Week 4 Flashcards
What is the course of the disease?
Cause (smoking, etc) -> abnormality -> Diagnosis -> prognosis -> treatment -> prevention -> health services
Once diagnosed its clinical stage
Outcome stage is after treatment
What’s the difference between pre-clinical and clinical stage?
Pre-clinical = Biologic onset of disease and symptoms first appear
Clinical = symptoms first appear, disease diagnosed, treatment, outcome
How to identify the onset of the disease?
With technology but usually hard to do which is why patients develop symptoms and then go to hospital
What is Prognosis?
- Prediction of future course of treated disease.
- Includes underlying natural history of disease being studied.
- Includes factors extrinsic to natural history (hypertension in heart attack, presence of other disease) of disease being studied.
- Includes thinking about the spectrum of disease.
- Usually thought of as domain of cohort study (to study different outcomes at different stages of the disease)
Clinical focus is on what?
recognition of disease.
Signs and symptoms lead usual diagnosis of disease.
Retrospectively, we often recognize clinical premonitory signs and symptoms, and that disease is probably clinically present though not diagnosed.
T or F: Diseases is usually present before presence of signs/symptoms.
True
What is Detection threshold (DT)?
Earliest time at which the disease can be detected by screening using available means. (ex: blood work, etc)
Defines a detectable pre-clinical phase if prior to clinical undiagnosed.
What is Therapeutic threshold (TT)?
Last time at which medical intervention has an important effect (usually curative) on altering the natural history of the disease
Longer detectable pre-clinical phase and therapeutic threshold late in natural history define a condition as….
amenable to screening.
- DT changes with progression of diagnostic modalities.
- TT is often empirically determined (Assumed early treatment works)
Chat says:
1. Longer Detectable Pre-Clinical Phase (DPCP):
A disease with a longer period where it can be detected before symptoms appear is more suitable for screening.
This allows enough time to intervene before it progresses too far.
- Therapeutic Threshold (TT) Late in Natural History:
If effective treatment is still possible later in the disease process, it supports screening efforts.
This means that even if a disease is caught at a later stage, treatment can still be beneficial.
T or F: Ideally we want to detect disease as soon as possible?
True, the faster the TT is in the preclinical stage the better
T of F: Predicting factors can alter onset of disease
True
Usually thought of as domain of case-control research.
When we look at they’re environment and risk factors we can ‘‘predict’’ what disease they could develop
Risk factors for Depression: Female, Alcoholism, Stress,
Physical comorbidities, Being poor, genetics,
Risk factors for Dementia: Female Aged, Alcoholism, Stress, Physical comorbidities, Cognitive decline
Depression can be a prognostic risk factor for dementia
What is the difference between Prognostic vs. risk factor?
Is acute hypertension a disease?
Risk factor: Aged, Overweight, Smoking, Salty diet, Alcoholism, Stress
Prognostic factor: Aged, Overweight, Smoking, Salty diet, Alcoholism, Hypertension, Stress, Diabetes
The outcome will be a heart attack
Chat says:
The key difference between a prognostic factor and a risk factor lies in when they influence the disease process:
Risk Factor (Before Disease Development):
A characteristic or exposure that increases the likelihood of developing a disease.
Present before disease onset.
Example: Smoking is a risk factor for lung cancer.
Prognostic Factor (After Disease Diagnosis):
A characteristic that affects the outcome of a disease after it has already developed.
Helps predict disease progression, survival, or response to treatment.
Example: Tumor stage is a prognostic factor for cancer survival—patients with early-stage tumors have better outcomes than those with late-stage tumors.
Summary:
Risk factors help predict who will get a disease.
Prognostic factors help predict how the disease will progress once it is diagnosed.
What are the caracteristics of a Prognostic study?
After disease occurs:
Patient sample
- Clarify characteristics of cohort (severity, etc)
T0/Time zero or baseline
- Address the inception, or status of disease (incident patient - first time cancer patient)
Follow-up
- Make sure the time is long enough for the onset of event. (for ex: work with hospital data base for the témoins)
Outcomes of diseases
- Document all the possible events not just death or disease. (different caracteristics of the disease, chronic outcome, etc) (the # of deaths is your outcome)
How to Describe prognosis?
Summarize the course of disease by a single rate;
- Convey a limited information by a single rate;
- Communicate with survival curve, survival analysis.
How to Evaluation prognosis?
- Case-fatality
- 5-year survival
- Observed survival
- Median survival time
- Relative survival
What is case-fatality?
The number of people who die of a disease divided by the number of people who have the disease
commonly used for short-term, acute diseases
What is the difference between Case-fatality vs. mortality rate (which includes both persons with and without the disease interested)
Case-fatality: only prognostic cohort
Mortality: general population
Why is the person-time test not good?
