Experimental therapies:
▪ Phase I studies determine the relationship between toxicity and dose schedules of treatment
▪ Phase II studies identify tumour types for which the treatment appears promising
▪ Phase III studies assess the efficacy of treatment compared to standard treatment, including toxicity
Showing posts with label statistics. Show all posts
Showing posts with label statistics. Show all posts
95% confidence intervals (95% CI)
The 95% confidence intervals (95% CI) around a value are the range within which there is a 95% chance that the true value lies. Similarly, the 95% CIs around a difference are the range in which there is a 95% chance that the true difference lies.
If the means of two groups have overlapping 95% CIs, then the two groups are not statistically significantly different. If the 95% CI of the difference between two groups overlaps zero, then the difference between the two groups in not statistically significant.
Statistical and clinical significance should not be confused. A very large study can generate very narrow 95% CIs (or very small p values) for very small differences, which may be of no clinical significance at all. By contrast, a small study may fail to show a statistically significant effect even if the effect is both large and clinically important.
If the means of two groups have overlapping 95% CIs, then the two groups are not statistically significantly different. If the 95% CI of the difference between two groups overlaps zero, then the difference between the two groups in not statistically significant.
Statistical and clinical significance should not be confused. A very large study can generate very narrow 95% CIs (or very small p values) for very small differences, which may be of no clinical significance at all. By contrast, a small study may fail to show a statistically significant effect even if the effect is both large and clinically important.
Bias in study design
Bias means a flaw in study design that leads to a built-in likelihood that the wrong result may be obtained. It cannot be controlled for at the analysis stage.
It can be extremely difficult to design studies without potential bias, particularly when there are complex interactions between exposures under study. Techniques such a
It can be extremely difficult to design studies without potential bias, particularly when there are complex interactions between exposures under study. Techniques such a
Medical Staristics: Sensitivity and specificity
Interpreting the sensitivity and specificity of a test depends on what you are using it for.
The poor specificity of this test means that it would be inappropriate to use it as reason for telling people they have HIV infection; in the at-risk population over 50% diagnosed positive will not have the disease. By contrast, the balance of risk is different in screening blood for HIV, where the risk of missing a positive case far outweighs the risk of discarding some blood units unnecessarily, and similarly in looking for bowel cancer.
The test would also potentially be appropriate in screening for head lice where the disease is not serious but you do not want to miss cases and the treatment is simple and safe.
Predictive value depends as much on prevalence of a condition as the sensitivity. The positive predictive value in young patients means most army recruits with a positive test would be false-positive, with more being false positive than true positive, but since very few are likely to have heart disease a preliminary screening test with a positive predictive value of 48% would be a reasonable test to use.
The poor specificity of this test means that it would be inappropriate to use it as reason for telling people they have HIV infection; in the at-risk population over 50% diagnosed positive will not have the disease. By contrast, the balance of risk is different in screening blood for HIV, where the risk of missing a positive case far outweighs the risk of discarding some blood units unnecessarily, and similarly in looking for bowel cancer.
The test would also potentially be appropriate in screening for head lice where the disease is not serious but you do not want to miss cases and the treatment is simple and safe.
Predictive value depends as much on prevalence of a condition as the sensitivity. The positive predictive value in young patients means most army recruits with a positive test would be false-positive, with more being false positive than true positive, but since very few are likely to have heart disease a preliminary screening test with a positive predictive value of 48% would be a reasonable test to use.
Type 1 Error in Medical Statistics
A type 1 error is formally defined as being where the null hypothesis (which is that there is no difference between the groups) was falsely rejected. In practice this means that the study claims to find a difference that does not really exist, i.e. the result is a statistical fluke.
The conventional cut-off for significance is P=0.05, or a 1-in-20 chance. Hence if 20 trials were conducted, you would expect to get one that was ‘positive’ by chance alone.
Source: Fig 1, Fig 2,
 |
| Fig 1: Type 1 and Type 2 error |
 |
| Fig 2: Type 1 and Type 2 error |
The conventional cut-off for significance is P=0.05, or a 1-in-20 chance. Hence if 20 trials were conducted, you would expect to get one that was ‘positive’ by chance alone.
Source: Fig 1, Fig 2,
Subscribe to:
Posts (Atom)