Depression, the all-too-often unacknowledged wallflower of cancer comorbidities, powerfully impacts outcomes. (One 2009 meta-analysis in the journal Cancer, for example, found all-cause mortality was a staggering 39% higher among depressed cancer patients (Cancer 2009;115:5349-61).
If depression were easy to spot and referral resources were readily available, it seems likely that modifying this potent risk factor for negative outcomes could significantly improve patients’ life spans, as well as the quality of their lives.
Unfortunately, though, abundant research raises serious questions about oncologists’ acumen at detecting a lurking threat to their patients, that doesn’t show up on a lab slip or physical examination. A 2001 study of 143 physicians found that their rate of detecting depression in 2,297 cancer patients had a sensitivity of 29% and specificity of 85% (Br. J. Cancer 2001;84:1011-15).
So what to do?
Oncologists need the equivalent of a flagged level on a laboratory slip to raise their suspicion that a patient may warrant a closer look to rule out depression, or, if it is confirmed, to prescribe appropriate medication and/or psychotherapeutic interventions.
An impressive recent study may shine light on a number of ways to improve detection of depression in a clinically realistic way, without adding undue time to clinical visits or burdening patients and oncologists with lengthy questionnaires (J. Affect. Disord 2012;140:149-60).
Dr. Alex J. Mitchell of Leicester Royal Infirmary led a multidisciplinary, multinational team of researchers exploring myriad depression screening tools to identify those that not only passed muster in terms of sensitivity and specificity, but would also pass the sniff test in real oncology clinics and practices because of the ease of their delivery and interpretation.
Among 19 screening tools evaluated in 63 studies, the researchers narrowed the field to include only those that had been validated by at least two independent studies, leaving 8 tools for comparison in their meta-analysis. Those tools had been subject to 56 diagnostic validity studies in more than 10,000 patients with all stages of cancer. Studies assessing depression in palliative care settings were less common, but 16 met the stringent criteria established by Dr. Mitchell and his team.
In nonpalliative settings, the weighted prevalence of depression was 17.6%, quite in line with the 15% point prevalence of depression in the first 2 years after a cancer diagnosis. For this population, two diagnostic tools were classified as offering level 2 evidence for case-finding in the meta-analysis: the 21-item Beck Depression Inventory II (BDI-II) and the use of a single-stem question, as simple as, "Are you feeling depressed?"
Not surprisingly, the more comprehensive BDI-II was more accurate and better at ruling out false positives. But it had lower acceptability, the team concluded.
For screening of all cancer patients, asking two stem questions elicited better results than just one, and this approach received a rating of Level 1b evidence with high acceptability. These questions can vary, but usually address mood in question one and interest in question two. For example, a common two-stem query might be:
– During the past 2 weeks (or since your last visit), have you felt sad, blue, or depressed?
– During the past 2 weeks (or since your last visit), have you lost interest in most things that you normally enjoy, such as hobbies, work, or other activities?
According to the study authors, "For every 100 people screened in advanced cancer, the two questions would accurately detect 18 cases, while missing only 1 and correctly reassuring 74, with 7 falsely identified [as having depression when they did not].
For every 100 people screened in nonpalliative settings, the BDI-II would accurately detect 17 cases, missing 2 and correctly reassure 70, with 11 falsely identified as cases."
As with any study’s conclusions, there were limitations to this one, among them that the two-stem question yields only a "modest" positive predictive value among cancer patients.
"These limitations," the authors concluded, "are not so great as to completely preclude clinical usefulness and it nevertheless is likely to outperform oncologists’ unassisted clinical ratings."
They added that it is unlikely that a one-size-fits-all tool will ever be found to address every oncologist’s needs in every oncology setting for patients of all ages with all stages of disease. Perhaps a quick screen (like the two-question approach), followed by a more elaborate diagnostic tool for those who answer "yes," would be best to ensure that cases have not been inappropriately included.
Moreover, no screen is worth its salt unless treatment is available that can change the course of the illness and its impact on outcomes.