Paperonium

← Back to the plain-language case

Abstract attributes the all-EGFR c-Met-high response rate (10/19, 52.6%) to the EGFR-mutant subgroup (n = 15)

J Clin Oncol (2022) · doi:10.1200/JCO.22.00739 · PMC9928626

Summary

The abstract reports that EGFR-mutation-positive (EGFR-M+) patients with c-Met-high tumors (H-score ≥ 225; n = 15) had an objective response rate (ORR) of 52.6%. No whole number of responders out of 15 gives 52.6% (7/15 = 46.7%, 8/15 = 53.3%). The figure is exactly 10/19, which the Results give as the ORR for all 19 c-Met-high patients, a group that includes three EGFR wild-type responders. The Results' own breakdown puts 7 responders among the 15 EGFR-M+ c-Met-high patients, an ORR of 46.7% (exact 95% CI 21.3 to 73.4). This is the abstract's only efficacy estimate for the subgroup the paper presents as of greatest clinical interest, and it overstates that estimate by 5.9 percentage points.

What the paper reports

Abstract, Results:

ORR for EGFR-M+ patients (n = 28) was 32.1%. Of EGFR-M+ patients, those who were c-Met high (n = 15) had an ORR of 52.6%.

Results, Efficacy, fourth paragraph (the paragraph describing Figure 2):

Among patients with c-Met–high scores (n = 19), one achieved a CR (1/19; 5.3%; EGFR-M+), nine achieved a PR (9/19; 47.4%; six EGFR-M+ and three EGFR-WT), five had SD (5/19; 26.3%; EGFR-M+), and four experienced progressive disease (4/19; 21.1%; three EGFR-M+ and one rare/unknown EGFR), indicating an ORR of 52.6% and DCR of 78.9%. No patients with lower c-Met scores achieved a CR; two of 16 had a PR (12.5%), 12 had SD (75%), and two had progressive disease (12.5%), indicating an ORR of 12.5% and DCR of 87.5%.

Results, Efficacy, second and third paragraphs:

For EGFR-M+ patients (n = 28), ORR was 32.1% (9/28; 95% CI, 15.9 to 52.4), with one CR (3.6%) and eight PR (28.6%).

ORR in EGFR-WT patients (n = 5) was 40% (2/5; two PR) and DCR was 80% (4/5; two PR and two SD). All three patients with rare/unknown EGFR mutations achieved SD.

Patients and Methods, Assessments, Efficacy:

ORR was defined as the proportion of patients with confirmed complete response (CR) or partial response (PR).

Results, Patient Characteristics:

Importantly, 53% were c-Met high (H-score ≥ 225).

Figure 2 legend:

HI indicates c-Met high (≥ 225; n = 19).

Context box, Relevance:

The subsets of greatest clinical interest are the patients with EGFR-M+ NSCLC who progressed on a prior EGFR TKI, and with high c-Met H-scores by immunohistochemistry.

Discussion, first paragraph:

This combination also showed promising efficacy, especially in c-Met–high EGFR-M+ patients, in agreement with each drug's mechanism of action.

What the paper's own numbers show

Method. I checked which response counts out of 15 could produce the abstract's 52.6%, and which denominators up to 40 could produce it at all. I then rebuilt the c-Met-high group from the Results breakdown by EGFR group and computed the ORR of its EGFR-M+ column, with the exact (Clopper-Pearson) interval the paper says it uses. As a check, I recomputed every ORR and disease control rate (DCR) in Table 3 and Table 4 from their printed counts, and tallied the per-patient labels in Figure 2 (EGFR group colour, c-Met ≥ 225 hatching, MET-amplification asterisk and best response, transcribed by eye). Every number below is printed by recompute.py.txt.

