Scrivere un articolo Statistica. Valter Torri Dip. Oncologia

Scrivere un articolo Statistica Valter Torri Dip. Oncologia

Qualità delle pubblicazioni Journal of Cerebral Blood Flow and Metabolism

Raccomandazioni sul reporting in base a tipo di studio clinico CONSORT REMARK STROBE QUORUM MOOSE STARD RTC Tumor Markers Studi Osservazionali Revisioni Sistematiche Diagnosi

CONSORT CHECKLIST 22 topics 5 sezioni Titolo Introduzione Metodi Risultati Discussione

CONSORT Diagramma di Flusso

Raccomandazioni sul reporting in base a tipo di studio preclinico

Errori e distorsioni Fase dello studio Possibili problemi Rimedi Priima: Diimensiione delllo studio Random error Diisegnii cross-over Diisegnii fattoriallii Arruollamento Durante: Conduzione Selection biias Performance biias Attriitiion bias Randomiizzazione Mascheramento Valutazione degllii endpoiint Detection biias Mascheramento Dopo: Analliisi statiistiica Selection biias Intentiion to treat The experimental design depends on the objectives of the study. In principle, a well-designed experiment avoids bias and is sufficiently powerful to be able to detect effects likely to be of biological importance.

Methods section: Il disegno sperimentale Experimental Unit Randomization Completely randomized designs Randomized complete block designs Blinding

Experimental units Each experiment involves a number of experimental units, which can be assigned at random to a treatment. The experimental unit should also be the unit of statistical analysis. Examples: If the treatment is given in the diet and all animals in the same cage therefore have the same diet, the cage of animals (not the individual animals within the cage) is the experimental unit Crossover experimental design may involve assigning an animal to treatments X, Y, and Z sequentially in random order, in which case the experimental unit is the animal for a period of time.

Methods section: Il disegno sperimentale Choice of Dependent Variable(s) Choice of Independent Variables or Treatments Uncontrolled (Random) Variables Replications (Factorial Experiments)

Factorial Design Factorial experiments have more than one type of treatment or independent variable (e.g., a drug treatment and the sex of the animals). The aim could be to learn whether there is a response to a drug and whether it is the same in both sexes (i.e., whether the factors interact with or potentiate each other). These designs are often extremely powerful in that they usually provide more information for a given size of experiment than most single factor designs at the cost of increased complexity in the statistical analysis. 14

Traditional design Purpose: Effect of drug on enzyme Design Controls 8 males Drug 8 males Stat: t-test / 14 degrees of freedom But, what about females? 15

Factorial Design Purpose: Effect of drug on enzyme Design B Controls 4 males 4 females Drug 4 males 4 females Stat: ANOVA Source DF MS Drug 1 xxx Sex 1 xxx Drug x Sex 1 xxx Error 12 xxx Same size, more info 16

Methods section: La dimensione campionaria Significance level Power Alternative hypothesis

Sample size & Power 6 per group: 12 per group:

Sample size & Power Desired power sample size Stringency of statistical test sample size Measurement variability sample size Treatment effect sample size

Dimensioni campionarie per dati continui e categorici Michael F. W. Festing and Douglas G.Altman ILAR Journal 2002 43(4): 245-258

Methods section: Analisi statistica Types of data Nesting (several observations per unit) Transformations Multiple Comparisons Serial Measurements Parametric and Nonparametric tests Correlation & regression

Parametric vs. non parametric tests Parametric tests are usually more versatile and powerful and so are preferred; however, they depend on the assumptions that: 1. the residuals have a normal distribution, 2. the variances are approximately the same in each group, 3. the observations are independent of each other

Transformations If the variances are not the same in each group and/or the residuals do not have a normal distribution, a scale transformation may normalize the data. A logarithmic transformation may be appropriate for data such as the concentration of a substance, which is often skewed with a long tail to the right. A logit transformation {log e (p/(1-p))} where p is the proportion, will often correct percentages or proportions in which there are many observations less than 0.2 or greater than 0.8 (assuming the proportions cannot be < 0 or > 1) A square root transformation may be used on data with a Poisson distribution involving counts when the mean is less than about five.

Results section: presentare i risultati Presentation of the Results Michael F. W. Festing and Douglas G.Altman ILAR Journal 2002 43(4): 245-258

Analisi di sopravvivenza rappresentazione secondo il modello di KM descrizione dei soggetti a rischio a tempi definiti. test statistico non parametrico (log-rank) descrizione dell HR e sui limiti di confidenza

Analisi della sopravvivenza: esempi Presentation of the Results: preclinical Drachman D. et al. Ann Neurol 2002;52:771 (25 ±1.5) (8±1.5)

Analisi della sopravvivenza: esempi Presentation of the Results: clinical Di Costanzo et al. JNCI 2008, 100:388