Not all disease have same deterioration rate
The first 6 months are the most challenging. When they over come those 6 months, they are good to live for a couple years
Stage 4 cancer = people die in less than 6 months
Diabetes = people can still live for a long time
Thats why person-time study is not good to use to compare diseases?
What is the lead time biais?
It’s when 2 populations die at the same time from the disease but one got diagnosed before the other
Slide 18
What are things we ask during prognosis?
What is the fraction of a population which will survive past a certain time?
Of those that survive, at what rate will they die?
Can multiple causes of death or failure be taken into account?
How do particular circumstances or characteristics increase or decrease the odds of survival?
What is survival analysis?
A study of the occurrence and timing of events. Covariates are studied to determine their effect(s) on survival duration. Although applicable for both retrospective and prospective data, they are best for the latter
Which Two features of survival analysis are not found in conventional statistics?
- censoring (you get some info but not all of it)
- time-dependent covariates (time-varying explanatory variables)
What does survival analysis analyse or look at?
Analyze survival data derived from laboratory studies of animals, clinical and epidemiological studies of humans, and other applications;
Predict the probability of response, survival, or mean lifetime;
Compare the survival distributions of experimental animals or human patients;
Identify risk and/ or prognostic factors related to event, survival, and the development of a disease.
What are Key elements of survival analysis?
Survival time
- The time to the occurrence of a given event. (time of diagnosis to time of death)
Censor
- Some patients may be alive or event-free at the end of the study period, or loss follow-up. The exact survival times of these subjects are known. (you have half the image but not clear. Ex: Some people are still alive after the 5-year follow window. You don’t know what happens to them after that.
Ex: people leave your study half-way through, you have no info on if they’re alive or not after they leave)
Event
- This event can be the development of another disease, response to a treatment, relapse, or death.
What is Censoring?
Data are collected over a finite period of time in a time-to-event study (always a time window); consequently, the time to the event may not be observed for all individuals in the sample
Ex: Some people are still alive after the 5-year follow window. You don’t know what happens to them after that.
Ex: people leave your study half-way through, you have no info on if they’re alive or not after they leave
What are Two major reasons for censoring?
- Some individuals will never experience the event (ex: they don’t get exposed)
- Some individuals will not experience it within the observation time (only after the observation window)
What are 2 examples of Censoring data?
Loss to follow up
- people leave or refuse to continue to be followed so we don’t know what happens to them when they drop-out
Event-free at the end of study
- we don’t know what happens to them after the observation window of the study is over
The amount of censoring is a function of:
Rate at which the event occurs
Length of the data collection period
The amount of censoring can be limited by what?
by the type of design but it cannot be eradicated
- The cost of lengthening the observation window must be considered
What are the Types of Censoring?
Non-informative
Informative
Right censoring
Left censoring
Interval censoring
What is Non-informative censoring?
Censoring mechanism occurs independent of event occurrence and risk of event occurrence
E.g., censoring occurs because data collection ends
Assume that all people who remain in the study after censoring occurs are representative of anyone who would have remained in the study had censoring not occurred
What is informative censoring?
Censoring that occurs because individuals have experienced the event or are likely to experience it
E.g., loss to follow-up
What is right censoring?
(Censoring to the right) The event of interest occurred after a certain time t, but at least the time to the t is clear
- The censoring time is prespecified and the same for all individuals
- Each individual has a specific censoring time
- Each individual is censored at a randomly selected time
What is left censoring?
(Censoring to the left) The event of interest occurred prior to a certain time t, but the exact time of occurrence is unknown.
- An event time is unknown because the beginning of time is not observed
- Often arises because the researcher has not paid sufficient attention to identifying the beginning of time during the study design phase
- Usually need to eliminate left-censored observations, either by changing the study design or by excluding them from the analysis
Event: the age at which a child learns to accomplish certain tasks at learning centers.
– Left censoring occurs if children can already perform the tasks when they start attending the learning centers.
What is interval censoring?
The event of interest occurred prior to a certain time t, but the exact time of occurrence is unknown.
What is Beginning of Time (T0)?
Everyone occupies only one state at the start of the study
Examples of beginning times
- Birth
- Precipitating event – e.g., hospitalization
- Arbitrary start time
Unrelated to event occurrence
E.g., date of randomization in a clinical trial
What is event occurrence?
An individual’s transition from one state to another
What are types of events?
Physical – disease presence; death
Psychological – depressed
Social – married; divorced
What are other caracteristics of event occurence?
States must be mutually exclusive and exhaustive
Most times we are only interested in two states, but some studies involve three or more states
Some states can be occupied only once in a lifetime (e.g., death), while others can be occupied again and again (e.g., pregnancy; hip fracture)
What is the time unit?
