A patient came to me recently with rectal cancer that had already spread to the liver. Multiple small deposits, both lobes involved. Not curable by surgery on the day the diagnosis was made — but very much treatable, potentially for years, if the sequence was correct. The plan I laid out, after the multidisciplinary discussion and the biomarker workup, was the version most modern oncology textbooks would recommend. Start systemic chemotherapy. Watch how the biology responds over the first three months. Then decide — with the surgical oncologist and the radiation oncologist in the same room — whether the rectum, the liver, or both, was worth an operation.
The family wanted a second opinion. That is not only reasonable, it is what I would want if the diagnosis were my parent. They went to a general surgeon in Hyderabad — one who advertised online as an onco-surgeon, though the training actually listed on the website did not read like a formally trained surgical oncologist. I do not know what the surgeon believed about themselves. I know what they told the family. Let's take the rectal tumour out first. Then you go get the chemo. And then, if the liver settles with chemo, we will think about the liver later.
Not one step of that plan is what stage IV rectal cancer with liver metastases needs. Not one.
The family came back to me the next day. I took a prescription pad and drew the sequence. Chemotherapy first, because the disease was in more than one place and a drug reaches every place. Response checked at three months. Only then does surgery enter — of the rectum, of the liver, or of both — and only if the response tells us it will help.
Somewhere in India this week, another patient with the same disease will not get that second conversation. They will walk into a local doctor who decides, in five minutes with a CT film, that the tumour is localised enough to remove. They will be operated. They will be handed a report, a scar, and a familiar sentence — you take on from here. The surgeon will believe it was the right thing. It will not have been.
This is not a story about incompetent surgeons. Almost all of them are technically excellent. It is a story about a mental model that has not caught up with the last decade of trials. The model is: cancer is a thing you can see, and the treatment is to physically remove the thing. Older than any of us. Ingrained in medical education, in patient expectation, in the aesthetics of curative care. When we can see it and reach it, we cut. When we cannot see it or reach it, we send it to the oncologist.
That model was true, once. It is no longer true across most of the cancers we treat.
Three words that matter — neoadjuvant, adjuvant, perioperative
Three words come up throughout this piece. Each names a decision about when a treatment is given, relative to the operation. They are not interchangeable.
Neoadjuvant means the treatment is given before the surgery. It shrinks the tumour, and — more importantly — it shows us how the biology responds to the drug. Information the operation itself would destroy.
Adjuvant means the treatment is given after the surgery. It goes after the microscopic disease left behind, once the surgeon has removed everything visible.
Perioperative means both — drugs given before and after the operation. The modern default in several tumour types.
For most of the last century, the sequence was surgery first, adjuvant after. The past decade has been a slow rewiring. Move the treatment forward. Assess how the cancer responds. Personalise the second half of therapy — or, in a small handful of cases, decide the surgery is not needed at all.
That is the argument this piece is about.
The pattern, in three signals
One. In one solid tumour — MMR-deficient locally advanced rectal cancer — the surgery itself has become optional. 49 out of 49 patients treated with a single immunotherapy drug for six months had no visible cancer left. 92 out of 100 remained cancer-free at two years. No surgery. No radiation. No chemotherapy. Cercek and colleagues, NEJM 2025. Already in NCCN v2.2026. In this specific disease, in this specific molecular subtype, operating first is medicine that has run out its clock.
Two. In three of the largest solid tumours — lung, breast, bladder — giving chemotherapy plus immunotherapy before the operation has flipped five-year survival curves that were considered ceilings.
- Non-small-cell lung cancer — 10 more patients per 100 alive at five years, when the drugs come before the operation (CheckMate 816, NEJM 2025).
- Triple-negative breast cancer — 9 more women per 100 cancer-free at five years (KEYNOTE-522, updated 2026).
- Muscle-invasive bladder — 7 more patients per 100 alive at two years (NIAGARA, NEJM 2024).
Ten. Nine. Seven. Every count is a person.
Three. The neoadjuvant window is prognostic information that can only be harvested once. Pathologic complete response — the disappearance of the tumour by the time the surgeon opens the specimen — is now the single most important predictor of long-term cure we have. In CheckMate 816, 95 out of 100 patients whose lung tumour disappeared before surgery were alive at five years. Only 56 out of 100 whose tumour did not disappear were. That gap — 39 patients per 100 — is information you cannot see after the operation, because there is nothing left to look at. Surgery-first destroys the window before we can look through it.
These are not fringe trials. They are the trials that guidelines are written from. And the sequence they mandate — chemotherapy or immunotherapy first, surgery in the middle, response-directed treatment after — is the sequence that most of my referred-in-after-surgery patients did not receive.
