Phẫu Thuật Tạo Hình Grabb & Smith’s (Bản Cập Nhật – Kevin C. Chung)

t-Test
The t-test, also known as student’s t-test, compares the means of two samples to tell you if the means
differ, and how much they differ from each other (if difference is significant). There are three primary
types of t-tests:
Independent t-test: Compares the means of two different (independent) samples (e.g., 6-month
postintervention scores in two different treatment groups).
Paired t-test: Compares the means of two related (dependent) samples (e.g., repeated measures of a
value in the same group, like pre-to-post-intervention). Also called the paired samples t-test,
dependent samples t-test, or repeated measures t-test.
Note: When using a paired t-test, the repeated measures tested should be from the same source,
such as the exact same patients, for both variables. Using data collected from slightly different
pre-to-post patient groups is unlikely to represent your sample’s true pre-to-post change.
Although it is tempting to include as many patients as possible in these types of analyses, this is
a misuse of the paired t-test. If your samples differ slightly, you may be able to use the
comparable nonparametric Wilcoxon signed-rank test. 10
One-sample t-test: Compares the mean of a single group against a known or hypothesized mean
(which you provide and should justify in your methods). It may also be used to find the statistical
difference between the sample mean and the sample midpoint. Also called the single sample t-test. 11
mebooksfree.net

ANOVA or Analysis of the Variance
When determining if differences exist between the means of ≥2 independent groups, ANOVA, or analysis
of the variance, can be used. ANOVA expresses this difference through the F statistic, a ratio of the
between-groups and within-groups variances. (Interestingly, the ANOVA can be used instead of the
independent t-test to compare two independent samples, with the resulting ANOVA F statistic equal to t
2 from the independent t-test.)
There are several data requirements you should check in order to use the one-way ANOVA correctly
(including confirming that your data are normally distributed). 12 Here we briefly address two: variance
and balanced sample size.
Variance: The amount of variability or range of values within a group. The within-groups variance
and between-groups variance should be approximately equal; otherwise, ANOVA results may not be
trustworthy. To analyze data with unequal variances between groups, it is better to use tests that do
not assume equal variances among populations, such as the Brown-Forsythe or Welch statistics. 13
Note: The Brown-Forsythe and Welch statistics are not nonparametric correlates to the ANOVA, that
is, the Kruskal-Wallis test, used for nonnormal data distributions. Brown-Forsythe or Welch statistics
are intended for use in parametric data sets with unequal variances.
Balanced sample size: When performing a one-way ANOVA, you should have at least six subjects (and
hopefully more) in each group; it is best to have balanced sample sizes in each comparison group, as
extremely unbalanced group comparisons increase the possibility of decreasing the validity of the
ANOVA’s F statistic.
We briefly describe four types of ANOVA tests here:
One-way ANOVA: Also called one-factor ANOVA, one-way analysis of variance, or between-subjects
ANOVA, the one-way ANOVA will tell you if there is a significant difference between the means of
any of the included groups. However, it will not tell you which group(s) is/are significantly different.
Additional post hoc testing (after running the ANOVA), such as with Dunnett C or the Tukey test, is
needed to determine this. 12,14 Dunnett C test compares each sample mean to a control mean (hence
the C) that you provide. If you do not have a control mean for comparison purposes, then the Tukey
or another post hoc test can be used.
Two-way ANOVA: A two-way ANOVA can be used to efficiently compare the impact ≥2 independent
variables may have on one dependent variable. This test provides information about the effect of
each independent variable by itself as well as the interaction between the two variables (e.g., effect
modification). Note: ANOVA techniques can be extended to three-way, four-way, and beyond,
provided your sample size is large enough.
Repeated measures ANOVA: The repeated measures ANOVA is used for variables measured more
than once for the same subject, like those that change over time. (In this case use of the standard
ANOVA is inappropriate because it does not account for the correlation between repeated measures
of the same variable.) The repeated measures ANOVA is used frequently in longitudinal studies
when subjects are measured repeatedly in regard to some outcome.
Multivariate analysis of variance (MANOVA): This test is essentially an ANOVA with several
continuous dependent variables. 14 It combines the selected dependent variables into a weighted
combination or composite variable to see how the new composite variable differs by groups or levels
of the independent variable.
mebooksfree.net

Regression Analysis
Regression analyses allow one variable (a dependent variable) to be predicted or estimated from one or
more other variables (independent variables). These mathematical models are one-directional, meaning
that the independent variable(s) can be used to predict the dependent (outcome) variables, but not the
other way around. Regression analysis is a useful tool in medical research and clinical practice because it
helps researchers determine who might be at risk for various illnesses, behaviors, etc., based on different
exposures. An example is predicting who may be at risk for postoperative complications based on their
preoperative comorbidities.
Several pieces of data are derived from regression analyses to describe the predictive relationship
between your independent and dependent variables. Three of these are slope, correlation coefficient, and
P-value:
Slope: Indicates where different patients may be mapped in your predictive model. In univariate
regression analyses, the slope indicates how much your dependent variable will change with each
unit change in your independent variable. In multiple linear regression analyses, the slope indicates
how much your dependent variable will change as a result of the combined influence of all your
independent variables.
Correlation coefficient (r): Also called r value, or Pearson product-moment correlation coefficient. This
indicates the strength of a predictive relationship. Values for r range from (+)1 to (−)1, where (+)1
represents a perfect positive correlation, (−)1 represents a perfect negative correlation, and zero
represents a perfect lack of any correlation. Correlation coefficients of ≥(+)0.7 or ≤(−)0.7 may be
considered strong. 15
P-value: Determines how likely it is that the correlation between your independent and dependent
variables is due to chance. P-values should always accompany the r values.
Several types of regression analyses are commonly used in biomedical studies. We briefly describe them
here with the goal of illustrating when each can be used.
Univariate or simple linear regression: Also called linear regression. Analysis performed using one
independent variable to predict the variable of interest (dependent variable). 15,16 In simple linear
regression, the outcome (dependent) variable is continuous. 16 These analyses are often performed
first to determine which independent variables warrant further investigation using more complex
and realistic regression models (e.g., multiple linear regression).
Multiple linear regression: Also called multivariable linear regression. Analysis performed using ≥2
independent variables (also called covariables or covariates) to predict the variable of interest
(dependent variable). 15 In multiple linear regression, the outcome (dependent) variable is
continuous. 16
Multivariate regression: Differs from multiple or multivariable linear regression in that it deals with
multiple dependent variables (e.g., repeated measures or clustered data) to predict an outcome. 16
Multiple logistic regression: Useful for identifying potential confounders. This differs from multiple
linear regression in that the outcome (dependent) variable of interest is dichotomous. In these
models, the independent variables are called covariates and can be categorical, ordinal, or
continuous.
Despite its implication of a directional influence, regression in of itself does not prove causality. 15
It is important to note that multivariate regression and multivariable regression are not the same. This can
be confusing as the two terms are sometimes used interchangeably although they represent two different
mebooksfree.net

