All-Cause Mortality Appears to Trump the Few Instances in Which Medical Screening Tests Work to Reduce Disease-Specific Deaths

17 January 2015

On a population-wide (sample) basis, if one is screened for common killers, one still won’t live longer?

Probably not, based on a statistical (meta-analysis) review of previously published randomized controlled trials:

We selected 19 diseases (39 tests) out of 50 diseases/disorders for which USPSTF [U.S. Preventive Services Task Force, best known for its mammography guidelines] provides screening evaluation.

Screening is recommended for 6 diseases (12 tests) out of the 19. We assessed 9 non-overlapping meta-analyses and 48 individual trials for these 19 diseases.

Among currently available screening tests for diseases where death is a common outcome, reductions in disease-specific mortality are uncommon and reductions in all-cause mortality are very rare or non-existent.

© 2015 Nazmus Saquib, Juliann Saquib, and John PA Ioannidis, Does screening for disease save lives in asymptomatic adults? Systematic review of meta-analyses and randomized trials, International Journal of Epidemiology, DOI: 10.1093/ije/dyu140 (advance access, 15 January 2015) (at Abstract) (extracts)

Table 2 in the paper is the critical one

It shows whether the analyzed controlled trials successfully demonstrated that screening prolonged life (as compared to the control group) both as to:

(a) not dying earlier from the specific illness tested for

(b) just dying from any cause.

In reading Table 2

Keep in mind that a 1.0 means the screening did not prolong life. Numbers lower than 1.0 mean that it probably did. But most of these results are so close to 1.0 that the difference between screening and not screening is negligible.

For those familiar with the scientific method and its statistical foundation, we are trying to demonstrate that chance (randomness) could not have generated the result. Hence the research team’s repeated references to the null hypothesis.

The key point

The take-away message is that screening-based “reductions in all-cause mortality are very rare or non-existent.”

And this stays true, even when screening for the 3 conditions that appear to benefit from early detection — “ultrasound for abdominal aortic aneurysm in men; mammography for breast cancer; fecal occult blood test and flexible sigmoidoscopy for colorectal cancer.”

In other words, one still croaks at the same time that the control groups did. Some other condition (apparently) has stepped in to take us down.

So, why all the hype about the benefits of screening?

Profit motive, clinicians’ wish to be helpful, patients’ fears — and the difficulty of doing carefully designed trials with large enough numbers of people to generate statistically meaningful results — combine to motivate us to do things that, as yet, have no demonstrated quantitative benefit.

Caveat

One of the points that study co-author John PA Ioannidis has made throughout his illustrious career is that acceptable proof depends on:

(a) well-designed studies (those that are alert to confounding variables)

and

(b) large enough patient numbers to detect infrequently occurring, but nevertheless meaningful, effects.

In the case of the above cited study, one could reasonably argue that results based on flawed, too small studies do not really get much stronger, when agglomerated together. The study authors are aware of this shortcoming:

To avoid uncertainty and a continuing conundrum in the world of screening for disease, we need to choose the appropriate study design and outcome, depending on the disease, to evaluate the effectiveness of screening tests.

We argue that for diseases where short- and medium-term mortality are a relatively common outcomes, RCT [randomized controlled trials] should be the default evaluation tool and disease-specific and all-cause mortality should be routinely considered as main outcomes.

Our overview suggests that even then, all-cause mortality may hardly ever be improved.

One can infer that the authors’ pessimism in believing that all-cause mortality cannot be improved is probably based on the idea that:

(i) the huge number of medical conditions that begin to operate toward the end of life

and

(ii) the equally large number of variations in the human genome and its interacting/influencing environmental factors

(iii) make it doubtful that we can statistically sort each in detectible ways.

The moral? — Medical screening of non-high-risk patients, as of 2014, appears not to reduce all-cause mortality

At best, according to the cited meta-analysis, only a few effective screening tests — “ultrasound for abdominal aortic aneurysm in men; mammography for breast cancer; fecal occult blood test and flexible sigmoidoscopy for colorectal cancer” — allow us to reduce the likelihood of taking that particular avenue out life.

But even these beneficial screenings do not prolong our lives. Something else apparently gets us (as a sample population), before we can take advantage of the hoped for extension of life.

Curious, huh?

Postscript — about conundrums in medicine

Some of my previous essays, regarding the difficulty of achieving certainty in the medical field:

here — statistical obstacles to knowing anything medical “for sure”

here — pharmaceutical industry invents bogus illnesses and cures

here — medical practice kills patients because research did not ask the right questions

here — systemic pattern of choosing meaningless research endpoints, combined with improperly designed risk analysis

here — failure of professional self-regulation in quality-delivery systems

here — conflicts of financial interest that distort the validity of clinical practice guidelines

here — structural funneling of knowledge through self-interest

here — resistance of the medical establishment and its patients (as of April 2013) to act upon evidence that the value of mammography screening in some age groups is outweighed by its harms

here — analysis of prostate cancer screening’s lack of benefit (as of April 2013) for many patients

In sum, we frequently do not know what we think we do.

PeteFree.com

All-Cause Mortality Appears to Trump the Few Instances in Which Medical Screening Tests Work to Reduce Disease-Specific Deaths