biostats

phase 1 clinical trials

-20-100 healthy volunteers or people with the disease state

-tests safety and dosage

phase 2 clinical trials

-100-300 people with the disease state

-testing efficacy and ADRs

phase 3 clinical trials

-300-3000 people with the disease state

-testing efficacy compared to other tx or placebo and ADRs

phase 4 clinical trials

-post-marketing surveillance after approval

-testing long-term and real-world efficacy and safety

new drug application (NDA)

comes after completion of phase 3 (those were INDs)

supplemental new drug application (sNDA) or supplemental biologics license application (sBLA)

used if there are any changes to the approval labeling

clinical trial disclosures

-protect the integrity

-disclose any potential conflict of interest (COI): financial incentives

sampling bias

participants are not representative of the whole pop

selection bias

baseline differences and prevents generalizability (can be from lack of randomization or strict exclusion criteria)

performance bias

groups not treated equally (no blinding)

detection bias

outcomes are measured differently (people assessing outcomes should be blinded)

attrition bias

patients with less severe disease have a higher dropout rate which skews results

reporting bias

decision to report a finding is influenced by the result

publication bias

only research with positive results is published

continuous data

-logical order with values that continuously increase by the same amount

-includes interval and ratio data

interval data

-continuous data that has no meaningful zero (0 does not equal none)

-ex: temp

ratio data

-continuous data that has a meaningful zero (0=none)

-ex: HR

discrete (categorical) data

-has categories

-includes ordinal and nominal data

ordinal data

categorical data that has arbitrary categories (names) and includes yes/no

ordinal data

-categorical data that is ranked and has logical order

-ex: pain scales (scales do not increase by the same amount like they do in continuous data)

when is using the mean preferred?

continuous data that is normally distributed

when is using the median preferred?

ordinal data or continuous data that is skewed

when is using the mode preferred?

nominal data

standard deviation (SD)

-how spread out the data is and to what degree it is dispersed away from the mean

-highly dispersed = larger SD

guassian (normal) distribution

-large sample sets of continuous data from a guassian or "normal" bell-shaped distribution

-it is symmetrical, mean/median/mode are the same value and the center point, and 68% of values fall within 1 SD of the mean and 95% fall within 2 SDs

skewed distributions

-not symmetrical

-when the number of values (sample size) is small and/or there are outliers

outliers

-have a large impact on the mean with a small number of values (use median instead)

-distortion of central tendency is decreased by collecting more values

skew

-the direction of the tail

-when outliers are the high values it is skewed to the right (positive skew)

-when outliers are low values it is skewed to the left (negative skew)

variable

data point that can be measured (age, sex, BP)

independent variable

changed by the researcher to determine whether it has an effect on the dependent variable (the outcome)

null hypothesis (h0)

-there is no statistically significant difference between groups

-we want to disprove or reject this

alternative hypothesis (HA)

-there is a statistically significant difference between groups

-we want to prove or accept this

alpha (a)

-alpha level is the standard for significance

-alpha (a) = error margin

-commonly set at 5% (or 0.05)

p-value

-compared to the alpha after the alpha is determined

-if p<0.05 the null is rejected and result is statistically significant

confidence intervals (CI)

-provides the same info about significance as p-value plus the precision of the result

-CI = 1 - a

-if alpha is 0.05 --> 95% CI

-if alpha is 0.01 --> 99% CI

interpreting CI if it includes/crosses O

-not statistically significant

-ex: -0.26-0.89

interpreting CI if it does not include/cross 0

-statistically significant

-ex: 18-58

interpreting CI by comparing ratio data (RR, OR, HR)

-if it includes 1 --> not stat sig

-if it does not include 1 --> stat sig

CI and estimation (extent and variability in the data)

-CI indicates you are 95% confident that the true value of the ARR for the general population lies somewhere within the range listed

-ex: 95% CI 6-35

-narrow CI = high precision

-wide CI = poor precision

type I error

-false-positive

-null hypothesis was rejected in error

-with a and p value <0.05 --> stat sig with probability of a type I error being <5%

type II errors

-false-negatives

-null hypothesis is accepted when it should have been rejected

study power

probability that a test will reject the null correctly (the power to avoid a type II error)

relative risk (RR) or risk ratio

-ratio of risk in exposed group / risk in control group

-risk = # of subjects in group with event / total # subjects in group

-RR = risk in tx group / risk in control group

interpreting the RR

-RR = 1 --> no difference

-RR > 1 --> greater risk in tx group

-RR < 1 --> lower risk in tx group

relative risk reduction (RRR)

calculated after the RR and indicates how much the risk is reduced in tx group compared to control group

