What Is a Kaplan–Meier Curve and How Do You Read It? A Guide to Interpreting Cancer Survival Data

One of the key factors determining how the success of a cancer treatment is assessed in clinical trials is the chosen endpoint. In early-phase clinical trials, indicators based on tumor volume, such as objective response rate (ORR) , take priority, whereas survival measures (progression-free survival and overall survival) have become more decisive in later-phase trials. In modern oncology, however, the effects of treatment on quality of life, tolerability, and long-term benefit are increasingly important alongside survival duration.

Therefore, particularly in the era of immunotherapies, antibody–drug conjugates (ADCs), and targeted therapies, additional endpoints such as Disease Control Rate (DCR), Clinical Benefit Rate (CBR), Health-Related Quality of Life (HRQoL), Time to Next Treatment (TTNT) and Duration of Response (DoR) offer a more refined assessment of clinical efficacy beyond conventional PFS and OS.

Looking back at the history of clinical oncology, treatment efficacy was initially measured solely by mortality rate . However, this approach was insufficient for detecting treatment differences early, particularly in slow-growing cancers or those requiring long follow-up. The “Kaplan–Meier method,” developed in 1958 by Edward L. Kaplan and Paul Meier , revolutionized oncology as a groundbreaking method for the statistical analysis of time-to-event data.

The Kaplan–Meier curve assesses each patient's survival time or time until an event individually and uses a concept known as “censoring.” This ensures that individuals who have not yet experienced the event (for example, a patient who is still alive or has not experienced progression) are not excluded from the analysis. Clinical trials can therefore generate more accurate survival estimates by evaluating all participants' data over time.

This statistical approach enabled the development of new endpoints addressing different clinical questions over time. In the chemotherapy era of the 1980s, Overall Survival (OS) was considered the “gold standard,” while, from the 1990s onward, targeted therapies brought intermediate endpoints such as Progression-Free Survival (PFS) and Time to Progression (TTP) to the forefront. Today, immunotherapies and biologic agents have enabled some patients to remain stable for long periods even without complete eradication of disease, leading to new measures such as Disease Control Rate (DCR), Clinical Benefit Rate (CBR) and Duration of Response (DoR) .

The concept of “efficacy” thus came to encompass not only survival duration, but also the duration of disease control, quality of life, and treatment tolerability. This evolution is a key milestone in the shift toward a patient-centered perspective in clinical research.

Today, advanced statistical models assess multiple endpoints together, rather than a single outcome, using methods such as joint survival modeling or multi-endpoint analysis to calculate the overall clinical impact. This enables a more comprehensive assessment of the true value a treatment offers patients.

Clinical Endpoints and Clinical Examples

• 1. Disease Control Rate (DCR):
  The proportion of patients who achieve a complete response, partial response, or stable disease with a treatment. For example, in the KEYNOTE-158 trial, a DCR of 33% was reported in microsatellite instability-high (MSI-H) solid tumors treated with pembrolizumab.
• 2. Clinical Benefit Rate (CBR):
  The proportion of patients who achieve a complete or partial response or maintain stable disease for at least 24 weeks. This measure is used particularly to describe the efficacy of endocrine therapies in hormone receptor-positive metastatic breast cancer.
• 3. Health-Related Quality of Life (HRQoL):
  Measured using questionnaires such as the EORTC QLQ-C30 or FACT-G. For example, in the CheckMate-214 trial, nivolumab + ipilimumab significantly improved HRQoL compared with sunitinib.
• 4. Landmark Overall Survival — Survival at a Specific Time Point:
  The proportion of patients alive at month 12 or 24. It is used as a rapid indicator of efficacy, particularly in early phase II trials.
• 5. Time to Treatment Failure (TTF):
  The time until treatment is discontinued for reasons such as progression, toxicity, patient preference, or death. It is used to measure treatment tolerability in real-world data.
• 6. Time to Next Treatment (TTNT):
  The time between one treatment and the next. For example, in indolent lymphomas or metastatic colorectal cancer, it is important for demonstrating the clinical durability of maintenance therapies.
• 7. Duration of Clinical Benefit (DoCB):
  The period during which disease remains stable or a response is maintained over the long term. It reflects long-term benefit in patients showing a “plateau effect” after immunotherapy.
• 8. Duration of Response (DoR):
  The time from achieving a complete or partial response until progression. For example, in the DESTINY-Breast03 trial, a median DoR of 25.1 months was reported in the trastuzumab deruxtecan arm, providing a substantial advantage over chemotherapy.