The tumours where the shift has happened
Breast — TNBC and HER2-positive. Every stage II or III triple-negative or HER2-positive breast cancer that goes to surgery first has forfeited two of the biggest survival levers of the decade. In TNBC, adding pembrolizumab to neoadjuvant chemo pushes 14 more women per 100 to no visible tumour by the day of surgery, and keeps 9 more per 100 cancer-free at five years (KEYNOTE-522). In HER2-positive residual disease — the group we can only identify by doing neoadjuvant first and then examining the specimen — switching to adjuvant T-DM1 keeps 14 more women per 100 cancer-free at seven years compared to trastuzumab (KATHERINE, NEJM 2024). Neither of those two levers is available without a neoadjuvant phase. You cannot personalise adjuvant therapy to a response you never measured. Pembrolizumab is CDSCO-approved for early TNBC. Trastuzumab, pertuzumab and T-DM1 are all available in India. Cost is the barrier the honest defenders name. Habit is the barrier the honest defenders do not.
Rectal — MMR-deficient. The single sharpest case in the entire literature. Six months of dostarlimab, 100 out of 100 patients had no visible cancer left. 92 out of 100 still cancer-free at two years. No surgery, no radiation, no chemotherapy. This was the Cercek MSK cohort. The 103-patient expanded series confirmed the finding. AZUR-1, the global confirmatory phase III, is running. In this molecular subtype, an operation is now a mistake — the equivalent of amputating a leg to treat an infection that an antibiotic would have cleared. And yet: dMMR/MSI testing is not reflex in most peripheral pathology labs in India. The ₹1,500 stain that identifies the eligible patient does not get run. The patient goes to the operating theatre, and the modern therapy is never offered.
Non-small-cell lung — resectable stage II to IIIB. CheckMate 816 is the flagship. Three cycles of nivolumab plus chemotherapy before lobectomy — 22 more patients per 100 with no visible tumour by the operating day, and 10 more per 100 alive at five years. Add KEYNOTE-671 (perioperative pembrolizumab, 23 more per 100 event-free at five years), AEGEAN (durvalumab), CheckMate 77T (nivolumab), and now RATIONALE-315 (tislelizumab, 10 more per 100 alive at five years, presented at ASCO 2026) — and this is a class effect, not a single-trial fluke. And still, in most Indian centres, a resectable NSCLC gets a lobectomy first, because that is what the operative slate expects.
Muscle-invasive bladder. The old standard was neoadjuvant cisplatin-based chemotherapy before radical cystectomy (Grossman, NEJM 2003). Retrospective Indian series suggest fewer than a third of muscle-invasive bladder cancer patients even received that. NIAGARA (Powles, NEJM 2024) has made the standard sharper — adding perioperative durvalumab to the neoadjuvant gem-cis backbone keeps 7 more patients per 100 alive at two years. Radical cystectomy is not the first step. It is the third step, after neoadjuvant chemotherapy and immunotherapy.
Esophageal — adenocarcinoma. For adenocarcinoma of the oesophagus and the gastroesophageal junction — the version we increasingly see in Indian metros — the argument was settled in January 2025 by ESOPEC (Hoeppner, NEJM 2025). Perioperative FLOT chemotherapy beat the older neoadjuvant chemoradiation regimen on every survival endpoint that mattered: median overall survival 66 months versus 37; five-year overall survival 51 out of 100 versus 39 — 12 more people per 100 alive a full five years later, from switching the sequence and swapping the modality. FLOT before, FLOT after, surgery in the middle. Not straight to esophagectomy.
Esophageal — squamous cell. For squamous-cell carcinoma — still the more common histology in India, driven by tobacco and hot beverages — ESOPEC did not enrol these patients. The older CROSS regimen (van Hagen, NEJM 2012; Eyck, JCO 2021 ten-year data) remains the standard: neoadjuvant chemoradiation followed by surgery, 13 more people per 100 alive at ten years compared to surgery alone. Different regimen, same principle. Not straight to esophagectomy.
Gastric / gastroesophageal. MATTERHORN (Janjigian, NEJM 2025) settled a debate the Asian centres held open for a decade — perioperative FLOT plus durvalumab tripled the rate of no-visible-tumour-at-surgery (19 per 100 versus 7) and reduced the risk of death by 22% compared to FLOT alone. And here the honest problem is doubled. In much of India, gastric cancer still goes to upfront surgery. But outside a handful of high-volume centres, the operation is often not a formally trained D2 lymphadenectomy — the Japanese-perfected technique that anchored every Asian trial and defines the modern oncologic operation. The patient loses on both sides. The systemic therapy comes late, and the surgery is not the surgery the data was generated from. FLOT is affordable. Durvalumab is not, but is available. The default sequence was defensible in 2015. It is not defensible in 2026.
Melanoma — resectable stage III. SWOG S1801 (Patel, NEJM 2023) showed that giving three doses of pembrolizumab before surgery — the same total drug the patient would have received anyway — added 23 more patients per 100 to the two-year cancer-free rate. NADINA (Blank, NEJM 2024) took the argument further: neoadjuvant nivolumab plus relatlimab cut the one-year risk of recurrence by more than two-thirds, and in 6 out of 10 patients made a full year of adjuvant therapy unnecessary. Same total drug, different sequence, better outcomes. In India, this matters most for acral and mucosal melanomas, which are our common variants. The reflex is still wide excision followed by block dissection.
