Scoring Methodology
OralCheck assigns a risk weight to each answer based on published odds ratios. This page documents how that scoring works, which sources it's built on, and where the model has known limitations.
Last reviewed: August 2026
Weight Derivation
Each risk factor is assigned a weight proportional to the natural log of its published odds ratio (OR), scaled by a constant k:
The scaling constant k = 4.47 was chosen so that daily tobacco use (the highest-weighted modifiable factor, OR 6.0× per Gandini et al. 2008) maps to a weight of 8. That anchors all other weights to a consistent ordinal scale from the same evidence base.
Where studies report a range of ORs, we use the midpoint or a conservative estimate. Symptoms like erythroplakia, leukoplakia, and non-healing ulcers are treated two ways at once: they add to the score, and they also override the tier to See-a-Dentist-Soon regardless of the total. They're better understood as possible in-situ pathology than as population-level risk exposures, so a reported symptom always surfaces the same recommendation even when every other factor is low.
Risk Factor Weights
| Factor | Published OR | Weight | Source |
|---|---|---|---|
| Betel quid / paan / gutka (current) IARC Group 1 carcinogen independent of tobacco | 7 – 10× | 9 | IARC Monograph 85, 2004 |
| Tobacco (daily) Anchor for the scaling constant k = 4.47 | 2.5 – 6.0× | 8 | Gandini et al., Oral Oncology, 2008 |
| Age 65+ Median age at diagnosis is 65 | ~4.0× (adjusted) | 6 | SEER, NCI; multivariable-adjusted |
| Current symptom lasting 2+ weeks | override | 6 | Napier & Speight, J Oral Pathol Med, 2008 |
| Tobacco (occasional) | ~3.0× | 5 | Gandini et al., Oral Oncology, 2008 |
| Alcohol (daily) IARC Group 1 carcinogen | ~3.0× | 5 | Bagnardi et al., Annals of Oncology, 2015 |
| HPV-related history Conservative estimate; OR ~15× for oropharyngeal specifically | 3 – 5× (blended) | 5 | Gillison et al., JAMA, 2008 |
| Immunosuppression, transplant, or prior head/neck radiation Blended across three heterogeneous exposures rather than taken from a single published figure. This is the least precisely derived weight in the instrument. | 2 – 4× (blended) | 5 | Engels et al., JAMA, 2011; Grulich et al., Lancet, 2007 |
| Age 55 – 64 | ~2.5× | 4 | SEER, NCI |
| Betel quid (past use) | ~2.5× | 4 | IARC Monograph 85, 2004 |
| Alcohol (weekly) | ~2.0× | 3 | Bagnardi et al., Annals of Oncology, 2015 |
| Family history (first-degree) | ~2.0× | 3 | Negri et al., Eur J Cancer Prev, 2009 |
| Diet low in fruit/vegetables | ~2.0× | 3 | Pavia et al., Oral Oncology, 2006 |
| Male sex at birth Derived from a deliberately conservative OR of 2.0, not the 2.6× SEER rate ratio, because part of the male excess is tobacco and alcohol exposure scored separately. | ~2.0× (conservative) | 3 | SEER, NCI (17.5 vs 6.6 per 100,000) |
| Possible symptom (unsure) | override | 3 | Napier & Speight, J Oral Pathol Med, 2008 |
| Never seen a dentist | detection proxy | 3 | SEER stage-at-diagnosis distribution |
| Age 35 – 54 | ~1.5× | 2 | SEER, NCI |
| Sex not stated Population average: roughly half of respondents who decline are male. | population average | 2 | SEER, NCI |
| Tobacco (former) Risk declines ~50% within 5 years of cessation | ~1.5× | 2 | Gandini et al., Oral Oncology, 2008 |
| HPV (not vaccinated, no known history) Unvaccinated status is a behavioral proxy for HPV-16 exposure risk, not a serostatus measure. OR reflects elevated background exposure probability in unvaccinated adults, not confirmed infection. | ~1.5× (proxy) | 2 | D'Souza et al., NEJM, 2007; population exposure estimate |
| Sun exposure (lips, unprotected) Lower lip squamous cell carcinoma; outdoor workers vs indoor reference | 2 – 3× | 2 | Perea-Milla López et al., Br J Cancer, 2003 |
| Last dental visit 3+ years ago | detection proxy | 2 | SEER stage-at-diagnosis distribution |
| Last dental visit 1–3 years ago | detection proxy | 1 | SEER stage-at-diagnosis distribution |
| Alcohol (rarely) | ~1.2× | 1 | Bagnardi et al., Annals of Oncology, 2015 |
| Sun exposure (regular, sometimes protected) | ~1.5× | 1 | Perea-Milla López et al., Br J Cancer, 2003 |
