What Gets You the Interview Is Not What Gets You Ranked

Sent to subscribers on 21 September 2026. Details may have changed since.

Training, research, and the specialty itself. Three things worth knowing this week.

Residency and Fellowship · Two different games, new NRMP data
Research · How to read an eight year update without being fooled
Cardiothoracic Surgery News · Four days in Seoul

Part One · Residency and Fellowship

Two Different Games

Programs open your file on Wednesday. Interview invitations follow within days. So this is the week to understand something the newest NRMP data makes unusually clear: the factors that get you into the room and the factors that get you ranked once you are in it are almost entirely different lists.

The NRMP published its 2026 Program Director Survey on August 26. It was fielded in March, in the eleven day window between rank list certification and Match Week, so outcomes could not color the answers. Program directors were asked which factors they used at each stage and how important each one was, on a scale of one to five.

To get the interview

89% Step 2 CK score · importance 4.0
87% Step 1 pass · importance 4.4
86% Specialty letters · importance 4.0

To get ranked

89% Interpersonal skills · importance 4.7
84% Interview interactions · importance 4.7
74% Feedback from residents · importance 4.5

Look at the right hand column. Not one of the three top ranking factors is on paper. All three are about how you are in the room, and the third is about how you are with people you may not think you are interviewing with at all. Three quarters of program directors say current residents' feedback shapes the rank list, and they weight it 4.5 out of 5. The resident who walks you to the conference room, the one who answers your questions over lunch, the one you sit next to at dinner: each of those is a scored interaction, whether or not anyone tells you so.

Paper gets you in. Once you are in, paper stops mattering, and the people you were least careful with may have the most to say about you.

Two more findings worth carrying into the season. First, being flagged for a Match violation was cited by only 20 percent of program directors for interview selection and 15 percent for ranking, yet it carried the highest mean importance of any factor in the entire survey, 4.7 and 4.8. Few programs encounter it. Those that do treat it as close to disqualifying. Second, for IMGs specifically, 84 percent of programs require Step 2 CK to have been taken for interview consideration, versus 76 percent for U.S. MDs, and among programs that require it, 32 percent set a target score for IMGs versus 26 percent for U.S. MDs. The bar is real and it is measurably higher.

One honest caveat. The response rate was 12.7 percent, 861 program directors. General Surgery contributed 59 responses. Integrated Thoracic Surgery contributed fewer than ten and is not broken out. Read the numbers as a well constructed signal, not a census.

What to do with this

Prepare for the interview day as two separate events. The formal interview is where your file is confirmed. Everything around it, the resident lunch, the tour, the pre interview dinner, the hallway, is where the rank list is actually written. Most applicants rehearse the first and improvise the second. Reverse that. Have two or three genuine questions ready for residents that you actually want answered, and treat their answers as data about the program, because that is what they are treating yours as.

Part Two · Research

How to Read an Eight Year Update Without Being Fooled

On Monday in Seoul, the ADAURA investigators presented eight year overall survival for adjuvant osimertinib in resected EGFR mutated stage IB to IIIA lung cancer, simultaneously published in the Journal of Thoracic Oncology. The headline was clean: 74 percent alive at eight years versus 58 percent with placebo in the primary stage II to IIIA population, hazard ratio 0.53. The drug was already standard of care. Nothing about that changes.

I want to use it for a different purpose. Long term updates of positive trials are among the most commonly misread documents in medicine, because the headline number arrives with an authority the underlying data does not always have. ADAURA is a good trial, and precisely because it is a good trial, its own report tells you exactly where to look. Here are the five questions I ask of any late follow up, worked on this one.

Five questions for any long term update

1. Was this analysis planned, or added later? Added later. The final planned OS analysis was January 2023. A protocol amendment in August 2023 asked surviving patients to consent to yearly follow up. The authors label the eight year result exploratory. That word is not decoration. It means the trial was not designed to answer this question with this data, and the result should be read as descriptive.

2. How mature is the data? Twenty six percent overall. Twenty percent of osimertinib patients had died, 31 percent of placebo patients. Three quarters of the events that will eventually define this curve have not happened yet.

3. Who was actually followed? Of 558 patients alive at the 2023 analysis, 431 had additional survival data, 77 percent. The other 127 are censored at their last known alive date from 2023. Then the detail that matters most: in the placebo arm, patients who provided follow up were far more likely to have been disease free in 2022 than those who did not, 56 percent versus 27 percent. In the osimertinib arm the gap was small, 81 versus 77. Censoring was not random, and it was more selective in one arm. The investigators reason that this likely inflated placebo survival, making the reported difference conservative. That is a defensible argument. It is also an argument, and you should know it is there.