Survival analysis is a fundamental statistical tool in modern clinical research, and the Kaplan–Meier (K–M) methodhas been standard in clinical oncology since Edward Kaplan and Paul Meier described it in 1958. This method calculates, step by step, the probability of a specified event (such as progression, recurrence, or death) occurring over time in a patient group, producing curves that visually represent the “probability of survival.”

Unlike conventional statistical analyses, the K–M curve accounts for all patients' data. If a patient leaves follow-up before the event occurs (for example, because of loss to follow-up, death from another cause, or early withdrawal from the trial), that patient is described as “censored.” Censoring does not invalidate the study's survival analysis because the Kaplan–Meier method retains these individuals' contributions up to that point. In this respect, it is the statistical model closest to real-world conditions.

1. Structure and Interpretation of the Curve

The Kaplan–Meier curve usually follows a staircase pattern. Each step represents the time when a patient experiences the event. As time progresses along the axis, the curve gradually falls, indicating an increasing probability of the event occurring. The point where the curve reaches 50% is defined as the median survival time . This value represents “the time during which half the population remains alive without experiencing the event.”

2. Comparing Two Curves: The Log-Rank Test and Hazard Ratio

When comparing survival curves for two treatment arms in clinical trials, the most frequently used method is the log-rank test. This test analyzes whether event rates differ significantly between the groups over time. If the resulting p-value is below 0.05, the difference between the groups is considered statistically significant.

Survival analyses also use the concept of the hazard ratio (HR) . The HR expresses the risk of an event (such as death or progression) in the treatment arm relative to the control arm:

• HR = 1 → No difference between the groups.
• HR < 1 → Lower event risk in the treatment arm (e.g., HR=0.70 → 30% risk reduction).
• HR > 1 → Higher event risk in the treatment arm.

3. The Clinical Value of Kaplan–Meier Curves

Kaplan–Meier curves are not merely statistical displays; they are also clinical tools for understanding the dynamics of treatment efficacy over time. The “plateau effect” seen particularly with immunotherapies—a high survival rate that remains stable over a long period—indicates long-term disease control in some patients. This is rarely seen with conventional chemotherapy and points to a subgroup of patients with prolonged responses to immunotherapy.

Kaplan–Meier curves are also frequently used in interim and subgroup analyses. For example, comparing separate curves for PD-L1-positive and -negative subgroups can reveal the molecular profile in which a treatment is more effective.

4. Visual Representation

Clinical endpoints are central not only to scientific research but also to drug approval processes. If an endpoint can demonstrate that treatment provides “real clinical benefit” to patients, the data carry the highest level of evidentiary value for both regulatory authorities (FDA, EMA) and health economics.

1. The FDA Perspective: The Concept of a “Surrogate Endpoint”

The US Food and Drug Administration (FDA) has, since 1992, formally defined the concept of a surrogate endpoint for accelerated approval of oncology drugs. Under this concept, even an endpoint that does not directly measure survival (such as tumor shrinkage, PFS, or DoR) can “reasonably predict” clinical benefit if it is highly correlated with overall survival (OS).

This principle enables the accelerated approval mechanism. For example:

• Pembrolizumab (Keytruda®) — received accelerated approval for MSI-H or TMB-high solid tumors based on objective response rate (ORR) and DoR data.
• Trastuzumab deruxtecan — received early approval following the DESTINY-Breast01 trial because of high DoR and DCR rates, with subsequent OS data converting this to full approval.
• Lenvatinib + Pembrolizumab combination — entered the approval process following assessment based on PFS and CBR in endometrial cancer.

The FDA describes such approvals in terms of “intermediate clinical benefit” and requires mature “hard endpoint” results, such as OS or quality of life, for full approval.