Pancreatic — borderline resectable. PREOPANC-1 (Versteijne, final analysis JCO 2022) turned a five-year survival of 6 per 100 into 20 per 100 in the borderline-resectable subgroup — 14 more patients per 100 alive at five years, simply from giving chemotherapy before the Whipple, not after. Not from a new drug. FOLFIRINOX is off-patent and affordable. PREOPANC-2 (Groot Koerkamp, Lancet Oncol 2025) refined the choice of neoadjuvant regimen. What both trials removed from the table is the surgery-alone option for borderline disease. And yet borderline resectability is often over-called by peripheral surgical judgment — and the Whipple happens on day two.
Head and neck. The largest cancer surgery culture in India, and the slowest to adopt the neoadjuvant shift. KEYNOTE-689 (Uppaluri, NEJM 2025) made perioperative pembrolizumab standard of care for PD-L1-positive resectable HNSCC. The regimen is not affordable everywhere. But there is an Indian answer that is: metronomic oral chemotherapy — pioneered at Tata Memorial Hospital, refined at Kerala's regional cancer centres, presented again at ASCO 2026 as triple oral metronomic therapy (Kapoor et al., abstract 17195). A neoadjuvant regimen that costs less than a family lunch, deliverable at any peripheral centre without an infusion unit. And still under-used.
The one honest exception
In advanced epithelial ovarian cancer, primary debulking surgery remains preferred when an experienced gynae-oncology team can reliably achieve a complete gross resection. The trials that established interval debulking as an alternative — EORTC 55971, CHORUS, SCORPION, and this year's CHRONO update at ASCO 2026 — did not overturn primary surgery. They said: when the surgical team cannot achieve a clean margin, when the tumour burden is very high, when the patient is too fragile, neoadjuvant chemotherapy first is safer.
The trouble is that most gynaecological cancer in Indian peripheral practice is operated by general gynaecologists, not fellowship-trained gynae-oncology surgeons. In that setting, the failure mode of primary surgery is suboptimal cytoreduction — leaving residual tumour behind that ovarian cancer cannot recover from. So even the exception to the neoadjuvant argument proves the pattern: outcomes are not defended by the surgeon's confidence. They are defended by the team's ability to deliver the correct sequence.
Why the reflex persists
Because the mental model of cancer as a physical object to be removed is intuitive, older than any of us, and reinforced every time the excised tumour is held up to the family in a jar to prove the operation was a success. Because a scan that shows no visible disease after surgery feels like proof of cure, even when the trial data says the trickiest disease is what the scan does not show. Because the surgical incentive structure — operative volume, technical mastery, immediate visible outcome — is easier to see and to reward than the medical-oncology decision to not operate for six months.
Because the peripheral centres that see the majority of Indian cancer patients still lack the multidisciplinary tumour boards where the sequencing decision would be made by three specialists together, not by one alone. Because a general surgeon in a district hospital who consults a medical oncologist before operating is doing an extra piece of work with no immediate reward, and one who does not is doing the operation they were trained to do.
And because we, as a professional culture, have not been sharp enough with each other about the cost. The oncology community writes review articles about neoadjuvant sequencing. It does not put up hoardings at surgical conferences that say: if you operate on this patient without a medical oncology consult first, you are shortening their life. Perhaps we should.
What the patient should ask
Before any cancer operation — before the paperwork is signed, before the anaesthesia consent is taken — the patient and family are entitled to one question. It is the shortest useful question in oncology.
Has a medical oncologist reviewed the plan?
If the answer is no, the operation can wait. Cancer is rarely a same-week emergency. The rare exceptions — obstructing tumours, uncontrolled bleeding, spinal cord threat — are obvious to any doctor who has spent time in oncology. What is not obvious is that most cancers, including the ones that feel urgent, are best served by an eight-to-ten-day pause for correct staging, biomarker testing, and multidisciplinary review. That pause almost never worsens the disease. Skipping it often does.
The second question is the one that saves the second life. If it is a solid tumour where neoadjuvant treatment is standard, why are we operating first? If the answer is a shrug, an appeal to the surgeon's judgement, or the phrase let's just get it out — the family is being asked to trust habit, not evidence.
The sequence is the medicine
The oldest lesson in cancer surgery — do not open, do not resect — is still true. It just needs an addition. Do not operate first, unless the evidence for this specific cancer, in this specific subtype, in this specific patient, says operate first. For an increasing majority of solid tumours, the evidence now says something else.
Not every cancer is cut first. Not every cancer is cut at all.
The reflex to reach for the scalpel is not going to disappear from Indian oncology in one year, or five. But the patients who benefit from the modern sequence are not going to wait for medicine's culture to catch up either. They are going to keep walking into oncology offices with fresh scars, and the oncologists are going to keep having the conversation that begins: if I had seen you before the operation, we would have done this differently.
That conversation is the quiet cost of the reflex. It is time the reflex was named.