analyses.
mebooksfree.net

Survival Analysis
Understanding how long patients may remain disease-free following treatment or survive given certain
predisposing factors, or how long an intervention may be effective, can be very informative for patients
and physicians. Survival analysis, or what is now more often referred to as Kaplan-Meier curves, are one
way to perform such analyses.
Kaplan-Meier curves were developed to help account for incomplete observations and differing survival
times. 17,18 This type of analysis is also called time-to-event analysis as successful survival (the event) does
not need to be limited to the absence of mortality. Instead, the event of interest can be any of many
different outcomes including a complication, the efficacy of an intervention, or resolution of an adverse
event, with the analysis observing the time it takes for the variable of interest to develop, occur, or
resolve. 17,18
Although there are a few methods to conduct survival analyses, the Kaplan-Meier method is generally
preferred. When used, data that should be included are the time-to-event (expressed as the median, not
the mean) confidence limits (as a measure of variability) and the number of patients who are included at
each point of survival (e.g., censoring versus duration of known survival). It should be noted that the
further away in time the results are from the beginning of the study, the greater the censoring
(attrition/dropout, death) rate and the more approximate the curve calculation becomes. Finally, when
comparing two or more survival curves, a suitable rank test should be used. 17
mebooksfree.net

Summary of Simple Parametric and Nonparametric Tests
There are some straightforward situations in which a nonparametric test can be substituted for a
parametric test. These are included in Table 19.1. However, when exploring options for more complex
ANOVAs and regression analyses, several additional factors, such as variable type, factor into the type of
test that should be used.
TABLE 19.1
Parametric and Nonparametric Tests Used for Various Statistical Comparisons 8
Analysis You Want to do
Example
Parametric Test
Nonparametric Test
Compare means between two independent
samples or groups
Group A versus
group B
Independent t-test (one-
tailed or two-tailed)
Mann-Whitney U test
Compare two quantitative measures taken from the
same individual
Pre-to-post-
intervention
differences
Paired t-test
Wilcoxon signed-rank test
Compare means between three or more samples
Group A versus
group B versus group
C
Analysis of variance
(ANOVA), or one-way
ANOVA
Kruskal-Wallis test, or one-
way ANOVA on ranks
Compare the distributions of a discontinuous
variable in two or more independent samples
Is x more likely to
occur given y?
Chi-square test
Fisher exact test
Now that we have reviewed several commonly performed types of analyses, we will review when you
should develop your statistical plan within the context of your protocol or study design.
mebooksfree.net

Analytic Approaches: A Priori Versus Post Hoc
The time at which you design your statistical plan, meaning before conducting the study or after all data
are collected, is important in identifying the presence or absence of true effects within your data set. We
review best practice guidance for including appropriately performed a priori and post hoc analytical
designs to ensure your findings are as trustworthy and meaningful as possible.
mebooksfree.net

A Priori Analyses
Definition
Best practices for research recommend designing your data analysis plan beforehand (a priori), or when
you plan the project or write up a proposal. This is considered the purest type of analysis and is intended
to achieve several goals:
Provide evidence regarding whether your hypotheses should be accepted or rejected.
Demonstrate the least amount of bias in the results obtained.
Prevent inappropriate use of statistics to obtain statistically significant results.
Help determine sample size and study power sufficient to detect an effect size, if present.
Help inform appropriate budget development based on determination of number of subjects needed.
mebooksfree.net

Appropriate Use
A priori analyses are designed in parallel with protocol development. Details of the analysis should be
included in a dedicated section of the protocol document (or grant proposal). The analysis should then be
completed according to this description following collection of the necessary data (e.g., interim analysis,
final analysis). Being faithful to your statistical approach is vital to minimizing bias and producing
reproducible results.
One concern when following a strictly outlined data analysis plan for which the results are not
statistically significant is that the study was meaningless (or worse: not publishable). However, this is not
necessarily the case. If you have carefully designed the study, meaning that it was appropriately powered
and you used appropriate statistical tests, your negative results are likely important and should be shared
with the field. a Certainly, you should not panic and turn your analysis into a fishing expedition intended
to identify statistically significant results. Although this is tempting given the time and costs to conduct
research, it is an inappropriate use of statistics, will potentially introduce false data into the literature and
clinical practice, and may ultimately hurt patient care. However, in certain circumstances performing
exploratory, post hoc analyses on an existing data set can be informative.
mebooksfree.net

Post Hoc Analyses
Definition
Post hoc analyses are any statistical analyses not described previously (a priori) in the protocol and
performed after data collection has been completed. Generally, these analyses do not address the study’s
primary or secondary aims. Instead, they help generate future hypotheses. Importantly, these results do
not provide proof regarding these newly generated hypotheses. 21
mebooksfree.net

Appropriate Use
Certain post hoc analyses are an integral part of your a priori analysis plan. Examples include post hoc
testing following ANOVA to determine which means differ significantly from each other, or conducting a
Bonferroni correction. 22 Although these types of tests technically meet the definition of post hoc analysis,
chances are you were able to plan for these prior to study implementation. This section is intended to
address unplanned analyses researchers would like to perform based on the collected data after the study
has been finished.
Researchers can conduct post hoc analyses and include these in publications, provided that this is clearly
stated in the study methods, results, and discussion of these results. 1 After completing a study and
reviewing your data, additional questions may arise. These may have been anticipated based on the
literature and included as exploratory study aims, or they may arise from the data itself. In these
situations, it is acceptable to perform exploratory analyses. The objectives of conducting post hoc analyses
include the following 21 :
Identification of unsuspected findings or associations
Development of new hypotheses
Informing the design of future prospective studies (e.g., inclusion/exclusion criteria, study design,
statistical power)
Additionally, retrospective studies, meta-analyses, and secondary analysis of existing data sets are all
subject inherently to post hoc analysis. Authors need to acknowledge the limitations of their data sets or
source data used (e.g., existing data sets, retrospective studies) and level of evidence of included studies
(e.g., meta-analyses), as relevant. This knowledge helps readers determine how much weight to give to
such studies. Finally, just as with your a priori analysis, meticulous documentation of your post hoc
analysis to demonstrate your process and facilitate reproducibility helps readers determine the relative
impact of your study.
mebooksfree.net