4. Which subgroup intervals cross one? Exon 19 deletion, hazard ratio 0.45, interval 0.29 to 0.68. L858R, hazard ratio 0.72, interval 0.46 to 1.11. Men, 0.74, interval 0.44 to 1.22. Stage IB, 0.50, interval 0.23 to 1.01. The report says benefit was directionally consistent across subgroups. That is true. Directionally consistent is not the same as demonstrated in each subgroup, and the L858R and male intervals include no effect.

5. What stopped being collected? Disease progression data after April 2022. From that point the trial knows whether patients are alive, not what happened to their cancer. Anything said about late recurrence patterns is inference.

None of this diminishes ADAURA. Three years of adjuvant osimertinib remains the standard and this update supports it. The point is that you can hold both things at once: this is the longest OS follow up of any global phase 3 adjuvant trial in this population, and it is an exploratory analysis with 26 percent maturity, 77 percent follow up capture, informative censoring the authors had to argue around, and subgroups whose intervals include one. A reader who sees only the first sentence and a reader who sees only the second are both wrong.

The abstract tells you what was found. The methods section tells you how much of it you are allowed to believe. Read the second one first.

Now the surgical part, because there is one.

A separate real world analysis presented at the same meeting found that only about one third of patients remain on adjuvant osimertinib at three years, and that stopping early was associated with more than double the risk of recurrence or death. The randomized trial gave everyone the chance to complete three years. Practice does not. The operation you perform is the entry ticket to this drug. Whether the patient actually collects the eight year benefit is decided in the eighteen months after they leave your clinic, mostly by people who are not you. If you are the surgeon, the most useful thing you can do at the six week visit may be to ask one question: are you still taking it.

What to do with this

Before you cite a long term update in a manuscript, a tumor board, or a conversation with a patient, run the five questions. It takes ten minutes with the paper open. If you can state the maturity, the follow up capture, and which subgroup intervals cross one, you understand the result. If you can only state the headline hazard ratio, you are repeating it.

Part Three · Cardiothoracic Surgery News

Four Days in Seoul

The World Conference on Lung Cancer ran September 12 to 15. Three results with direct surgical consequence.

SWOG S1827 MAVERICK · Presidential Symposium, September 13

Prophylactic cranial irradiation in small cell lung cancer just lost its default status

304 patients with small cell lung cancer and no brain metastases after initial therapy were randomized to MRI surveillance alone or MRI surveillance plus prophylactic cranial irradiation. MRI alone improved the primary endpoint, cognitive failure free survival, hazard ratio 0.60 with a 90 percent confidence interval of 0.46 to 0.78, in both limited and extensive stage disease. Preliminary overall survival at 128 deaths showed no significant difference, hazard ratio 0.90, interval 0.67 to 1.20. Omitting PCI did what you would expect to intracranial control: brain metastases at six and twelve months were 21 and 30 percent with MRI alone versus 6.9 and 15 percent with PCI. The investigators concluded MRI surveillance alone should be the standard. Two things to hold onto: the trial reports 90 percent intervals by design, and the survival analysis is preliminary. PCI became standard in a pre MRI era. That era is over, but the final survival answer is not in yet.

NAUTIKA1, ALK positive cohort · Primary analysis

Eight weeks of alectinib before resection: surgery stayed feasible, and a third of nodes converted

48 patients with resectable stage IB to IIIB ALK positive disease received eight weeks of neoadjuvant alectinib, then surgery, optional platinum chemotherapy, and two years of adjuvant alectinib. Major pathologic response 55.8 percent, complete pathologic response 18.6 percent, conversion to ypN0 in 31.1 percent, R0 resection in 95 percent of operated patients, with low reported rates of perihilar fibrosis or intraoperative complication. Two year event free survival was 94 percent at a median follow up of 20.8 months. Phase 2, single arm, and a registration path based on survival is unlikely, per the presenter. But the operational finding matters for anyone who operates: a targeted agent given before resection did not make the resection harder, and it changed nodal stage in a third of patients before the surgeon ever saw the field.

ARTEMIS-008 and TAISHAN-302 · Two phase 3 trials

Relapsed small cell lung cancer finally moves past topotecan

Two separate antibody drug conjugates targeting B7-H3, risvutatug rezetecan and tambotatug pelitecan, each reduced the risk of death by 54 percent compared with topotecan in relapsed small cell lung cancer in independent phase 3 trials presented the same week. Neither is approved. Two concordant phase 3 results against the same target is the kind of signal that changes what the thoracic surgeon's tumor board is discussing by next spring.

Whether you are walking into interview season, deciding how much weight a long term update deserves in your next manuscript, or working out what four days in Seoul change in your own practice, that is the kind of thinking AskDimitri is built for.

Work through it with AskDimitri →

Editor's note: in October 2026 the figures in this letter were checked against the sources listed below, and the text was corrected where the letter as sent differed from them.

Until next week,

Sources

A new letter goes out every week, to subscribers first. Join to get the next one.