Transcript presentation here

Overdiagnosis as an extreme form of
length-time bias
Breast cancer screening as example
Juan Gérvas, MD, PhD, Equipo CESCA, Madrid, Spain
[email protected] @JuanGrvas



Overdiagnosis is an extreme form of length-time bias and
occurs when screening detects indolent tumors that meet
diagnostic criteria for cancer but would not otherwise have
come to clinical attention during the lifetime of the individual.
Overdiagnosis artificially inflates the case survival rate.
Overdiagnosed patients obtain no benefit while being
subjected to all of the harms of treatment.
Reference:
Failure of Researchers, Reviewers, Editors, and the Media to Understand Flaws in Cancer Screening
Studies.Berry DA. Cancer. 2014;(Article first published online: 12 JUN 2014):1–8. doi: 10.1002/cncr.28795.
A
B
S
C
A = Tumor onset
B = Beginning of the period in which breast cancer can be diagnosed with
screening mammography
C = Start of the clinical phase (signs-symptoms) of breast cancer
D = Death because breast cancer, o for any cause
BC = Latency (sojourn), period for early detection
s = Critical point of irreversibility
SB = Lead time bias.
BC = Length time bias when BC is longer than usual, or even close or equal to "n"
(being "n" the number of years of health expectancy at the time of diagnosis).
Overdiagnosis "stricto sensu" means that BC is equal to "n".
D
In some cases C could be before B (and the latency period becomes negative)
A
B
S
C
A = Tumor onset
B = Beginning of the period in which breast cancer can be diagnosed with
screening mammography
C = Start of the clinical phase (signs-symptoms) of breast cancer
D = Death because breast cancer, o for any cause
BC = Latency (sojourn), period for early detection
s = Critical point of irreversibility
SB = Lead time bias.
BC = Length time bias when BC is longer than usual, or even close or equal to "n"
(being "n" the number of years of health expectancy at the time of diagnosis).
Overdiagnosis "stricto sensu" means that BC is equal to "n".
D
DEFINITIONS
1. Lead-time bias. The increment in survival among screen-detected cases that
equals the difference in time of detection and when the cancer would have been
detected otherwise. Labeling individuals earlier as having cancer means they will
live longer after having been so labeled.
2. Length bias. The increment in survival among screen-detected cases
because of the over selection of slowly growing cancers. Cancers that are slowly
growing have a longer period between the time a tumor can be detected on a
mammogram and when it will become evident from symptoms.
3. Sojourn (latency) period. The time a tumor can be detected on a
mammogram until when it will become evident from signs-symptoms. The
distribution of sojourn time in breast cancer is unknown. Sojourn time varies,
depending on factors such as age and breast density. Typical estimates range
from 4 years to 7 years.
4. Critical point of irreversibility. Starting period of invasive cancer (metastasis
or development of a clon which will "kill" the female without treatment).
CONCLUSIONS (I)




Overdiagnosis is not an error of diagnosis (overdiagnosis is not
a false positive)
Overdiagnosis is an error of prognosis.
We do not know why overdiagnosis is so common and, worse, we
cannot separate tumors that are overdiagnosis (we cannot at the
time of diagnosis identify "overdiagnosis tumors“)
In the future, genetic intratumoral heterogeneity (the presence of
multiple clones, intratumoral genetic chaos) might help
References:
•
•
•
Bewildering genetic intratumor heterogeneity: Evolution through space and time. Available in:
http://cancerres.aacrjournals.org/content/72/19/4875.short
Cancer intra-heterogeneity. Can the degree of genetic chaos accurately predict of tumour recurrence? Probably. Available in:
http://scienceblog.cancerresearchuk.org/2014/10/09/lung-cancer-evolution-a-journey-through-space-and-time/
Overdiagnosis. Tracking the clonal origin of lethal prostate cancer. Available in: http://www.jci.org/articles/view/70354
CONCLUSIONS (II)


We don’t know why overdiagnosis is so common and, worse, we
cannot separate tumors that are overdiagnosis (we cannot at the
time of diagnosis identify "overdiagnosis tumors").
In the future, genetic intratumoral heterogeneity (the presence of
multiple clones, intratumoral genetic chaos) might help. At least
in mice, mechanical properties of tumors differs substantially
from normal cells and tissues so macroscopic stiffness of breast
tumors predicts metastasis.
References:
•
Macroscopic stiffness of breast tumors predicts metastasis. Available in: http://www.nature.com/srep/2014/140701/srep05512/pdf/srep05512.pdf