An adult outpatient with lung cancer is about to start a new chemotherapy regimen. The pre-chemotherapy platelet count is 380 × 10⁹/L. Haemoglobin, leukocyte count and BMI are all unremarkable, and no red-cell growth factor is in use.
Lung cancer scores 1 point and a platelet count of 350 × 10⁹/L or more scores 1, so the Khorana score is 2.
What 2 means depends on which line you read it against. In the 2008 derivation paper, a score of 1 to 2 is the intermediate risk group; "high" starts at 3.1 But the two large placebo-controlled trials in this population, CASSINI and AVERT, both enrolled patients with a score of 2 or more,23 and the 2022 ITAC guidelines and 2019 ISTH SSC guidance describe that same ≥2 line, as does the 2020 ASCO guideline.459 This guide sets the two thresholds side by side, then explains why published rates for the same band differ so much from study to study.
Illustrative case: synthetic, and any criterion not mentioned is absent.
At a glance
- The Khorana score adds five factors (cancer site, platelets, haemoglobin or red-cell growth factor, leukocytes, BMI) for a total of 0 to 6, and places an outpatient starting a new regimen in one of three risk groups for symptomatic VTE.
- The paper's bands are 0 / 1–2 / ≥3. The later trials and three guidance documents use ≥2. Both are real, and they answer different questions.
- Published rates for each band come from studies with different endpoints, follow-up windows, cohorts and methods. In the derivation cohort and in a large meta-analysis, most clots occurred in patients scoring below 3.
What the Khorana score measures
The score sorts adult outpatients at the start of a new chemotherapy regimen by how often symptomatic venous thromboembolism followed in the derivation cohort. Khorana and colleagues built it in 2008 from a prospective US registry: 2,701 patients in the derivation set and 1,365 in a validation set drawn from the same study.1 Patients were enrolled between 2002 and 2005 and followed for up to four cycles of chemotherapy.1
The outcome is symptomatic VTE (deep-vein thrombosis or pulmonary embolism) diagnosed by the treating clinician and not independently adjudicated.1 Follow-up was a median of 2.5 months (73 days), not a fixed horizon, and 75% of events occurred in the first two cycles.1
It is a points model. Its C-statistic was 0.7 in both the derivation and validation sets.1 In 40,218 Danish patients starting chemotherapy it was 0.60.6
Criteria and points
| Item | Points | Definition used in Curie Calc |
|---|---|---|
| Cancer site: stomach or pancreas | 2 | Primary site of the cancer being treated. Stomach includes gastro-oesophageal junction. |
| Cancer site: lung, lymphoma, gynecologic, bladder or testicular | 1 | Bladder includes upper-tract urothelial. Any other site scores 0, including breast, colorectal, prostate, head and neck, brain, kidney, myeloma, oesophageal, cholangiocarcinoma and ampullary cancer. |
| Platelets | 1 | Pre-chemotherapy platelet count of 350 × 10⁹/L (350,000/µL) or more. |
| Haemoglobin or red-cell growth factor | 1 | Baseline haemoglobin below 10 g/dL (100 g/L), or use of a red-cell growth factor. Counts once if either applies. |
| Leukocytes | 1 | Pre-chemotherapy leukocyte count of more than 11 × 10⁹/L (11,000/µL). Exactly 11 is not counted. |
| BMI | 1 | 35 kg/m² or more. |
Only one site is scored, so the maximum is 6. The original score does not address two primary cancers; Curie Calc scores the primary site of the cancer being treated. The paper groups totals into three bands: low (0), intermediate (1–2) and high (3 or more).1 It does not publish a rate for each individual score.
Source: Khorana 2008, Table 3 and Results.1
Details worth checking
Which genitourinary cancers score a point?
In the scoring table, bladder and testicular only. The paper is not consistent with itself here. Its scoring table (Table 3) lists "lung, lymphoma, gynecologic, bladder, testicular" as the 1-point sites, while its Results text and Table 2 describe the group as genitourinary excluding prostate, which would also take in kidney and other urothelial cancers.1 Curie Calc keys to Table 3, so kidney cancer scores 0, as do brain tumours and myeloma. Upper-tract urothelial cancer is treated as bladder.
The authors noted that brain, renal and myeloma patients were too few in their cohort, and suggested the high-risk-site score "may be generalized" to brain tumours.1 AVERT later used a modified score that gave points to brain tumours, myeloma and kidney cancer.5 A score computed with a modified site list is a different score from the one whose rates the paper reports.