| Family history (distant relative) | ~1.3× | 1 | Negri et al., Eur J Cancer Prev, 2009 |
| Fruit/vegetables a few times a week | ~1.3× | 1 | Pavia et al., Oral Oncology, 2006 |
| HPV status unknown | ~1.2× (proxy) | 1 | D'Souza et al., NEJM, 2007; population exposure estimate |
| Immune or radiation history (unsure) | ~1.2× (proxy) | 1 | Engels et al., JAMA, 2011; population exposure estimate |
Why Each Question Is Asked
Every question earns its place by changing the score in a way the published literature supports. This is the reasoning behind each one, including the two where the evidence is weakest.
Age
Incidence climbs steeply after 40 and the median age at diagnosis is 65. Age is the strongest non-modifiable predictor in the instrument, and it also works as a proxy for cumulative mucosal exposure to every other carcinogen on this list, which is why it carries weight even for someone who reports nothing else.
Sex at birth
Men are diagnosed at roughly 2.6 times the rate of women. Some of that gap is historically higher tobacco and alcohol use, which this screener already scores separately, so applying the full 2.6× would count the same exposure twice. The weight is derived from an OR of 2.0 instead. Sex at birth is asked rather than gender because that is what the incidence data recorded.
Tobacco
The anchor for the whole scale: the scaling constant was chosen so that daily use maps to a weight of 8. Both combustible and smokeless forms count. Former use keeps a non-zero weight because risk declines gradually after cessation rather than resetting, roughly halving over five years.
Alcohol
A Group 1 carcinogen for the oral cavity with a clear dose-response relationship. The carcinogen is acetaldehyde, ethanol's metabolite, which is why frequency of drinking matters more to the score than what is being drunk.
Betel quid, paan, and gutka
Asked separately from tobacco rather than folded into it, because betel quid is carcinogenic with or without added tobacco. It is the single highest-weighted exposure in the instrument. It is asked of everyone rather than gated behind a region question, since the screener is used well outside the countries where use is most common.
HPV
HPV-16 is now the leading cause of oropharyngeal cancer in adults under 50. A public tool cannot ask for serostatus, so vaccination status and reported history stand in as behavioral proxies. This is the weakest inferential step in the instrument, and the weight is set far below the true odds ratio for confirmed HPV-16 to reflect that.
Sun exposure
This one is about the lower lip specifically, not the inside of the mouth. Lip squamous cell carcinoma behaves more like a skin cancer than an oral one, and outdoor workers carry two to three times the risk of indoor workers.
Current symptoms
Treated differently from everything else on this page. A persistent sore, patch, lump, or difficulty swallowing is better understood as possible existing pathology than as a risk exposure, so it both adds points and overrides the tier outright. Someone with a symptom and no other risk factors still gets told to see a dentist.
Family history
A first-degree relative with head and neck cancer roughly doubles risk. Part of that is shared environment rather than inheritance, which is why the weight is modest and a distant relative counts for much less.
Immune and radiation history
Added in August 2026 after external clinical review flagged it as the clearest gap in the instrument. Organ and bone marrow transplant, prior head and neck radiation, and conditions or medications that suppress immunity are grouped into one question because they converge on the same mechanism: reduced immune surveillance of the oral mucosa, or in the radiation case, a field already exposed to a known carcinogen. They are grouped rather than split because any one of them leads to the same recommendation.
Diet
Low fruit and vegetable intake is associated with roughly double the risk. This is the weakest causal evidence among the modifiable factors, heavily confounded with smoking and socioeconomic status, and it is weighted accordingly.
