Alessandro Cucchetti

• MD
• Professor (Assistant) at University of Bologna

Dear reader.

We would like to try to translate a method commonly used in ecology in a specific field of medicine. We believe that MARK software can be sutiable and we would like to use it for repeated observations during a screening program for cancer detection.

Is there anyone out there who has a good experience with this model and has the mind open enough to be able to perceive its applicability outside of ecology?

Best wishes

Complications related to viral hepatitis, alcohol-related and non-alcoholic liver disease, are the main reason for seeking gastroenterologists and hepatologists advice. In addition, hepatocellular carcinoma often arise on the ground of hepatitis, representing the fifth most common cancer in men and the ninth in women. In 2015, the World Health Organization estimated that 325 million people were living with chronic hepatitis infections (hepatitis B or C) worldwide and that globally, 1.34 million people died of viral in 2015.

In front of this global health problem, gastroenterologists, hepatologists and hepato-biliary-pancreatic (HBP) surgeons, are daily involved in the clinical routine in taking difficult clinical decisions. As Sir William Osler quoted: “medicine is a science of uncertainty and an art of probability” and no doctor returns home from a busy day at the hospital without the nagging feeling that some of his/her diagnoses may turn out to be wrong, or some treatments may not lead to the expected cure. Probability is a recurring theme in medical practice and the ability of dealing with risk and uncertainty can be elicited through a special kind of intelligence. In 2012, The UK psychologist Dylan Evans defined it as “risk-intelligence” that is "a special kind of intelligence for thinking about risk and uncertainty", at the core of which is the ability to estimate probabilities accurately.  

Consequently, doctors are routinely asked to make predictions, and their predictions would lead to a consistent payoff when regarding a patient’s life. At the basis of “wise” medical decisions, physician’s experience surely plays a vital role. However, doctors can assume that their competency in a given area can be significantly higher than it really is. Such illusory superiority, is described as the Dunning – Kruger effect, a meta-cognitive bias leading to a discrepancy between the way people actually perform and the way they perceive their own performance level. The concept of “risk-intelligence” relies on the confidence that each subject has with their own knowledge, thus returning accurate probability estimates, and a “wise” doctor should be aware that he/she do not known, thus, returning high risk-intelligence.

To date, little is known about risk-intelligence and the Dunning – Kruger effect between doctors, and, especially, among hepatologists, a specialty strongly involved in important clinical decisions. With this aim we conducted a survey to test how risk-intelligence affects medical decision making in this particular clinical setting and whether the Dunning – Kruger bias can effectively affect these physicians.

If you are a gastroenterologist, hepatologist or HBP surgeon please help us in investigate this issue by completing the following survey:

I have r-project with meta and metafor packages.

I know how to perform meta-analysis of mean differences but I need to report also the meta-analyzed means (and sd) of the two comparison groups.

Doeas anyone know how to do this?

Regret theory postulates that the optimal (therapeutic) strategy is the action that results in the least amount of regret in case it is proven wrong. Faced with a cirrhotic patient with intermediate HCC the clinician must choose between the treatment suggested by the guidelines (trans-catheter arterial chemo-embolization - TACE) and surgery. The therapeutic strategy to adopt will depend on expectations from surgery (as well as from TACE). If expectations from surgery are too high, an eventual failure (patient death) will result in a great sense of regret. Regret can be used as a measure of expectations from surgery and present study showed how it can be applied in the decision-making of intermediate HCC.

The regret model assumes that a patient should be submitted to surgery if the probability of survival is equal or greater than a specific value (threshold probability - Pt) that can be derived by weighting the two components of regret: the result of a wrong action (regret of commission) and the failure to act (regret of omission). Assume that for a specific HCC patient the physician elicits a regret of omission of 70 and a regret of commission of 60; the threshold value can be calculated as 46.2% (Pt=1/(1+regret of omission/regret of commission)). If the patient survival, at a specific temporal end-point, is superior to 46.2% the physician should opt for surgery. Otherwise, TACE will be the least regretful strategy to adopt.

When three hypothetical clinical cases were send to 20 hepatologists and 20 surgeons asking for regret of omission and commission elicitation, results showed a significant separation among physicians' threshold values. However, the application of the regret-model has put in accordance more than 90% of clinicians on the optimal therapeutic strategy to adopt for each of the three clinical cases.

I am working on an accelerated failure time model in the setting of liver transplantation using survreg package for R. I have no problem in fitting the model but I can not obtain median survivals on the basis of covariates.

If I use only one variable (example recipient age: young=0, old=1) I have no problem using the formula: median <-predict(weibull.aft, newdata=list(recipient_age=c(0,1)), type=’quantile’,p=0.5), but how can I obtain the same results when fitting a model with more than one variable?

I would like to perform a trial where a reduction of 33% of the event (from 30% to 20%) is expected with the new treatment. I will perform 3 interim analyses (at 25%, 50% and 75% of enrollment) and I need to calculate futility boundary for stopping trial.

There are many functions in R-project but I am unsure of what to search for.

Can anyone help me?

I am trying to apply cure models in the setting of hepatocellular carcinoma. Reading pertinent literature, non-mixture models fit my needs well, because this model has a background in modelling of tumor recurrence.

Now the question is: what survival should be considered? Patient survival (event=death) or disease-free survival (event= death or tumor recurrence)?

From my point of view, cure should be defined for patients who are alive and without tumor, so disease-free survival has to be used. Is that right?

In the case of cancer, some patients will have fast-growing, aggressive tumors with short asymptomatic periods and rapid progression from symptoms to death. Other patients will have slower-growing, less aggressive tumors that are less likely to metastasize and, therefore, have a better prognosis. These less aggressive tumors have a longer asymptomatic period and are therefore more likely to be identified in a screening program.

When a cohort identified by screening (e.g., mammography) is compared with a cohort identified by clinical presentation (e.g., palpable mass), less aggressive tumors will be overrepresented in the screening cohort, and more aggressive tumors will be overrepresented in the clinical presentation cohort. Even in the absence of therapy, the cohort identified by screening will have a better prognosis. A screening program may appear to improve survival when in fact it has only preferentially selected out the subgroup with the best prognosis.

Is there anyone who knows that some work has addressed this problem with a mathematical approach easy to understand?