Adjusted Analyses: to Control or Not to Control
Despite the best planning, it is not always possible or practical to design a study to minimize your
between-groups variation. In these situations, adjusted analyses may be used. Study results may be
expressed as unadjusted (aka, crude) or adjusted. Adjusted refers to controlling for potential overestimation
of the influence of any single independent variable on your outcome when several variables may be at
work. This can also be thought of as controlling for bias. In addition to adjustment, we address a few
factors that can influence your findings and make adjustment necessary, such as confounding, effect
modification, and interaction.
mebooksfree.net

Adjustment
Adjustment is needed when several independent variables may be influencing your outcome of interest.
To perform an adjusted analysis, you first need to know how to choose the independent variables to
include. This can be done by (1) reviewing the literature to determine what variables have been identified
in other similar studies (the usual suspects) and (2) performing a preliminary unadjusted or univariate
analysis to help identify variables that may be affecting the outcome variable. Because unadjusted
analyses are prone to alpha (type I) error (overestimating significance), it is normal for certain variables
that looked promising after unadjusted analysis (have significant P-values) to become statistically
insignificant after an adjusted analysis.
Adjusted analyses reduce the combined alpha value to mitigate the chance for alpha error. This provides
more realistic estimations of the relationships between variables. One example of such a multiple-
comparison correction is the Bonferroni correction. 23
Adjusted analysis can be very helpful when comparing two different groups. Here different means
unmatched. The advantages of these analyses are you may use them on groups with nonidentical sample
sizes and to account for characteristics you were unable to match based on study design, recruitment, or
your source data.
mebooksfree.net

Overadjustment and Unnecessary Adjustment
When faced with many interesting variables, it can be challenging to limit the number to include for an
adjusted analysis. However, selecting too many variables by which to adjust your analysis can result in
overadjustment and/or unnecessary adjustment. Although they differ slightly and their influence is
relative, b both can introduce bias. 24,25 .
Overadjustment: Occurs when your controlling variable introduces, rather than eliminates, bias. 25 It
can also cause problems by subsetting your groups to such an extent that the results are not
meaningful and can put researchers at risk for beta error (false-negative results). 24 Simple first steps
to help prevent overadjustment are (1) taking care to select independent variables based on
knowledge (e.g., those substantiated by the literature or indicated by your data set via univariate
analysis) and (2) having ≥10 subjects for each dependent variable included in a multivariate adjusted
analysis. 24
Unnecessary adjustment: Similar to but different from overadjustment. It is unlikely to introduce
additional bias (aka, bias-neutral adjustment) but can affect precision, here meaning a clear
understanding of the influence of the exposure on the outcome. 24,25 Unnecessary adjustment occurs
when the variable used to control an analysis is a variable (1) unrelated to the relationship being
investigated, (2) that only causes the exposure, (3) that is created as a result of the exposure but is not
involved in the causal (exposure-to-outcome) pathway, or (4) that causes the outcome but is not
involved in the causal (exposure-to-outcome) pathway. 24 Importantly, its role is relative, b can be
subtle, and may go unnoticed.
mebooksfree.net

Confounding, Effect Modification, and Interaction
Confounding, effect modification, and interaction are three similar yet different events in which two (or
more) variables influence the outcome of interest in a way that makes the resulting apparent association
inaccurate.
Confounding bias: Comes from a variable that influences both the outcome of interest and the
independent variable (or exposure) being assessed, but is not involved in the pathway between the
two. This confounding factor makes an inaccurate association between the exposure and the
outcome. An example is breast cancer treatment assignment and postoperative complications. As
patients’ cancer stage increases, they are more likely to receive more aggressive treatment. Such
patients would have more complications due to both their more aggressive treatment and their
advanced cancer stage, not merely treatment alone.
Although not a statistical panacea, adjusted analyses help control for potential confounders. 26
Another way to minimize confounding is by using propensity scores. 26,27 Propensity scores are
intended to replace many different pieces of baseline data with one summary score and are
calculated using a specific type of logistic regression model. 27 The propensity scores can then be used
to help match patients between your groups, or as an independent variable for which you can adjust
your analysis. In other words, propensity score analysis attempts to retrospectively recreate
randomization. 27 However, these potential solutions only work for known confounders and cannot
address unknown confounders. 27
Effect modification: Comes from a variable that influences the outcome but not the exposure and
occurs when an exposure has a different effect in different subgroups. This third variable can
potentially mask (conceal) or increase (sometimes multiplicatively) the true impact of the exposure
on the outcome. In this situation, analysis of pooled data can be misleading. Effect modification can
be checked using multivariable analysis or stratification to control for the potential effect modifier.
Tools that can be used to measure the impact of effect modification include risk ratio, odds ratio, and
risk difference. 28
Interaction: Occurs when two exposures or interventions influence an outcome. 29 Although similar,
interaction and effect modification are different; while they often are, the two terms should not be
used interchangeably. 29
mebooksfree.net

Stratification
Another method by which you may control your comparative analyses between populations is
stratification. It allows you to further subdivide your comparison groups to better match control and
treatment groups. 27 Although stratification can work well in large populations with a few variables (and
thus, a few strata), it rapidly falls apart when there are many baseline characteristics for which you want
to stratify. One way you may address the problem of excessive population subsetting is to stratify by
propensity scores. 27
mebooksfree.net

Presenting Data
After you have designed your analysis; selected your tests; completed the study, data collection, and
analysis; and assessed your results for bias, you get to the final reward: presenting your hard-earned data
for publication. When presenting data, it is helpful to remember a few things:
You or your team is the expert on the study you just completed, so be sure to present your audience
with a complete picture of the study.
Keep in mind the most important or interesting finding(s) and how to present them in the most
meaningful way.
Use simple language, particularly if you are presenting something complex; your audience will
thank you for it.
We have included some tips here for how you can present a comprehensive view of your findings in
concise and meaningful ways.
mebooksfree.net

Formatting Findings
Statistical results tend to be most meaningful when several measures describing the same finding are
presented together. This provides a more comprehensive picture of your sample, its relative
heterogeneity or homogeneity, and the potential strength of the effect you found. Some recommended
formats for presentation of statistical findings follow:
Mean ± SD, n = __ (95% CI:_-_), P-value = __
Median (IQR), (95% CI:_-_), P-value = __
Odds ratio (OR), (95% CI:_-_), P-value = __
Writing your findings in this way or including these data in a well-conceived table can improve your use
of journal space and be more effective than wordier verbal descriptions. This efficiency can also enable
you to share more information about your findings by freeing up your word count.
mebooksfree.net

