6) and control (Apgar .6), and that they
represented the “exposure” and “nonexposure” factors, respectively.
Hence, this study can be classified as a case control study type.

SUGGESTED READING
Basic statistics for clinicians, 1995. Can. Med. Assoc. J. ,[Link]. .
Estrela, C., 2001. Metodologia Cientı́fica: Ensino e Pesquisa em Odontologia. Editora Artes
Médicas, Sao Paulo, Brazil.
Everitt, B., 2006. Medical Statistics from A to Z: A Guide for Clinicians and Medical Students,
second ed. Cambridge University Press, Cambridge, UK.