Quantity Location Reported Recomputed from the paper's own numbers
ORR, EGFR-M+ and c-Met high (n = 15) Abstract, Results 52.6% 7/15 = 46.7% (exact 95% CI 21.3 to 73.4); 1 CR + 6 PR from the Results breakdown
Possible ORRs with n = 15 (derived) — 7/15 = 46.7%, 8/15 = 53.3%; none rounds to 52.6% (52.6% × 15 = 7.89)
Fractions with denominator ≤ 40 that give 52.6% (derived) — 10/19 and 20/38 only
ORR, all c-Met high (n = 19) Results, Efficacy p4 52.6% (1 + 9)/19 = 52.6% (CI 28.9 to 75.6); matches
DCR, all c-Met high Results, Efficacy p4 78.9% 15/19 = 78.9%; matches
Size of the EGFR-M+ c-Met-high group Results, Efficacy p4 n = 15 (Abstract) 1 CR + 6 PR + 5 SD + 3 PD = 15; matches the abstract's n
DCR, EGFR-M+ and c-Met high (derived) not reported 12/15 = 80.0%
Responders in the c-Met-high group who are not EGFR-M+ Results, Efficacy p4 three EGFR-WT PR 10 − 7 = 3
Upper limit for the EGFR-M+ c-Met-high ORR Table 3 (9 EGFR-M+ responders in all) — 9/15 = 60.0%, if every EGFR-M+ responder were c-Met high
EGFR-M+ c-Met-high responders in Figure 2 Figure 2 labels — 1 CR + 6 PR = 7 among 14 plotted (7/14 = 50.0%)
ORR, c-Met lower (n = 16) Results, Efficacy p4 12.5% 2/16 = 12.5% (CI 1.6 to 38.3); matches
c-Met-high share of efficacy set Patient Characteristics 53% 19/36 = 52.8%; matches
ORR / DCR, EGFR-M+ (n = 28) Table 3; Results p2 32.1% [15.9 to 52.4] / 85.7% [67.3 to 96.0] 9/28 and 24/28; same values and CIs
ORR / DCR, EGFR-WT (n = 5) Table 3 40.0% [5.3 to 85.3] / 80.0% [28.4 to 99.5] 2/5 and 4/5; same values and CIs
ORR / DCR, rare/unknown (n = 3) Table 3 0 [0.0 to 70.8] / 100 [29.2 to 100] 0/3 and 3/3; same values and CIs
ORR / DCR, total (n = 36) Table 3; Results p1 30.6% [16.3 to 48.1] / 86.1% [70.5 to 95.3] 11/36 and 31/36; same values and CIs
Table 4 subgroup ORR / DCR Table 4; Results p2-p3 31/77, 33/93, 27/73, 39/100 4/13, 10/13; 5/15, 14/15; 4/15, 11/15; 5/13, 13/13; all match

All of the paper's other response rates reproduce from their printed counts, including every Clopper-Pearson interval in Table 3. The abstract's 52.6% is the only one that cannot be formed from its stated denominator. Its digits equal the all-EGFR c-Met-high rate (10/19) given in the Results, and the 15 is the size of the EGFR-M+ column of that same breakdown. The abstract thus pairs the n = 19 rate with the n = 15 label.

The figure for the EGFR-M+ c-Met-high subgroup does not depend on whether responses were confirmed. Table 3 (confirmed responses) and Figure 2 (best response) both show 9 EGFR-M+ responders in the whole cohort (1 CR, 8 PR). Seven of them carry the c-Met ≥ 225 hatching in Figure 2, the same 7 the Results paragraph lists.

Figure

The abstract's 52.6% equals the all-EGFR c-Met-high rate 10/19; the EGFR-M+ column of the same breakdown gives 7/15 = 46.7%

Figure 1. Where the abstract's number comes from. Top left: the abstract sentence, with the value and the subgroup size it states in orange. Top right: the Results breakdown of the 19 c-Met-high patients by best response and EGFR group (counts as printed in the text). Bottom: the two response rates that breakdown supports, in blue. Pooling all columns gives (1 + 9)/19 = 52.6%, the value the Results report for all c-Met-high patients and the value the abstract prints. Using only the EGFR-M+ column, the group of 15 the abstract names, gives (1 + 6)/15 = 46.7%. No count out of 15 can give 52.6%.

On a common ORR axis, the abstract value sits on the n = 19 estimate and between the two possible n = 15 values

Figure 2. Objective response rates on one axis, with exact 95% CIs. Blue dots and bars are recomputed from the paper's counts; the orange diamond is the value the abstract reports for EGFR-M+, c-Met-high patients. The grey ticks above the first row mark every rate that 15 patients can produce (k/15); 52.6% falls between 7/15 and 8/15. The thin orange line at 52.6% passes through the all-c-Met-high estimate (10/19), not the EGFR-M+ one. The correction moves the subgroup estimate down by 5.9 points. It remains well above the rate in patients with lower c-Met scores (2/16) and above the rate in all EGFR-M+ patients (9/28), with wide, overlapping intervals.

Why it matters

Explanations considered

Status

Suggested correction

In the abstract, replace

Of EGFR-M+ patients, those who were c-Met high (n = 15) had an ORR of 52.6%.

with

Of EGFR-M+ patients, those who were c-Met high (n = 15) had an ORR of 46.7% (7/15).

if the Results breakdown counts confirmed responses, or with the confirmed count out of 15 if it does not. Alternatively, keep 52.6% and describe it correctly: "Patients who were c-Met high (n = 19), regardless of EGFR status, had an ORR of 52.6%." The authors may also wish to say whether the c-Met-stratified paragraph of the Results reports best or confirmed responses, and to reconcile it with Table 3 (35 v 36 patients; 12 v 11 responders; EGFR-WT and rare/unknown counts). In the Discussion, they may wish to check the MET-amplified ORR of 62.5% against the six MET-amplified patients in Figure 2 (3/6 = 50.0%), and in the Results the "two had SD (50%)" of five patients (2/5 = 40%).