Correlation Versus Causation
When discussing findings there is some confusion surrounding the terminology used to indicate
relationships between variables. Here we briefly clarify correlation and causation.
Correlation (or association): This term indicates that there appears to be a relationship between your
exposure (independent variable) and outcome of interest (dependent variable), but that you are
unable to determine if one causes the other. Correlation and association are important when
conducting studies in which it is not possible to collect every variable that could affect an outcome
and to help indicate what variables should be investigated further through more vigorous studies
(e.g., controlled clinical trials).
Causation: This term indicates that there is a quantifiable relationship between your exposure
(independent variable) and outcome of interest (dependent variable), and that you can prove that
one is causing the other. Causation is difficult to prove in complex interactions, such as situations in
which many different exposures may contribute to an outcome over many years.
mebooksfree.net

A Word About Data Collection
When discussing statistics, it is necessary to touch upon data collection, or considerations regarding your
source data set. For one thing, your findings can only ever be as good as the data used to derive them. An
example of this is a beautifully designed meta-analysis of level 3 evidence papers; regardless of how
perfectly the meta-analysis is performed, the resulting paper’s level of evidence cannot exceed that of its
source material (per this example, level 3 evidence).
The old term still applies: garbage in, garbage out. Therefore, collecting good data and understanding what
your data can or cannot tell you are vital to understanding what you can ask of it.
mebooksfree.net

Potential Database Tools
Research can be more expensive than anticipated, whether conducting a relatively small retrospective
review or a phase 3 clinical trial. This causes many people to make use of free (or prepaid) resources for
the purposes of storing data. Though understandable, researchers must be aware of the strengths and
weaknesses of commonly used tools. We provide a summary of some commonly used electronic data
collection tools and their strengths and weaknesses in Table 19.2. This is not intended to be
comprehensive; rather, this table touches upon points meant to help researchers in selecting the tool most
appropriate for them.
TABLE 19.2
Strengths and Weaknesses of Some Common Data Collection Tools
Data
Collection
Tool
Strengths
Weaknesses
Excel 30,31
Inexpensive or free, easily accessible, nearly universal across operating systems
and versions, can be password-protected.
Easily corrupted (e.g., missorted),
users have to create their own data
dictionary.
REDCap 32
Web-based for easy access between team members, has tiered access for users,
inexpensive or free, easily updated by study team if changes need to be made to
DB design, creates data dictionary for users. Designed for research use. Can be
HIPAA-compliant.
Flat (not relational) database, will
need to pay for database design/setup
if no team members know how to do
this.
Access 33
Included in many MS Office packages, has capacity to hold relatively large
amounts of data, creates data dictionary for users.
Designing queries can be taxing, not
inherently HIPAA-compliant.
FileMakerPro
Has tiered access for users, DB can be changed easily by study team if needed,
can be made to create data dictionary for users.
Paid licenses are needed for all users.
Will need to pay for database
design/setup if no team members
know how to do this.
Regardless of the tool used, some forethought and design need to go into creating your data collection
tool. Once created, it should be tested. You can do this by entering data using a few sample or test cases.
This step can be very helpful in the long-term as it will illustrate any problems that may exist (e.g.,
missing or redundant data fields, imprecise prompts or descriptors, forms that do not work correctly) and
the changes that may be needed. 30 It is much more efficient and cost-effective to recognize and correct
any potential problems early than after all your data have been collected.
Finally, although it can be challenging to identify funding on the front end of a project, it may be more
cost-effective to pay for a database that is unlikely to corrupt data or that can provide structure to enable
easy and uniform data entry by individuals of varying medical or research knowledge. Realizing halfway
through or at the end of a project that your source data and/or analyses are corrupted or flawed is enough
to make anyone give up research.
mebooksfree.net

Data Collection
Like other disciplines, there is a science to data collection to ensure accuracy, completeness, and patient
privacy.
Good practices for data collection for industry, as published in the Food and Drug Administration’s good
laboratory practice, follow the acronym ALCOA to identify elements of data quality. 34 ALCOA stands
for:
Attributable: Data entries and any changes can be traced back to the person who made the entries or
changes.
Legible: Data entries and any changes are easily read, even after being corrected. This pertains to hard
copies (paper) and electronic records (entry history).
Contemporaneous: That data collection and entry occurs as close in time as possible to the observation.
Original: The first and therefore generally considered to be the most accurate and reliable recording
of data.
Accurate: The recorded data correctly reflect the observation.
Following these principles puts you on the right track for high-quality data collection that you can trust
months or years later.
mebooksfree.net

Preexisting Data Sets
There can be great value in secondary analysis of preexisting data sets. Some examples are:
Access to sample sizes to which you might not otherwise have access
Leverages preexisting resources making research more affordable
May consist of data from more than one institution, state, or country
Larger sample size may make study population more comparable to general population
However, researchers must be aware of the potential drawbacks of preexisting data sets. Some examples
are:
Potential for missing data
Data collected may change from year to year, making multiyear comparisons challenging or
infeasible
May require significant data preparation and cleaning prior to analysis
May not include variables most informative for testing your hypothesis
When using a preexisting data set, you must familiarize yourself with the data, how it was collected, and
obtain and review the user’s guide or primary protocol documenting how the data were collected. Taking
these steps will educate you as to what questions you can ask appropriately of the data set. Design of an
appropriate research question, thoughtful design of a post hoc analysis plan, and transparency in stating
how the study was conducted are important parts of reporting findings for studies conducted using these
resources.
mebooksfree.net

Data Preparation
Although it is best to proactively prevent data entry errors, some entry errors are inevitable. Data
preparation, also called data cleaning, is a crucial and potentially time-consuming part of ensuring your
data are as amenable to analysis as possible.
Data preparation is when you review and format your data in a spreadsheet to make it compatible with
the program the statistician will use. 30,35 This is intended to enable direct importing of the data from a
shared spreadsheet into the statistical software program without any unintended changes being
performed automatically by the analysis software on your data. It can include many different levels of
review, from formatting to editing abnormalities. Importantly, this is not manipulation of your data:
corrections may be made, but data values should not be changed. Changes should be tracked such that it
is possible to compare the originally exported data set with the cleaned version. 35 Some examples of data
cleaning activities:
Indicating missing data versus zero values. Blank fields can present problems during analysis. Therefore,
missing data may have a specific entry such as a “.” (period) instead of remaining blank. Fields
whose values are left unaddressed may be assigned values by the analysis software and are likely to
yield inaccurate results.
Variable name length. Software programs often have character limits for their field names. This can
cause some overly long names to be cut off and potentially unrecognizable. Checking, editing names
down, and tracking changes in your data dictionary can prevent this from occurring.
Data accuracy. Ensuring that data are correctly and consistently displayed (no typographical errors,
use of letters in number fields, revisiting abnormal values to ensure they are real results, etc.).
Additional variable categorization. Although data were collected on a precise level, it can be helpful to
recategorize certain variables into larger classes prior to analysis (e.g., BMI into BMI classes). Such
choices may be made based on the literature.
It is important for the study team to complete this work because (1) it can be time-consuming and (2) you
are more likely than your statistician to be an expert on your data and how it should be included in
analyses.
To best streamline the data preparation process, ask your statistician about desired formatting or
formatting requirements before preparing and sharing your data.
mebooksfree.net

