Not Estimable: Why the Median Time-to-Event Is Sometimes Out of Reach ?

 In clinical trials and survival analyses, statements like:

• “The median DoR was not estimable with a median duration of follow-up of 26.6 months,” or

• “The median PFS and OS were not estimable”

typically mean that insufficient events (e.g., disease progression, death, or end of response) have occurred at the time of data cutoff to accurately estimate the median using Kaplan-Meier survival analysis.

Here’s a breakdown of the reasoning:

---

🔍 What Does "Not Estimable" Mean?

In the context of Kaplan-Meier curves, median is the time at which 50% of patients have experienced the event (e.g., progression, death, end of response).

• If fewer than 50% of patients have had the event by the data cutoff, then the median cannot be computed because the curve hasn’t dropped to 50%.

• So the median is “not estimable” (NE) or not reached.

---

📌 Common Scenarios for NE (Not Estimable):

1. Durable responses or long survival:
   If a large proportion of patients are still alive or responding, the median DoR/PFS/OS can't yet be determined.

2. Short follow-up relative to survival or response duration:
   Even if the median follow-up is 26.6 months, the actual number of events might be very low, especially in:

   • slow-progressing tumors

   • early in the trial

   • highly effective treatments

3. Heavy censoring:
   When many patients are censored (e.g., lost to follow-up or still ongoing without events), the Kaplan-Meier curve plateaus, making it hard to estimate median values.

---

✅ Example Interpretation:

“The median DoR was not estimable with a median follow-up of 26.6 months.”

👉 This likely means:

• Most patients were still in response at data cutoff.

• Less than 50% had disease progression or relapse, so Kaplan-Meier curve didn’t drop to 50%.

• Hence, median DoR is not estimable at this timepoint.

---

📊 What Can Be Reported Instead?

If median is not estimable, trials often report:

• KM estimates at fixed timepoints (e.g., 12-month PFS rate)

• Range of duration or maximum observed values

• Hazard ratios (if comparing two groups)

• Time-to-event plots showing how many are still censored

🔬  Statistical Theory Behind “Not Estimable”

🧮 Definition of Median (Survival Analysis)

In Kaplan-Meier (KM) analysis, the median survival time is the point at which the survival probability drops below 50%. Formally:

$$\text{Median} = \inf \left\{ t : \hat{S}(t) \leq 0.5 \right\}$$

Where:

• $\hat{S}(t)$ is the estimated survival function from the KM estimator.

• $t$ is time from randomization or response onset.

📉 Kaplan-Meier Estimation

The KM estimator is a step function that declines only when events occur. Censored data (no event at cutoff or lost to follow-up) do not cause drops in the curve but reduce the number of patients at risk.

---

Why Is the Median “Not Estimable”?

🔄 Scenario A: Long-Tail Survival

Suppose a targeted therapy yields long-term response or disease control in many patients.

Consequence:
KM curve has not yet crossed 50% event rate, even after a long follow-up → Median survival is not estimable.

🕓 Scenario B: Short Follow-up for Slow Disease

Imagine a disease with slow progression (e.g., indolent lymphoma) and a follow-up of 12–24 months.

Consequence:
Insufficient events accumulate → Median is not reached.

🏃 Scenario C: Highly Censored Data

In early-phase trials, especially basket or single-arm trials, many patients may:

• Still be on treatment

• Drop out for reasons unrelated to progression

• Be followed for <6 months

Consequence:
Survival curve plateaus → hard to estimate median.

Heavy Censoring = Median NE

🔁 Example Case:

You have 10 patients:

Patient	PFS (months)	Event?
A	6.5	Yes
B	7.0	Yes
C	10.2	No (Censored)
D	11.0	No (Censored)
E	12.4	No (Censored)
F	14.0	No (Censored)
G	15.8	No (Censored)
H	16.9	No (Censored)
I	20.3	No (Censored)
J	21.5	No (Censored)

👉 Only 2/10 had events
👉 KM curve never drops to 50%
👉 Median PFS = Not Estimable

What Should Be Reported Instead?

When median is not estimable, regulators and publications often report:

Metric	Description
KM estimates at fixed timepoints	e.g., 12-mo PFS = 73%, 24-mo OS = 88%
Range of observed durations	e.g., DoR ranged from 8.2 to 36.5 months
Censoring rate	% of patients censored due to ongoing treatment or loss
Hazard ratio (HR)	If comparing to control, e.g., HR = 0.54 (95% CI 0.29–1.02)
Restricted Mean Survival Time (RMST)	Area under survival curve up to time τ

---

📚 Clinical Trial and Regulatory Context

🏛️ In FDA/EMA Submissions:

• "NE" does not mean invalid — it can be clinically meaningful.

• If responses are durable and censoring is low, regulators may view NE positively.

• Some trials use predefined KM landmarks (e.g., 12-mo PFS) instead of median.

💊 Real-World Examples:

1. Vemurafenib in BRAF V600–mutant cancers (basket trial):

“Median DoR was not reached at the time of analysis. 68% of responders were ongoing at 18 months.”

2. Atezolizumab in NSCLC (IMpower110):

“Median OS not estimable in PD-L1 high group; 24-month OS rate was 60%.”

---

🧠 Key Takeaways

• “Median not estimable” is common and expected when:

  • Few events have occurred.

  • Treatment is durable.

  • Disease is slow-progressing.

  • Study has short follow-up or heavy censoring.

• It reflects data immaturity, not data quality.

• Reporting alternative metrics (landmarks, RMST, HR) is standard practice.

• Interpretation depends on context: a positive signal in oncology trials, especially immunotherapy or targeted therapies.