2. The EMA Approach: “Conditional Marketing Authorization”

The European Medicines Agency (EMA) has similarly operated, since 2006, a conditional marketing authorization system. In small patient populations or situations of urgent unmet need, the EMA may consider early endpoint data such as PFS, DoR, or HRQoL sufficient to authorize marketing. However, the manufacturer must provide final OS or long-term safety data within a specified period.

For example:

• Nivolumab — received conditional EMA approval in 2015 for metastatic renal cell carcinoma based on PFS and HRQoL data, converted to full approval in 2017 with OS data.
• Sacituzumab govitecan — received accelerated EMA approval in TNBC because of a high DoR rate in 2021.

3. Current Trends: Composite Endpoints and Real-World Data

Regulatory authorities now favor multidimensional measures of impact over a single endpoint. Composite endpoints combine PFS, HRQoL, and toxicity data in their assessment. Parameters such as Real-World Evidence (RWE) and Time to Next Treatment (TTNT) also provide valuable information on real-world effectiveness beyond randomized clinical trials.

For example, in the KEYNOTE-091 trial, PFS superiority was demonstrated in patients receiving adjuvant pembrolizumab. Although OS data were not yet mature, long-term HRQoL stability was included in the decision-making process.

4. Impact on Clinical Practice

These developments have led oncologists to ask not only “how many additional months of life does it provide?” but also “what is that life like?” when choosing treatment. Particularly in palliative and advanced cancer settings, measures such as duration of response (DoR) or time between treatments (TTNT) have clinical value in optimizing the balance between a patient's quality of life and treatment burden.

These parameters are also critical indicators influencing reimbursement decisions in health policy. In Türkiye, the Social Security Institution (SGK) uses the clinical endpoint data underpinning international regulatory decisions as a reference, particularly for immunotherapy and targeted agents.

Kaplan–Meier curves and clinical endpointsare not only numerical tools for understanding the efficacy of oncology treatments, but also a “language” that allows us to read the biological behavior of disease over time. Modern oncology seeks to answer not only “how many patients experienced tumor shrinkage?” but also “which patients maintained disease control, for how long, and what was their quality of life during that time?” Measures such as PFS, OS, DoR, CBR, and HRQoL have therefore become central to clinical decisions.

Assessing clinical data with Kaplan–Meier curves visualizes how durably a treatment works over time. Early separation of the curves reflects rapid drug efficacy, whereas a prolonged parallel course later on reflects the “plateau effect” seen with immunotherapies—a subgroup of patients with prolonged responses. This visual analysis is not merely a statistical indicator; it is also the clinical reflection of biological mechanisms.

Today, the FDA and EMAseek far more in their approval processes than a one-dimensional answer to “does it prolong life?” Multidimensional indicators of clinical benefit—“how long was disease controlled?”, “was quality of life preserved?”, and “is the response sustainable?”—have become the new language of regulation. This approach gives the patient-centered paradigm in cancer treatment a statistical foundation.

Meanwhile, the growth of real-world evidence (RWE) and the development of digital patient monitoring systems allow us to move beyond conventional clinical endpoints. Metrics derived from electronic health records, such as time-to-next-treatment (TTNT) or real-world progression-free survival (rwPFS) , are now assessed as complementary to clinical trial data.

This shift also matters ethically. Statistical significance alone is no longer considered sufficient; the concept of “clinical significance reflected in the patient's quality of life” is coming to the forefront. This approach enables clinical trials to be interpreted from a human perspective, beyond numbers alone.

References

• FDA Oncology Center of Excellence. Clinical Trial Endpoints for the Approval of Cancer Drugs and Biologics. 2024 Update.
• Broglio KR et al. “Understanding clinical trial endpoints and their implications for clinical practice.” J Clin Oncol. 2023;41(2):115–124.
• Schwartz LH et al. RECIST 1.1 — “Response Evaluation Criteria in Solid Tumors.” Eur J Cancer 2022;163:20–28.
• EMA Guidelines on the Evaluation of Anticancer Medicinal Products in Man. EMEA/CHMP/205/95 Rev.6, 2023.