Data Dictionaries
A data dictionary is a comprehensive document that includes the full and abbreviated names of all of
your data fields, the possible answers for coded or categorized data, and indications for which fields are
labeled as free-text fields. This document acts as a translation book to ensure you and your statistician are
referring to and using the same field(s) when discussing your study and performing analyses.
You must be aware when selecting a data management tool which ones will and will not generate a data
dictionary. Importantly, if your data management tool does not create a data dictionary, you will need to
create one and keep it updated. This can require a considerable amount of effort and should not be
underestimated.
mebooksfree.net

When to Consult a Statistician
Many individuals new to research ask when they should consult a statistician. To misquote a phrase, you
should consult one early and often. As previously stated, a carefully designed study including an a priori
statistical plan is vital to producing meaningful results or convincing a review committee to fund your
project. 36
mebooksfree.net

Finding a Knowledgeable Statistician
Though not necessary, it is helpful to work with statisticians who understand your area of study or who
can understand it based on similar work in a different field. Familiarity often means that the statistician
will be able to apply his or her time more efficiently to your project. If the statistician is unfamiliar with
your area of study, be prepared to orient him or her. This will go into how much time you need to spend
with the individual during the planning phase.
First, there may be several types of statisticians who are available to you:
Academics (master’s or PhD level)
Students studying statistics (earning their MPH, PhD)
Students formally overseen by PhD-level statisticians
Private or ad hoc personnel
A member of the team who is comfortable with statistics
The complexity of your study, importance of results, and availability of funding may dictate the level of
statistician you approach. It may also be advisable to obtain references for their work and/or to look up
past publications to determine if the statistician’s past experience aligns with your research needs. There
are advantages and drawbacks to each of the aforementioned individuals: be aware that going with the
inexpensive option may require more oversight and knowledge on your part, whereas going with a
highly qualified individual may require significant wait-time prior to gaining assistance and payment for
their time.
mebooksfree.net

Keeping the Statistician On-Board
There are several approaches researchers can take to engage a statistician. It is best to ask directly
regarding the relative importance of hourly compensation, documentation of utility across an
organization, and authorship as any of these may apply.
mebooksfree.net

Compensation for Time
As with all skilled positions, hourly rates for statisticians can vary depending on their level of education
and years of experience. If you are looking for a less costly but reasonably reliable option, blended rates
may be available for students who are overseen by a PhD-level statistician. It is important to understand
how skilled a statistician you need for your project so that your efforts are adequately supported.
mebooksfree.net

Authorship
Depending on whether your statistician is an academician interested in the area of your study, you may
be able to offer authorship for their involvement. Certain types of studies rely heavily on the statistician
and warrant prominent authorship. Such studies include secondary analyses of preexisting databases
(e.g., NSQIP, NIS) or in-depth meta-analyses of multiple studies. In these cases, the statistician’s
authorship may need to be as high as co-first authorship. If the statistical consultation is used to confirm
your statistical plan, you may only need to include the statistician’s contribution in the acknowledgments
section. It is important to fully understand the effort request you are making of the statistician so that you
can make an appropriate offer of authorship.
mebooksfree.net

Documenting Organization Utility or Benefit
In certain academic environments, biostatistics core facilities are available. These can be funded in any
number of ways (indirect costs, federal grants, etc.) and often include reporting metrics to justify ongoing
funding. If this is the case, a certain number of in-kind hours may be available to faculty, students, or
staff. (In the case of grant design, these free hours are also intended to assist with budgeting for future use
of the statistician’s time when data collection is complete.) In these situations, a combination approach
that includes compensation and authorship in published articles may be needed: be sure to ask.
mebooksfree.net

Statistical Plans: Get Them in Writing
Research, and particularly prospective clinic trials in human subjects, can take much longer than
anticipated to complete. For example, delays in IRB approvals, start-up, and recruitment; dropout or lost-
to-follow-up rates; and stringent inclusion/exclusion criteria can all contribute to slower study completion
timelines. Given all the other details to which one must pay attention during a research study or clinical
trial, it is easy to forget one’s analysis plan.
If you have not written out your a priori analysis plan, chances are you will not remember it when you
finally have data to analyze. Additionally, you need to have the analysis plan in writing in case the
original biostatistician is no longer available to you at the time the study is completed. Because statistics
can be somewhat open to interpretation, different statisticians may create different yet valid approaches
for analyzing the same set of data. This is fine unless it renders your collected data inappropriate.
mebooksfree.net

Questions
1. PRS has published a study examining the pre-to-post change in patient satisfaction with breast
reconstruction comparing a new implant sling technique (n = 13) with conventional implant
placement (n = 28). What is the most appropriate test to use to determine if a statistically significant
difference in satisfaction exists between groups?
a. Paired t-test
b. Independent t-test
c. Wilcoxon signed-rank test
d. One-way ANOVA
2. Which of the following scenarios would require an increase in the sample size?
a. Changing the significance level from 0.05 to 0.01
b. Changing the power of the study from 90% to 80%
c. Determining that the difference in the means is less than originally calculated
d. Determining that the standard deviation of the means is less than originally calculated
3. A P-value of 0.05 indicates which of the following?
a. 5% chance the null hypothesis is correct
b. 5% chance the null hypothesis was incorrectly rejected
c. 95% chance the null hypothesis was correctly accepted
d. 95% chance the null hypothesis is correct
4. A study is performed to determine the outcomes between two different breast reconstruction
techniques. One technique is performed by surgeon A at hospital X and the second technique is
performed by surgeon B at hospital Y. Surgeon A has significantly better patient satisfaction scores
based on a P-value of less than 0.01. Which type of bias is at risk for affecting the data?
a. Skewing
b. Post hoc analysis
c. Confounding bias
d. Overadjustment
5. A study team conducts a prospective, randomized, placebo-controlled trial designed to determine
the effect of an experimental medication on flap necrosis rates. While completing the analysis of the
study aims, the team notices an apparent correlation between participant BMI and postoperative
infection rates. After conferring with the protocol principal investigator, the team decides to publish
these additional findings in a separate manuscript. Which type of bias is at risk for affecting the data?
a. Skewing
b. Post hoc analysis
c. Stratification
d. Overadjustment
1. Answer: c. The sample size is small indicating that a nonparametric analysis should be performed
because it cannot be assumed that the data are normally distributed. The pre-to-post change in
patient satisfaction indicates that the data are dependent (repeated measurements in the same
patients) and should be analyzed as such.
2. Answer: a. When performing a power calculation four pieces of information are needed: the event
rate, the smallest clinically important difference, power (to detect the difference), and significance
level. Decreasing power (answer b) reduces the sample size needed but increases the chance of beta
error. Lowering the significance level (answer a) decreases the likelihood of alpha error and increases
the number of subjects needed in the study. This is because you are seeking a greater level of
mebooksfree.net