Is the leukocyte cutoff 11 or more, or more than 11?
More than 11. Table 3 reads "more than 11 × 10⁹/L".1 A count of exactly 11.0 × 10⁹/L scores 0. The platelet item is the other way round, "350 × 10⁹/L or more", so a count of exactly 350 scores 1.1 Haemoglobin is "less than" 10 g/dL and BMI is "35 or more". Each threshold has its own direction.
When is red-cell growth-factor use counted?
In the original study, up to the start of cycle 2. The haemoglobin item combines low baseline haemoglobin and growth-factor use into one point.1 The paper recorded growth-factor use up to cycle 2, so in the derivation data this item was not purely a pre-treatment measurement, unlike the platelet and leukocyte counts, which were taken before chemotherapy.1 The item scores once even if both parts apply.
Does "2.5 months" mean the rates are 2.5-month risks?
Not exactly. The paper reports rates "over a median of 2.5 months" of follow-up, which ranged from 5 to 364 days.1 That is a median follow-up in a cohort tracked for up to four cycles, not a fixed 2.5-month horizon, and it is shorter than the six-month window used by most later studies.
Khorana 2 or 3? The two thresholds side by side
The 2008 bands describe observed risk groups; the ≥2 line comes from trial entry criteria and later guidance. The paper had no "≥2" group and published no rate for one.1
| Framework | Body and year | What it uses | The opening case (score 2) |
|---|---|---|---|
| Derivation bands | Khorana 20081 | 0 low · 1–2 intermediate · ≥3 high | Intermediate |
| CASSINI trial entry | Khorana 20192 | Score ≥2, described as high risk | Meets the score criterion (one of several entry criteria) |
| AVERT trial entry | Carrier 20193 | Score ≥2, described as intermediate-to-high risk (modified site list) | Meets the score criterion (one of several entry criteria) |
| ITAC guidelines | Farge 20224 | Score ≥2 identifies intermediate-to-high risk | In that group |
| ASCO guideline | Key 20209 | Score ≥2 before a new regimen, described as high risk (kept in the 2023 update) | High risk |
| ISTH SSC guidance | Wang 20195 | Notes that both main trials enrolled at ≥2 | At the trials' line |
| ASH guidelines | Lyman 20217 | Low, intermediate or high classification based on a validated tool, naming the Khorana score, complemented by clinical judgement | Classification by the tool |
The same patient is "intermediate" in one row and at or above the line in the others. The framework travels with the number: a result reported as "Khorana 2" is clearer when it says which reading is meant.
ASCO's 2023 guideline update re-reviewed two questions only, perioperative thromboprophylaxis and VTE treatment, so the outpatient question from its 2020 guideline was not revisited and no threshold changed in that update.89
Curie Calc shows the paper's three bands and lists the ITAC, ISTH, ASCO and ASH positions beside the result as context.
Reading the result: 7.1%, 11%, or 4.1%?
The three most-quoted figures for a "high" Khorana score measure different things. The rows below differ in endpoint, window and cohort, so they are not directly comparable with each other; read each row on its own terms.
| Source | Cohort | Endpoint | Window | Score 0 | 1–2 | ≥3 |
|---|---|---|---|---|---|---|
| Khorana 2008, derivation1 | 2,701 US outpatients starting chemotherapy, 2002–2005 | Symptomatic VTE, clinician-diagnosed | Median 2.5 months | 0.8% | 1.8% | 7.1% |
| Khorana 2008, validation1 | 1,365 patients, a random split of the same registry | Symptomatic VTE, clinician-diagnosed | Median 2.5 months | 0.3% | 2% | 6.7% |
| Mulder 2019, meta-analysis10 | 55 cohorts, 34,555 ambulatory patients (6-month data for 27,849) | Radiologically confirmed symptomatic or incidental VTE | 6 months | 5.0% | 6.6% | 11.0% |
| Overvad 2022, registry6 | 40,218 Danish outpatients starting chemotherapy | VTE recorded in national registries, death as a competing risk | 6 months | 1.5% | 2.8% | 4.1% |
Group averages in each cohort, not forecasts for an individual patient.
