Table 1.

Comparison between measures to summarize trend: annual absolute change in survival (AAC_S) and the annual percent change in the conditional probability of cancer death (APC_D).

CharacteristicsAAC_S: Annual Absolute Change in SurvivalAPC_D: Annual Percent Change in the Conditional Probability of Death
Underlying measureS(i,y): Cumulative survival, i.e., the probability of surviving cancer after i years from diagnosis, for patients diagnosed in year yP(i,y): Conditional probability of dying of cancer in interval i given alive at the beginning of the interval, for patients diagnosed in year y
Formula (approximation)Average of the survival difference {S(j,y+1)-S(j,y)} for calendar years y in the joinpoint segment100*{P(i,y+1)-P(i,y)}/P(i,y) = 100*[exp(B)-1] where B is the coefficient in the joinpoint segment
UnitPercentage points (pp): difference of two percentagesPercent: relative change of percentages. Similar to annual percent change (APC) for rates
ExamplesMoving from 40% to 42% means a 2–percentage point increase in cumulative survivalMoving from 40% to 38% means a 5 percent decrease in the annual probability of cancer death
InterpretationAAC_S(5) = 2%: The 5-year cancer survival is increasing on average 2 percentage points for each subsequent year of diagnosis.APC_D = −5%: The annual probability of dying of cancer is decreasing by 5% for each subsequent year of diagnosis, similar to a 0.95 relative risk of dying of cancer in year Embedded Image compared with year y.
Does it vary by time since diagnosis?Yes.No.
AAC_S(1) ≠ AAC_S(5)APC_D is the same for 0–1 year, 1- years, … from diagnosis.
Motivation/SummaryMore clear prognosis interpretation versus more awkward mathematical derivation.Clear mathematical interpretation versus challenging prognosis interpretation