confidence with which to correctly reject the null hypothesis (a 99%, or 99 in 100, chance).
3. Answer: b. A P-value is a measurement of the strength of the evidence against a null hypothesis;
the smaller the P-value, the stronger the evidence that you should reject the null hypothesis.
However, the P-value indicates the percent (%) chance that the null hypothesis would be incorrectly
rejected, in this case a 5%, or a 1 in 20, chance.
4. Answer: c. Many different factors may be influencing differences in patient satisfaction between
the two groups of patients being compared, including the surgeon’s skill level and interpersonal
skills, the breast reconstruction technique employed, the hospital, the clinic and nursing staff, and
more. Although this may serve as an initial hypothesis-generating study to help indicate how to
better ascertain what precisely is causing these differences, this study alone is not very informative.
5. Answer: b. The analysis exploring a correlation between participant BMI and postoperative
infection rates was not a part of the originally approved protocol’s specific aims or a priori statistical
plan. By definition, this analysis was done post hoc and is subject to all the shortcomings inherent to
post hoc analyses. Although it can be acceptable to publish post hoc findings, the study team needs
to clearly state this fact and may not assume that a post hoc finding has proven a true correlation.
Instead, post hoc findings can be considered hypothesis-generating and informative to future study
designs.
mebooksfree.net

References
Moher, D, Hopewell, S, Schulz, KF, et al. CONSORT 2010 explanation and elaboration: updated guidelines for reporting parallel group
randomised trials. Int J Surg. 2012;10(1):28-55.A published statement developed and refined by experts in clinical research to guide investigators
toward best practices when designing and conducting clinical trials. This statement is widely endorsed by leading medical journals and international
editorial groups.
Ioannidis, JPA. The proposal to lower P value thresholds to .005. JAMA. 2018;319(14):1429-1430.
Leek, J. On the scalability of statistical procedures: why the p-value bashers just don’t get it. Simply Statistics, A Statistics Blog by Rafa
Irizarry, Roger Peng, and Jeff Leek.
bashers-just-dont-get-it/. Published February 14, 2014. Accessed May 9, 2018.
Wasserstein, RL, Lazar, NA. The ASA’s statement on p-values: context, process, and purpose. Am Stat. 2016;70(2):129-133.The American
Statistical Association’s statement articulates in nontechnical terms a few select principles that could improve the conduct or interpretation of
quantitative science, according to widespread consensus in the statistical community.
Kirby, A, Gebski, V, Keech, AC. Determining the sample size in a clinical trial. Med J Aust. 2002;177(5):256-257.
Sedgwick, P. What is intention to treat analysis? BMJ. 2013;346:f3662.
AButhman. Dealing with non-normal data: strategies and tools. iSixSigma.
non-normal-data-strategies-and-tools/. Accessed May 7, 2018.
Greenhalgh, T. How to read a paper. Statistics for the non-statistician. I: different types of data need different statistical tests. BMJ.
1997;315(7104):364-366.The first in a series of articles excerpted from the book, How to Read a Paper: The Basics of Evidence-Based Medicine.
This article provides clinicians with a checklist of preliminary questions to help assess the statistical validity of a scientific publication.
Mean and Standard Deviation. BMJ. readers/publications/statistics-square-one/2-mean-and-
standard-deviation.
Statistical Consulting, Kent State University Libraries. SPSS Tutorials: Paired 
 Samples t Test.
Updated 
 May 24, 2018. Accessed May 25, 2018.
Statistical Consulting, Kent State University Libraries. SPSS Tutorials: One Sample t Test.
Updated May 25, 2018. Accessed May 25, 2018.
Statistical Consulting, Kent State University Libraries. SPSS Tutorials: One-Way ANOVA.
Updated May 25, 2018. Accessed May 25, 2018.
Reed, JF III, Stark, DB. Robust alternatives to traditional analysis of variance: Welch W*, James J I*, James J II*,Brown-Forsythe BF*. Comput
Methods Progr Biomed. 1988;26(3):233-237.
Bird, KD. Simultaneous contrast testing procedures for multivariate experiments. Multivariate Behav Res. 1975;10(3):343.
Greenhalgh, T. How to read a paper: statistics for the non-statistician. II : “Significant” relations and their pitfalls. BMJ. 1997;315(7105):422-
425.The second in a series of articles excerpted from the book, How to Read a Paper: The Basics of Evidence-Based Medicine. This article
provides clinicians with a checklist of preliminary questions to help assess the statistical validity of a scientific publication.
Hidalgo, B, Goodman, M. Multivariate or multivariable regression? Am J Public Health. 2013;103(1):39-40.
Mathew, A1, Pandey, M, Murthy, NS. Survival analysis: caveats and pitfalls. Eur J Surg Oncol. 1999;25(3):321-329.
Rich, JT, Neely, JG, Paniello, RC, et al. A practical guide to understanding Kaplan-Meier curves. Otolaryngol Head Neck Surg. 2010;143(3):331-
336.
Duyx, B, Urlings, MJE, Swaen, GMH, et al. Scientific citations favor positive results: a systematic review and meta-analysis. J Clin Epidemiol.
2017;88:92-101.
Carroll, HA, Toumpakari, Z, Johnson, L, Betts, JA. The perceived feasibility of methods to reduce publication bias. PLoS One.
2017;12(10):e0186472.
Elliott, HL. Post hoc analysis: use and dangers in perspective. J Hypertens Suppl. 1996;14(2):S21-S24.
EWWeisstein. “Bonferroni correction.” From MathWorld – A Wolfram Web Resource.