Three differences sit behind the spread:
- Window. The Khorana 2008 figures cover a median 2.5 months; the other two sources cover six months.1610
- Endpoint. The derivation counted symptomatic clots only. In the meta-analysis, 60% of the included studies also counted clots found incidentally on imaging, which the authors note differs from the derivation study's outcome.10
- Method and setting. The Danish study identified clots from registry codes and treated death as a competing risk; in the same data, Kaplan-Meier analysis overestimated incidence by up to 23%.6 Its rates were the lowest of the three, and its authors note they were lower than in key randomised trials.6
For the ≥2 line itself, Overvad reported a six-month risk of 3.6% in patients scoring 2 or more.6
Curie Calc shows the derivation and validation figures (0.8%, 1.8%, 7.1% and 0.3%, 2%, 6.7%), with the endpoint, median follow-up and cohort size in every result line. They are the paper's own rates for its own model, and all use the same endpoint and window.
Most clots occur below 3. In the derivation cohort, 40% of patients who developed VTE had scored 3 or more.1 In the meta-analysis, 23.4% of patients with VTE in the first six months had scored 3 or more; the authors write that "most events occur outside this high-risk group."10 A low or intermediate band is a group average and does not rule out a clot.
Worked examples
Synthetic cases. Any criterion not mentioned is absent. Each result was computed by the calculator's code.
| Case | Calculation | Total | Learning point |
|---|---|---|---|
| Pancreatic cancer starting a new regimen, BMI 36 kg/m², labs unremarkable | Pancreas 2 + BMI 1 | 3 (high) | A very-high-risk site plus one other item reaches the paper's high band. |
| Breast cancer, pre-chemotherapy leukocytes exactly 11.0 × 10⁹/L, haemoglobin 9.6 g/dL | Breast 0 + leukocytes 0 + haemoglobin 1 | 1 (intermediate) | Leukocytes must be above 11; breast is "any other site". |
| Kidney (renal cell) cancer, pre-chemotherapy platelets 360 × 10⁹/L | Kidney 0 + platelets 1 | 1 (intermediate) | Under the Results text's "genitourinary excluding prostate", kidney would score 1 and the total would be 2. |
What the score doesn't tell you
The score was built for adult outpatients starting a new chemotherapy regimen.1 Curie Calc lists these as populations where it should not be used: inpatients; patients not starting a new regimen, including those on continuous single-agent chemotherapy; acute leukaemia or after stem-cell transplant; pregnancy or lactation; under 18; and as a test for a suspected current clot.
It leaves out prior VTE, central venous catheters, current anticoagulant or antiplatelet use, specific chemotherapy agents, cancer stage, performance status, recent surgery, immobility and D-dimer. It does not assess bleeding risk.
Performance varies by cancer and by treatment.
- Lung cancer. The meta-analysis found the score less informative in lung and haematological cancers.10 In pooled trial control arms it did not separate risk in lung cancer (odds ratio 1.1) while it did in other cancers (3.2),11 and in the Danish registry the ≥2 line did not stratify lung, hepatobiliary or pancreatic, or gynecologic cancer.6
- Checkpoint inhibitors (ICI). The evidence is mixed. In 3,946 Danish patients starting a checkpoint inhibitor, six-month VTE incidence (with death as a competing risk) was 2.1%, 2.6% and 3.7% for scores 0, 1–2 and ≥3, cohort averages,12 while a Vienna cohort of 672 patients found no association with the score,13 and a lung-cancer cohort found it "did not identify high-risk ICI-treated patients."14
- Targeted therapy. The derivation registry enrolled patients in 2002 to 2005.1 A 2025 prospective cohort that included patients on targeted agents (19.7%) reported a C-statistic of 0.53 for Khorana across all therapies, without a separate result for targeted agents.15
Newer scores exist. EHR-CAT reported a C-statistic of 0.70 to 0.71 against 0.63 for Khorana in 732,594 US patients,16 and the Vienna CATScore, which uses tumour type and D-dimer, reported 0.69 in 598 patients on contemporary therapy.17 Both validations were published after the ASH 2021 and ITAC 2022 guidelines cited above.471617
Frequently asked questions
What does a Khorana score of 2 mean?
In the 2008 paper, a score of 1 or 2 is the intermediate risk group. In its derivation cohort (n=2,701), 1.8% of that group developed symptomatic VTE over a median 2.5 months, and 2% in its validation cohort (n=1,365), a group average rather than an individual forecast.1 A score of 2 is also the entry line used by the CASSINI and AVERT trials, and the threshold the 2022 ITAC guidelines use for intermediate-to-high risk.234 The 2020 ASCO guideline uses the same line before a new regimen and describes it as high risk.9
Is a high Khorana score 3 or more, or 2 or more?