-RRR = (% risk in control - % risk in tx) / % risk in control

-another way --> 1 - RR (decimal form)

RRR interpretation

if RRR is 43% --> tx group is 43% less likely to have outcmes

absolute risk reduction (ARR)

-includes reduction in risk and incidence rate of the outcome

-ARR = % risk in control - % risk in tx

ARR interpretation

-if ARR = 12% then 12/100 patients benefit from tx

-can use the inverse to determine NNT or NNH

number needed to treat (NNT)

-number of patients who need to be treated for a certain period of time in order for one patient to benefit

-NNT = 1 / (risk in control - risk in tx)

-another way --> 1 / ARR (decimal)

number needed to harm (NNH)

-number of patients who need to be treated for a certain period of time in order for one patient to experience harm

-same formula as NNT

rounding rules for NNT and NNH

-NNT: anything greater than whole number rounds up (52.1 would be 53)

-NNH: anything greater than a whole number rounds down (41.9 would be 41)

odds ratio

-used in case-control studies to estimate the risk of unfavorable events associated with an intervention

-OR = AD/BC

hazards ratio (HR)

-used instead of risk in survival analysis (analysis of death or disease progression)

-the rate at which an unfavorable event occurs within a short period of time

-HR = hazard rate in tx group / hazard rate in control group

OR and HR interpretation

-1 --> event rate is the same

->1 --> event rate in tx group is higher

-<1 --> event rate in tx group is lower

statistical tests for continuous data

-one-sample t-test: data from single sample group is compared to known data

-paired t-test: single sample group used for pre/post measurement (serves as their own control)

-student t-test: when the study has 2 independent samples (tx and control groups)

statistical tests for categorical data

chi-square test

correlation

-to determine if one variable changes or is related to another variable

-does not prove a causal relationship

regression

-used to describe the relationship between a dependent variable and 1 or more independent variables

-common in observational studies where multiple independent variables need to be assessed or need to control for many confounding factors

-3 types = linear (continuous), logistic (categorical), cox (categorical in survival analysis)

sensitivity

-the true positive

-how effectively a test identifies patients with the condition

-a test with 100% sensitivity will be positive in all patients with the condition

specificity

-the true negative

-how effectively a test identifies patients without the conditions

-a test with 100% specificity will be negative in all patients without the condition

equivalence trials

the tx has roughly the same effect as the old tx

non-inferiority trials

new tx is no worse than the current standard

forest plots

-used for meta-analysis

-provide CIs for difference data or ratio data

when interpreting statistical significance using a forest plot

-the boxes show the effect estimate

-diamonds at the bottom represent pooled results from multiple studies

-horizontal lines through the boxes show the length of the CI for that particular endpoint

-the vertical solid line is the line of no effect --> it is set at zero for difference data and at one for ratio data

comparing difference data

result is not stat sig if the CI crosses zero so the vertical line is set at zero

comparing ratio data

result is not stat sig if the CI crosses 1, so the vertical line is set at 1

case-control studies

retrospective comparisons of cases (with disease) and controls

cohort studies

retrospective or prospective comparisons of patients with an exposure to those without an exposure

randomized controlled trials

prospective comparison of patients randomly assigned to groups

ECHO model

-economic = direct, indirect and intangible costs of the drug compared to medical intervention

-clinical = medical events that occur as a result of the tx or intervention

-humanistic = consequences of the disease or tx as reported

direct medical costs

drug preparation and admin, inpatient direct costs, outpatient direct costs

direct non-medical costs

travel, household costs (childcare), home health aids

indirect costs

lost work time, low work productivity, morbidity, mortality

intangible costs

pain, suffering, anxiety, fatigue

incremental cost-effectiveness ratios

-change in costs and outcomes when 2 tx are compared

-ICR = (C2-C1) / (E2-E1)

-C = costs, E = effects

cost-minimization analysis (CMA)

-when 2 or more interventions have demonstrated equivalence in outcomes and the costs are being compared

-use is limited given its ability to compare only options with demonstrated equivalence outcomes

-outcome unit = demonstrated or assumed to be equivalent in comparative groups

cost-benefit analysis (CBA)

-comparing benefits and costs of an intervention in terms of monetary unit

-outcome unit = dollars

cost-effectiveness analysis (CEA)

-compare the clinical effects of 2 or more interventions to the respective costs

-advantage = outcomes are easier to quantify since they are similar to outcomes seen in clinical trials

-most common

-disadvantage = inability to directly compare different types of outcomes

-outcome unit = natural unit (life-years gained, BP)

cost-utility analysis (CUA)

-form of CEA that includes things like quality-adjusted life years (QALYs) and disability-adjusted life years (DALYs)

-outcome unit = QALY or others