Guyatt, G, Walter, S, Shannon, H, et al. Basic statistics for clinicians: 4. Correlation and regression. CMAJ. 1995;152(4):497-504.
Schisterman, EF, Cole, SR, Platt, RW. Overadjustment bias and unnecessary adjustment in epidemiologic studies. Epidemiology.
2009;20(4):488-495.
VanderWeele, TJ. On the relative nature of overadjustment and unnecessary adjustment. Epidemiology. 2009;20(4):496-499.
Cook, JA, Ranstam, J. Statistical models and confounding adjustment. Br J Surg. 2017;104(6):786-787.
Adamina, M, Guller, U, Weber, WP, Oertli, D. Propensity scores and the surgeon. Br J Surg. 2006; 93: 389-394.
VanderWeele, TJ, Robins, JM. Four types of effect modification: a classification based on directed acyclic graphs. Epidemiology. 2007;18(5):561-
568.
VanderWeele, TJ. On the distinction between interaction and effect modification. Epidemiol. 2009;20(6):863-871. Errata in: Epidemiology.
2011;22(5):752, and 2010;21(1):162.
Juluru, K, Eng, J. Use of spreadsheets for research data collection and preparation: a primer. Acad Radiol. 2015;22(12):1592-1599.
Elliott, AC, Hynan, LS, Reisch, JS, Smith, JP. Preparing data for analysis using Microsoft Excel. J Investig Med. 2006;54(6):334-341.
mebooksfree.net

Harris, PA, Taylor, R, Thielke, R, et al. Research electronic data capture (REDCap): a metadata-driven methodology and workflow process
for providing translational research informatics support. J Biomed Inform. 2009;42(2):377-381.
Schneider, JK, Schneider, JF, Lorenz, RA. Creating user-friendly databases with microsoft access. Nurse Res. 2005;13(1):57-75.
TJKuhn. ALCOA: a standard for evidence. GxP zone: good practice for the pharmaceutical industry from the quality assurance perspective.
Posted July 23, 2008. Accessed June 28, 2018.
Van den Broeck, J, Cunningham, SA, et al. Data cleaning: detecting, diagnosing, and editing data abnormalities. PLoS Med. 2005;2(10):e267.
De Muth, JE. Preparing for the first meeting with a statistician. Am J Health Syst Pharm. 2008;65(24):2358-2366.
mebooksfree.net

PA R T I I
mebooksfree.net

Skin and Soft Tissue
20. Skin Care and Benign and Malignant Dermatologic Conditions
21. Thermal, Chemical, and Electrical Injuries
22. Principles of Burn Reconstruction
23. Radiation and Radiation Injury
24. Lasers in Plastic Surgery
mebooksfree.net

C H A P T E R 2 0
mebooksfree.net

Skin Care and Benign and Malignant
Dermatologic Conditions
John H. Bast and Alex K. Wong
Key Points
Understand the role of Fitzpatrick skin types and their role when choosing a chemical peel regimen
or laser treatment
Understand the spectrum of dermatitis and be able to distinguish eczema from psoriasis
Identify risk factors for development of NMSCs
Be able to formulate a surgical treatment plan for melanoma
The skin serves as an important interface to the external environment as both a physical and an
immunologic barrier. Over the course of a lifetime, the skin undergoes characteristic changes as a result of
aging and the constant damage and remodeling that takes place. Plastic surgeons will frequently
encounter patients with concerns regarding the skin because it is the most visible of all the organs. It also
plays an important role in the manifestation of disease processes through dermatologic symptoms. An
example of a well-established form of external insult is ultraviolet (UV) radiation. UV radiation has been
implicated as a strong risk factor for the development of both nonmelanoma skin cancers (NMSCs) and
melanoma skin cancer, but also contributes to other benign skin lesions that develop. Latitude plays an
important part in UV radiation exposure with lower latitudes having more direct and greater UV
radiation exposure. 1 Although an external stimulus is often required, phenotypes of an individual also
play a factor in the development of diseases affecting the skin. Inadequate DNA repair mechanism
(xeroderma pigmentosum), those lacking melanin (albinism and lower Fitzpatrick score [I-IV]), and
immune deficiency have all been implicated in increased risk for skin cancer. In the United States, a
significant portion of health care dollars is spent managing the effects of UV damage to the skin in the
form of skin rejuvenation therapies as well as skin cancer treatment. 2 Skin cancer has the highest
incidence worldwide of all cancers. 3 There is also significant expenditure in the prevention strategies in
the form of sunscreen and other agents for skin resurfacing to reverse the effects of sun damage. This
chapter will review the effects of UV damage on the skin as well as go over some benign skin conditions
that may present to the clinician for evaluation. It will also cover the most common skin cancers and the
treatment strategies around those cancers.
mebooksfree.net

Skin
Anatomy
The skin is composed of two distinct layers—the epidermis, which is the most superficial layer, and the
dermis, which is the deeper layer. Below this is a layer of subcutaneous fat. Glabrous skin is found on the
palm and soles. It is without hair and has much thicker layer than the other skin of the body (Figure 20.1).
Hair-bearing skin is found throughout the body.
FIGURE 20.1 Glabrous versus nonglabrous (hair-bearing) skin layers.
Reproduced with permission from Bear, MF, Connors, BW, Paradiso, WA. The somatic sensory system. In: Bear, MF, Connors, BW, Paradiso, WA, eds.
Neuroscience: Exploring the Brain. 4th ed. Philadelphia, PA: Wolters Kluwer; 2016:415-482. Figure 12-1.
There are four layers of epidermis in the skin, and the glabrous has an additional fifth layer. The
epidermis is primarily composed of keratinocytes along with melanocytes, Langerhan cells, and Merkel
cells. The dermis is divided into two layers, with the more superficial papillary dermis and the deeper
reticular dermis. The dermis holds its integrity through a dense matrix composed of collagen and elastin
fibers that give both strength and elasticity. The dermis also houses the other skin appendages such as
hair follicles, sebaceous glands, and the apocrine and eccrine sweat glands.
With photoprotection, the skin will become thin, dry, smooth, and unblemished with aging. Aging
photodamaged skin is characterized by fine and course wrinkles, roughness, laxity, scaling, dryness,
telangiectasia, and dyschromia with an increase in the frequency of both benign and malignant lesions.
Solar elastosis describes the photodamaged extracellular matrix of the dermis with abnormal elastin and
decreased, disorganized collagen fibrils.
mebooksfree.net

