Tuesday, July 31, 2007
3 disjointed topics this Tuesday:
Two from the latest Journal of Clinical Oncology.
First is one looking at cancer survival trends. It is based on data from the SEER database and presents 5 and 10 year survival estimates for multiple types of cancer. The authors use what is described as a novel and accurate model, "modeled period analysis," to make the survival estimates. Needless to say there is complicated statistical modeling going on here which I don't understand; the authors' claim is that this method provides more accurate information about survival because it is more reflective of recent trends (in this case up to 2003) - if anyone can explain why MPA is good in human-readable language please leave a comment. Why I'm bringing this up though is because it provides the most recent data in long-term cancer mortality for US patients, and presents data stratified by type and stage of cancer, as well as age and sex, giving us a good prognostic 'gestalt' for our patients.
The general findings regarding trends in cancer survival is that for many cancers long term survival has incrementally improved since 1988 (up a few percent). Survival for lung and pancreatic cancers however remains dismal (16% & 7% respectively at 5 years).
The other article is one on survival prediction for patients receiving radiation therapy. This was a prospective German study of 216 people with incurable cancer receiving palliative radiotherapy. The researchers gathered various clinicians' (including their institution's tumor board as a whole) survival predictions and measured a bunch of other variables (performance status, certain symptoms, opioid use, etc.). They found that clinicians' survival estimates weren't great. They asked the clinicians to place people in 3 categories (<1>6 months to live) and they were accurate a little more than half the time. When inaccurate they tended to overestimate survival and were worst at predicting those who would die in a month. I personally think these categories are a little unfair and one wonders if their results would have been different if they were <3mo,>6 months. Nevertheless all of this is in keeping with previous findings (docs aren't great predictors, tend to overestimate survival). Also in keeping with previous findings was that certain objective findings were independently associated with worse survival in their multivariate analysis: brain mets, dyspnea, poor performance status, high LDH, high WBC count. More interesting is that they also found that the use of opioids was an independent predictor of mortality in multivariate analysis (& had about the same magnitude of risk as brain mets). I can't recall opioid use (or pain itself) panning out as an independent risk factor in other similar studies looking at prognosis in advanced cancer. Christian, as our resident prognosis go-to guy, any thoughts? This is all likely a statistical dead-end of course but assuming it's not and assuming opioids weren't the cause of death per se one wonders what opioid use is a marker of here - identification by a doc that they are 'end stage' and strong analgesic use is appropriate? Severe pain itself predictive of mortality?
BMJ has a piece by Chochinov on dignity conserving care for all patients. It is a general piece, directed at a broad range of clinicians, and (perhaps mimicking BMJ's ongoing "The A,B,C's of..." series) presents the A,B,C,D's (attitude, behaviors, compassion, dialogue) of dignity in routine patient care. I never know whether I should despair or rejoice when I see pieces like this. Rejoice because they are being published in widely read and prestigious journals. Despair because I am reminded that we (physicians) do in fact really need to be reminded of Chochinov's very basic assumptions and recommendations (basic but of course accurate/important): patients want/need to be cared for as whole people, we should talk to our patients about their lives and strive to understand how their illnesses affect them.
"Treating a patient's severe arthritis and not knowing their core identity as a musician; providing care to a woman with metastatic breast cancer and not knowing she is the sole carer for two young children; attempting to support a dying patient and not knowing he or she is devoutly religious—each of these scenarios is equivalent to attempting to operate in the dark."
To amplify his point (...without this we'd be operating in the dark...): there is nothing special or squishy or superfluous about this, it's just good medical care.
And finally, Pain Medicine presents a randomized, placebo controlled (sham therapy) trial of therapeutic touch for reducing pain and anxiety during stereotactic breast biopsies. They describe sham TT as:
"Sham TT employed the approach of Quinn—practitioners having no conscious intent to help and counting backward by "serial 7s" silently during "treatment"—was administered by providers neither trained nor experienced in actual TT delivery but who completed two 3-hour training sessions instructing them only on hand movements simulating TT. In this way, variables such as support from another person not involved with the procedure and patient distraction during SCB can be controlled for."
No differences were found between the real TT group and the sham group in any of the outcomes, suggesting any benefits from TT are from caregiver attention or other similar mechanism. There has been ongoing concern about publication bias in CAM studies (negative studies not being published) so I'm glad to see decent, controlled trials like this getting published. On the other hand, this isn't the first controlled trial out there showing TT to not work (or its fundamental assumptions to be flawed), and one could argue why bother doing more studies?