We’re working a project on cancer cachexia. There is no “test” (or even modern consensus defintion) for cancer cachexia (tissue wasting) but there are different methods available for assessing whether an individual is suffering from cancer cachexia (e.g. Fearon criteria: >5% weight loss over 6 months or BMI <20 & >2% weight loss over 6 months or sarcopenia).
We have calculated prevalence of cachexia for people in hospice care using several of these criteria. These different estimates range from ~75% to 90%. Given the lack of a consensus defintion we would like to combine our estimates. My thought here is a Bayesian multilevel model (similar to a meta-analysis). Cachexia prevalence would be predicted by the individual model estimates (and associated errors).
This seems reasonable to me but I’m seconded guessing myself and if anyone has any other suggestions etc that would be useful.
Unfortunately you’ll find no validation of the Fearon criterion. Validation would require demonstration of the irrelevance of how far above or below the threshold a patient’s value s fall. Much better science can be done using the raw data, i.e., a triplet of variables (log(height), log(initial weight), log(current weight)). You can combine those 3 numbers into a single number of you have a dataset with ultimate outcomes to predict from the 3 variables, or you have a current hard physiologic or patient function measure to train the 3 variables to.
Thanks. I appreciate there is no validation of the Fearon criteria & unfortunately there is no validation of any criteria (e.g. Evans, GLIM, modified Glasgow prognostic scale) for classifying someone as cachectic. It is an ongoing problem in the field because there is no agreed upon definition nor functional/physiological cutoffs. There is no “current hard physiologic or patient function measure to train the 3 variables to”… there is just (effectively) clinical opinion as to whether someone is or is not cachectic. But the condition definitely exists!
What we’re trying to do characterise our hospice pts in terms of prevalence of cachexia with the ultimate aim of rasing the profile of cachexia and improving end of life care. We have used the Fearon criteria and the 3 others mentioned above to estimate this prevalance and (problems with these estimates notwithstanding) I wondered about the best way to combine them. They are the “best we can do” at the moment.
Given a lack of e.g. sensitivity, specificaty etc I thought the “meta-analysis” type approach might be best.
You mention that the purpose of your project is to improve end of life hospice care. This is a really important goal. And since cancer cachexia is distressing to patients and families, it seems like a good topic to study. But which methods are most suited to this type of research ?
I guess the first question is how you would define “improvement” in end-of-life care in this situation. Is the goal of your project:
To try to find an intervention that reduces the incidence of end-of-life cachexia?; OR
To focus on ways to help families and patients to accept and understand it, with the goal of reducing psychological distress?
According to UptoDate, there have been a lot of different definitions of cancer cachexia syndrome over the years. Unfortunately, it doesn’t seem like there’s strong evidence to support the efficacy of any particular intervention. Before designing more studies, it seems important to ask why this is the case. Is it because we simply haven’t yet found and tested an intervention with intrinsic therapeutic efficacy? OR is it because it’s not realistic to expect experimental methods to be able to disentangle the intrinsic efficacy of any therapeutic intervention (medical or psychological) from statistical noise when “syndromes” are used as trial inclusion criteria ? The End of the “Syndrome” in Critical Care
I guess the first question is how you would define “improvement” in end-of-life care in this situation. Is the goal of your project:
To try to find an intervention that reduces the incidence of end-of-life cachexia?; OR
To focus on ways to help families and patients to accept and understand it, with the goal of reducing psychological distress?
Both of these are important. The first is probably less important at end-of-life because by then cachexia (and the underlying pathology) are untreatable. However identification of key pathways/processes could help us moderate the severity of end-of-life cachexia. That would be useful. Much work here is done in pre-clinical models and is not necessarily very applicable to humans (it seems).
At end of life cachexia is a concern for both people with cancer (or other chronic diseases) and their families/carers. Psychological issues around body image and quality of life are important. So your latter point is important at end-of-life. Obviously there is nothing that pre-clinical models can really contribute here!
Currently we just want an estimate of cachexia prevalence in our context. There is no estimate at the moment for end-of-life cachexia (partly due to definitions; see my prev reply to Frank’s point) and partly due to a bias for studying cachexia whilst (putatively) treatable disease is present. It’s hard to resource clinical or psychological or nutrition/exercise care when you don’t really have a sense of the scale of the problem although you know it’s there.
Best,
i
ps - thanks for the link to the other thread. Interesting.
Wouldn’t it be correct to say that the condition doesn’t actually exist but that it’s a continuum of disease. Just as a diagnosis of type II diabetes doesn’t mean that much, but the severity of diabetes does? This argument is from https://www.acpjournals.org/doi/10.7326/0003-4819-149-3-200808050-00010
Since there doesn’t seem to be much evidence that specific interventions beyond standard supportive care (loosely defined as nursing/nutritionist/OT/PT/social work/spiritual care) improve the severity of cachexia syndrome or reduce patient/family distress, I assume that the “resourcing” you’re referring to is resourcing for routine supportive care (?) Are you saying that funders underestimate the cost of providing routine supportive care for cachexia syndrome in hospices, such that clarifying syndrome prevalence could support calls for more funding for supportive care (?)
If I’ve interpreted your goal correctly, then, rather than trying to mash several disparate definitions of cachexia syndrome together to get a prevalence estimate, is there any reason why you couldn’t do an observational study in which you assess each patient using both the most conservative definition AND the most liberal definition of cachexia syndrome, and then compare the amount of supportive care required for patients who satisfy only one versus both definitions ? This exercise might give you a min/max estimate of the funding that’s needed to support these patients (?)
correct to say that the condition doesn’t actually exist but that it’s a continuum of disease.
A fair point. This was also captured in Ken Fearon’s 2011 consensus paper as pre-cachexia, cachexia & refractory cachexia (discretisation of continuous variables!). Pre-cachexia is a key point to intervene but the condition remains invisible at that point.
Creating labels as they did is misleading, counterproductive, and information-losing. It requires you to add more variables to the analysis to make up for this loss of information. Some researchers will go so far as to claim there are subtypes of the disorder, rediscovering the lost information caused by binning.