Fitzpatrick Skin Type
Fitzpatrick skin type classifies the skin into six phenotypes based on the melanin content of the skin
(color), the inherent pigmentation of the skin, and the sensitivity of the skin to sun damage. 4 Lower
Fitzpatrick score indicates less melanin in the skin and hence a greater sensitivity to sun damage. This
leads to an increased risk for skin cancer in these individuals. The Fitzpatrick classification system has
also been applied to other modalities of skin resurfacing such as chemical peels and laser resurfacing as a
guide to how the skin will respond and the propensity for pigmentation damage and scarring after
resurfacing procedures. 4,5
mebooksfree.net

Sunscreen
There has been a clear and consistent link of sun exposure history (UV) and skin cancer. UV radiation
that reaches the earth’s surface is UVA and UVB. 95% is in the form of UVA; however, UVB is the most
carcinogenic through direct photochemical damage to cellular DNA and to the DNA repair mechanism of
the cell. Primary means of preventing damage from UV include ways that minimize the UV that can
penetrate the skin and cause photodamage, especially at a young age. Avoiding UV radiation exposure at
peak times (10 am–4 pm) and wearing protective outer clothing are examples of primary prevention.
Sunscreens protect the skin either chemically or physically. Chemical sunscreens absorb UV (usually
UVB) radiation with the most common sunscreen ingredient being para-aminobenzoic acid (PABA),
benzophenones, and cinnamates. Physical sunscreens work by reflecting UV radiation and producing a
protective barrier. They are usually opaque and include agents such as zinc oxide and titanium dioxide.
Sunscreens that block both UVA and UVB are best; however, the UV protection listed on the sunscreen is
for the UVB only, and many do not effectively block the UVB. Appropriate application ensures the
protective effect of the sunscreen.
mebooksfree.net

Retinoids
Retinoids (derivative of vitamin A) are well studied for the protective effects in the skin and reversal of
photoaging. They work by binding to nuclear receptors and modulating cellular gene expression.
Clinically, this leads to epidermal thickening and increased collagen in the dermis because of decreased
breakdown and increased collagen synthesis in the dermis. 6 They are available as topical (e.g., tretinoin
or Retin-A) and can be used to treat areas of actinic damage. Retinoids are also available systemically
(e.g., isotretinoin) and can be used to treat acne and rosacea; however, they are known teratogen and thus
must be used under close monitoring in women of childbearing years. Side effects such as pruritus,
burning, erythema, and peeling are usually seen with onset of treatment and can lead to noncompliance,
especially in patients with a history of sensitive skin. Photosensitivity is another common side effect of
topical therapies, and patients are advised to avoid sun exposure when starting therapy.
mebooksfree.net

Chemical Peels
Chemical peel strength is typically classified based on the depth of effect with deeper peels having
longer-lasting effects: superficial, medium, or deep. Superficial peels cause exfoliation of the epidermis
without deeper penetration. Medium peels affect the epidermis and penetrate into the papillary dermis,
and deep peels penetrate into the reticular dermis where they cause realignment of collagen. Peels are
used for a wide variety of skin conditions ranging from management of photodamage and melasma to
hyperpigmentation, rosacea, and acne. Deeper peels tend to have more profound and longer-lasting effect
compared with superficial peels. Common superficial peels include glycolic acid and salicylic acid,
whereas TCA (20%–35%) and Jessner’s are considered medium peels. Phenol and TCA (45%–50%) are
examples of deep peels. Peels are commonly used in conjunction with other skin resurfacing modalities
for improved outcomes. 7
mebooksfree.net

Laser Resurfacing
Laser resurfacing relies on the principles of selective thermal destruction (photothermolysis) to disrupt
damaged areas of skin and to induce remodeling. Ablative laser resurfacing causes wounding of the skin
to the dermis and then allows wound healing from dermal regenerative cells. Examples include the CO2
laser and erbium:yttrium-aluminum-garnet laser. Fractional lasers create microthermal zones of injury
that are spatially confined rather than confluent. This preserves some epidermis and shortens the
recovery time and diminishes the side effects. Healing occurs from the surrounding unaffected follicular
units. 8
mebooksfree.net

Evaluation and Management of Common Skin Disorders
Skin and Soft-Tissue Infections
Bacterial Infections
As a barrier to the external environment, the skin is in constant balance with the external bacteria. When
this balance is shifted, as in an open wound, bacteria proliferate, which can manifest as cellulitis and/or
an abscess. Cellulitis is caused by a break in the skin barrier that allows bacteria access to the deep
dermis. It manifests as spreading erythema and warmth. 85% of the time, no bacteria are cultured from
cellulitis, and treatment should cover the staph and strep of the skin flora with consideration of
methicillin-resistant Staphylococcus aureus (MRSA) in certain at-risk groups. 9 An abscess is a collection of
pus in the dermis/subcutaneous space. In the immunocompetent patient, the most common cause of
abscess is S. aureus but can be multifloral. In the immunocompromised patient, other more atypical
bacteria and fungi need to be considered.
mebooksfree.net

Bites and Stings
Bites and stings from mammals, spiders, reptiles, ticks, and insects are a common source of bacterial
infection. By causing a break in the normal protective skin barrier and inoculation of bacteria into the
deeper layers, the skin flora invades into the dermis. When treating bites from mammals and reptiles, the
oral flora of the animal should be considered for targeting therapy. Tetanus is also a risk, and appropriate
treatment is warranted depending on immunization history. Bites can also deliver toxins that can
produce both local and systemic effects. The brown recluse spider, for example, causes wide local tissue
destruction from the enzyme sphingomyelinase D. 10 Systemic toxicity from envenomation includes
neurotoxicity and coagulopathy and can be quickly fatal. Early treatment with respiratory management,
treatment of shock, and antivenom administration can be lifesaving.
mebooksfree.net

Infestations
Infestations including pediculosis (from lice [e.g., Pediculus humanus capitis, Pediculus humanus humanus,
and Phthirus pubis]), scabies, infection from bedbugs, schistosome cercarial dermatitis, and myiasis are
another source of skin lesion/irritation. Lice present with excoriated erythematous papules along with
pruritus and regional lymphadenopathy. Permethrin is first-line treatment for the condition. Scabies
(Sarcoptes scabiei var homonis) involves female mites living within epidermal burrows that create a
hypersensitivity reaction after 3 to 4 weeks (Figure 20.2). Usually the web space, wrists, and/or genitalia
are pruritic and erythematous. Topical permethrin 5% is the treatment of choice. Schistosomiasis is
manifested by pruritic erythematous plaques after the larvae penetrate the skin. Typically treatment is
symptomatic because most are animal schistosomes.
mebooksfree.net