Does the Khorana score work in lung cancer?
Does the Khorana score apply to patients on immunotherapy?
Sources
- Khorana AA, Kuderer NM, Culakova E, Lyman GH, Francis CW. Development and validation of a predictive model for chemotherapy-associated thrombosis. Blood. 2008;111(10):4902-4907. (opens in new tab)
- Khorana AA, Soff GA, Kakkar AK, et al. Rivaroxaban for thromboprophylaxis in high-risk ambulatory patients with cancer. N Engl J Med. 2019;380(8):720-728. (opens in new tab)
- Carrier M, Abou-Nassar K, Mallick R, et al. Apixaban to prevent venous thromboembolism in patients with cancer. N Engl J Med. 2019;380(8):711-719. (opens in new tab)
- Farge D, Frere C, Connors JM, et al. 2022 international clinical practice guidelines for the treatment and prophylaxis of venous thromboembolism in patients with cancer, including patients with COVID-19. Lancet Oncol. 2022;23(7):e334-e347. (opens in new tab)
- Wang TF, Zwicker JI, Ay C, et al. The use of direct oral anticoagulants for primary thromboprophylaxis in ambulatory cancer patients: guidance from the SSC of the ISTH. J Thromb Haemost. 2019;17(10):1772-1778. (opens in new tab)
- Overvad TF, Ording AG, Nielsen PB, et al. Validation of the Khorana score for predicting venous thromboembolism in 40 218 patients with cancer initiating chemotherapy. Blood Adv. 2022;6(10):2967-2976. (opens in new tab)
- Lyman GH, Carrier M, Ay C, et al. American Society of Hematology 2021 guidelines for management of venous thromboembolism: prevention and treatment in patients with cancer. Blood Adv. 2021;5(4):927-974. (opens in new tab)
- Key NS, Khorana AA, Kuderer NM, et al. Venous thromboembolism prophylaxis and treatment in patients with cancer: ASCO guideline update. J Clin Oncol. 2023;41(16):3063-3071. (opens in new tab)
- Key NS, Khorana AA, Kuderer NM, et al. Venous thromboembolism prophylaxis and treatment in patients with cancer: ASCO clinical practice guideline update. J Clin Oncol. 2020;38(5):496-520. (opens in new tab)
- Mulder FI, Candeloro M, Kamphuisen PW, et al. The Khorana score for prediction of venous thromboembolism in cancer patients: a systematic review and meta-analysis. Haematologica. 2019;104(6):1277-1287. (opens in new tab)
- van Es N, Ventresca M, Di Nisio M, et al. The Khorana score for prediction of venous thromboembolism in cancer patients: an individual patient data meta-analysis. J Thromb Haemost. 2020;18(8):1940-1951. (opens in new tab)
- Overvad TF, Skjøth F, Piazza G, et al. The Khorana score and venous and arterial thrombosis in patients with cancer treated with immune checkpoint inhibitors: a Danish cohort study. J Thromb Haemost. 2022;20(12):2921-2929. (opens in new tab)
- Moik F, Chan WE, Wiedemann S, et al. Incidence, risk factors, and outcomes of venous and arterial thromboembolism in immune checkpoint inhibitor therapy. Blood. 2021;137(12):1669-1678. (opens in new tab)
- Icht O, Darzi N, Shimony S, et al. Venous thromboembolism incidence and risk assessment in lung cancer patients treated with immune checkpoint inhibitors. J Thromb Haemost. 2021;19(5):1250-1258. (opens in new tab)
- Vladić N, Englisch C, Berger JM, et al. Validation of risk assessment models for venous thromboembolism in patients with cancer receiving systemic therapies. Blood Adv. 2025;9(13):3340-3349. (opens in new tab)
- Li A, Jafari O, Lam BD, et al. Validation of a risk score for cancer-associated thrombosis using nationwide EHR data. JAMA Netw Open. 2025;8(11):e2544428. (opens in new tab)
- Englisch C, Nopp S, Moik F, et al. The Vienna CATScore for predicting cancer-associated venous thromboembolism: an external validation across multiple time points. ESMO Open. 2025;10(2):104130. (opens in new tab)
This article is educational reference material for healthcare professionals. It does not provide medical advice, diagnosis, or treatment recommendations, and it does not decide thromboprophylaxis for any individual patient. The underlying Curie Calc Khorana calculator carries independent clinical review; this article is drawn from that reviewed source and the cited studies and guidelines.