Цитата:
Сообщение от Iluhin
Но как им удалось предсказать будущее?
|
Легко (шутка). Думаю, дело в следующем: они оговорили в дизайне продолжительность follow-up'а, на которую было расчитано исследование, но какую-то часть пациентов наблюдали значительно дольше (при этом, раз наблюдали, то и УЗИ делать им, естественно, приходилось). Часть пациентов из наблюдения на этом незапланированном продленном этапе из исследования выпала, но эти цензурированные случаи технически не являются помехой для оценки кумулятивной частоты события. Правда, оценивать такую оценку нужно с большой осторожностью.
Подробнее - здесь: [
Ссылки доступны только зарегистрированным пользователям ]
Цитата:
Most clinical trials have a minimum follow-up time, at which point the status of each patient is known. The survival rate at this point becomes the most accurate reflection of the survival rate of the group. Survivor function at the far right of a Kaplan-Meier survival curve should be interpreted сautiously, since there are fewer patients remaining in the study group and the survival estimates are not as accurate.
For example, in a study looking at the survival of patients receiving treatment for stage I lung cancer, the median follow-up was 40 months. 4 However, 1 of the 302 patients had a follow-up at 10 years, and so a survival rate of 92% (95% confidence interval (CI), 88% to 95%) at 10 years was presented based on this one patient. Ninety-two percent at 10 years appears to be a very good estimated survival rate. However, with such a small subset of patients at this time point, the Kaplan-Meier estimates can be misleading and should be interpreted with caution. Carter et al. 5 pointed out that if this remaining patient had an event the following month, the survival probability would dramatically drop to 0% using the Kaplan-Meier calculations, with the CI 0% to 0%. Thus the estimations of survival from Kaplan-Meier analyses are most accurate at the time point when most patients are still present in the study. In this example, the median follow-up of the study was 40 months, and so the quoted survival rate of 92% (95% CI, 88% to 95%) is a better representation of the 4- or 5-year survival rate than the 10-year rate.5 This type of error can be avoided if authors include of patients at risk (remaining subjects in the study) for each interval.
|