Dental care
Not a risk factor at all, and the page is explicit about that. It is a detection-delay proxy. Over 60% of oral cancers are found at stage III or IV, and the two-minute visual exam that finds them earlier happens at a dental visit. Its weight comes from clinical rationale, not from an odds ratio, and it is the only entry in the table for which that is true.
Tobacco + Alcohol Interaction Term
When both tobacco and alcohol use are present at meaningful levels (daily or weekly), the model adds an interaction bonus of +3 points.
The reason is that tobacco and alcohol together produce a multiplicative rather than additive increase in oral cancer risk. Combined regular use raises risk roughly 15× above baseline, while simply summing the individual weights only gets you to about 9×. The +3 captures that gap.
Sources:Hashibe M, et al. Cancer Epidemiology, Biomarkers & Prevention, 2009 (heavy combined users: OR 35.8×); Bagnardi V, et al. Annals of Oncology, 2015.
Risk Tier Thresholds
The maximum possible score is 61 (all highest-risk answers plus the interaction bonus). A few reference points: a daily smoker alone scores 8 (moderate); tobacco + alcohol + interaction lands at 16 (elevated); betel + tobacco + alcohol + interaction reaches 25 (high). Male sex at birth alone scores 3, which is still low.
Regardless of score, any symptom lasting 2+ weeks (a persistent sore, red or white patch, unexplained lump, or difficulty swallowing) overrides the tier to High and triggers a prompt for immediate evaluation.
Thresholds are anchored to those reference profiles rather than rescaled whenever the maximum changes. Adding the sex and immune history questions in August 2026 raised the maximum from 53 to 61, and rescaling proportionally would have moved the High boundary to 26, quietly demoting an unchanged betel + tobacco + alcohol profile from High to Elevated. A new question should not make an existing profile look safer, so the boundaries stayed where they were.
It is worth being blunt about where these four numbers came from, because nothing derived them. They were set by judgment against the reference profiles above: a daily smoker should not read as low, someone using tobacco and alcohol together should land clearly above them, and a profile carrying several Group 1 exposures should reach the top tier. They were then checked against the distribution of plausible answer combinations so that the tiers were not all but empty or all but full. That is a defensible way to set a cutoff for a tool whose job is to prompt a dental visit, and it is not the same thing as a validated threshold. There is no sensitivity, no specificity, and no ROC curve behind these boundaries, because producing those requires outcome data this tool has never had access to.
One Score, Two Diseases
Both clinicians who reviewed this raised the same thing independently: oral cavity and oropharyngeal cancer have different dominant causes, and a single blended score serves neither cleanly. Two obvious fixes were considered and rejected. Splitting the output into two scores would build the oropharyngeal one on a single question, and a one-question score is not a score. Restricting the tool to oral cavity and dropping the HPV question would remove the fastest-growing part of the disease, which is the opposite of useful for the under-50s it most affects.
So the score stayed whole and the explanation changed. Each question is tagged with the disease it actually speaks to: tobacco, alcohol, betel, diet and sun exposure to the oral cavity; HPV to the oropharynx. Age, sex, family history, immune history, symptoms and dental attendance raise risk at both sites, so they are excluded from the comparison entirely rather than dragging every profile toward whichever site has more questions attached to it. A site is named only when its weight is at least twice the other's; below that the page says the profile spans both, because manufacturing a lean out of a one-point difference would be worse than saying nothing.
This is a change to what the tool explains, not to what it computes. The weights are untouched, and the open question, which belongs to the clinicians and not to this page, is whether the HPV weight should rise now that a reader can see the score is site-blended.
The Survival Figures This Site Quotes
Several pages quote five-year survival by stage, because the gap between finding oral cancer early and finding it late is the entire reason this tool exists. These are the current figures and what they do and do not mean.
| Stage at diagnosis | 5-year relative survival | Share of cases |
|---|---|---|
| Localized Has not spread beyond the original site. | 88.7% | 26% of cases |
| Regional Has reached nearby structures or lymph nodes. | 69.7% | 55% of cases |
| Distant Has spread to distant parts of the body. | 36.0% | 12% of cases |
| All stages Everyone diagnosed, including the 6% whose stage is unknown. | 69.9% | — |
Two things worth being precise about. These are SEER summary stages, which are not the same as AJCC Stage I through IV; this site used to present them as Stage I and Stage IV survival, which attributed one staging system's numbers to another, and that has been corrected. And these are relative survival rates, meaning survival compared with people of the same age and sex without the cancer, not a prediction for any individual.