Smallest possible unit of time available to monitor event
- No single metric for time is universally appropriate
- Goal is to measure time as precisely as possible
- Continuous time data versus discrete time data
The distribution of survival times is usually described or characterized by what three functions?
- survival function
- probability density function
- hazard function
What is survival function?
S(t)=number of people survived/total
S(t)=P (an individual survives longer than t)
= P (T>t)
From the definition of the cumulative distribution function F(t) of T ,
S(t)=1-P (an individual fails before t)
= 1- F(t)
Here S(t) is a non-increasing function of time t with the properties
1 for t=0
0 for t=∞
The probability of surviving at the time zero is 1 and that of surviving an infinite time is zero.
Cumulative survival rate
What is Probability density function?
f(t)=number of people died/(total*survival interval)
Survival time T has a probability density function defined as the limit of the probability that an individual fails in a short interval or simply a probability of failure in a small interval per unit time.
What is Hazard function?
h(t)=number of people died/[survival interval(people survived at beginning-(people died/2))]
H(t) = -logS(t)
The hazard function h(t) of survival time T gives the conditional failure rate.
T or F: The cumulative hazard function can be any value between zero and infinity.
True
What is Time-to-event data?
Known by a variety of names: lifetime data analysis, reliability analysis, time-to-event analysis, event history analysis, survival analysis
Time-to-event data focuses on what?
Focus on occurrence and timing of events
– Characterizing the distribution of time to an event for a given population, comparing this time among different groups or testing the relationship between covariates and time to the event
Time-to-Event analyzes what?
Target Event
- Whose occurrence is being studied
Beginning of Time
- An initial starting point when the study participant has not yet experienced the target event
Metric for Time
- A meaningful scale in which event occurrence is recorded
What are the 2 main methods to study survival analysis?
Descriptive method
Analytic method
What are the 3 subtypes of Analytic method?
Parametric method: i.e., exponential, Weibull
- Distribution of data is time-consuming or not economical or no theoretical distribution adequately fit
Semi-Parametric method: Cox proportional hazards (PH) model
- Assumes the hazard function for different individuals is proportional to a baseline hazard function
Nonparametric method: Kaplan Meier estimator
- Makes no assumptions about the underlying survival time distribution
Because of possible censoring, summary statistics such as the mean and standard deviation cannot be used and may not have the desired statistical properties
What Other methods can we use to describe the data?
One way is to estimate the underlying population distribution of time to the event
Once the distribution is estimated parametrically or non-parametrically, we can then use it to estimate other quantities, such as median survival time
- Life-table
- Kaplan-Meier
- Cox proportional hazards (PH) model
- Non-parametric estimation
What is life-table?
Distribution of survival times is divided into a finite number of intervals
For each interval, we compute the number and proportion of cases or objects that
- entered the respective interval ‘alive’,
- failed in the respective interval (e.g., number of cases that died),
- were lost to follow-up or censored in the respective interval.
How do we Calculate the survival function?
When there is no censoring (no incomplete info on individuals)
– Direct method
When there is censoring
- We do not know event time of individuals who are censored, therefore use an indirect method
S(tj) = S(tj-1)[1 – h(tj)]
S(tj) = [1 – h(tj)] [1 – h(tj-1)] [1 – h(tj-2)] … [1-h(t1)]
What is a Kaplan-Meier estimate?
A plot of the Kaplan–Meier estimate of the survivor function is a step function, in which the estimated survival probabilities are constant between adjacent death times and only decrease at each death.
To produce a plot of the survival curves for each group of interest
What are Assumptions when using life tables and Kaplan-Meier method?
No secular (temporal) change in the effectiveness of treatment or in survivorship over time
Follow-up of persons enrolled in the study
– Life table: people who lost to follow-up is the same as those who are followed-up
Uniform distribution of risk and withdrawal during each time interval
What is the Log-rank test?
The comparison of the survival curves of two groups should be based on a formal non-parametric statistical test
Test the difference of survival function between two groups
Why is the Cox Proportional Hazard models considerable flexibility and is widely used?
The log-rank test cannot be used to explore (and adjust for) the effects of several variables.
Adjustment for variables that are known to affect survival may improve the precision with which we can estimate the treatment effect.
Cox regression analysis will yield an equation for the hazard as a function of several explanatory variables.
Cox regression is considered a ‘semi- parametric’ procedure because the baseline hazard function, h0(t) does not have to be specified. Since the baseline hazard is not specified
What is the Median survival time?
the length of time that half (50%) of the study population survives
How do you calculate the relative survival?
Relative survival = observed survival in people with the disease/Expected survival if disease were absent
Slide 63