Source: NCI SEER Cancer Stat Facts, Oral Cavity and Pharynx Cancer. Survival covers 2016 to 2022; case and death projections are for 2026; incidence rates cover 2019 to 2023.
What This Tool Cannot Tell You
A risk score is easy to over-read. These are the specific things a result from OralCheck does not mean, stated plainly so nobody has to infer them.
- It cannot tell you whether you have oral cancer. It never sees your mouth. It reads back the exposures you reported and nothing else.
- A low result is not a clean bill of health. Somewhere around a quarter of oral cancers occur in people with none of the conventional risk factors this screener asks about, so a low score paired with a symptom still means see a dentist.
- It cannot distinguish oral cavity from oropharyngeal cancer. These are epidemiologically distinct diseases with different dominant causes, and the single score blends them.
- It cannot account for what you did not report or did not know. Tobacco and alcohol use are consistently under-reported on self-administered questionnaires.
- It has not been validated against a clinical outcome dataset. Every weight is derived from published literature, but the instrument as a whole has never been calibrated against a prospective cohort, and it should not be described as validated.
- The score is not a probability. It is an ordinal stratification, so a score of 20 is not twice the risk of a score of 10 and does not correspond to a percent chance of developing anything.
- It cannot replace the exam. The actual screening for oral cancer is a short visual and tactile examination performed by a dentist or physician. The entire purpose of this tool is to make that appointment more likely to happen.
Known Limitations
- 1OralCheck hasn't been validated against a clinical outcome dataset. The weights are grounded in published literature, but they haven't been calibrated against a prospective cohort.
- 2The weights are pooled from studies conducted in different populations, and then applied to whoever happens to open the page. The tobacco figure comes from a meta-analysis of smoking cohorts, the betel figure from populations in South and Southeast Asia where use is common, the HPV figure from a US case-control study. Summing log-odds across them assumes those effect sizes transport to the person answering, and that assumption is untested. It is the single largest source of uncertainty in the score, larger than any individual weight being slightly wrong.
- 3The tier boundaries were set by judgment rather than derived from data. They are anchored to reference profiles and sanity-checked against the distribution of possible answers, but no outcome dataset informed them, so the tool cannot state a sensitivity or specificity for any tier.
- 4The screener still produces one score for two epidemiologically distinct diseases. Tobacco and alcohol dominate oral cavity cancer; HPV-16 dominates oropharyngeal cancer. The results page now says which of the two a person's own factors point at, but the score itself is not split, and the HPV weight remains a blended compromise: it is set at 5, from a conservative pooled odds ratio, where the published association for confirmed HPV-16 with oropharyngeal cancer specifically is roughly 15. That means an HPV-driven profile is understated by design, and the results page says so rather than hiding it.
- 5The combined immune and radiation question groups three exposures with genuinely different effect sizes into one weight. Someone on long-term immunosuppressants after a transplant and someone who had head and neck radiation decades ago score identically, which is a simplification made to keep the screener short.
- 6The screener relies on self-reported data. Tobacco and alcohol use are commonly under-reported.
- 7Dental visit frequency is included as a detection-delay proxy, not as a causal risk factor with a well-characterized OR. Its weight is based on clinical rationale rather than direct epidemiological derivation.
- 8Sex is captured as a single binary with an opt-out. The instrument has no way to represent someone whose exposure history doesn't match the population the underlying incidence data was drawn from.
References
- 1.Gandini S, et al. Tobacco smoking and cancer: a meta-analysis. Oral Oncology. 2008;44(7):617–638.View on PubMed / publisher →
- 2.Bagnardi V, et al. Alcohol consumption and site-specific cancer risk: a comprehensive dose–response meta-analysis. Annals of Oncology. 2015;26(1):39–55.View on PubMed / publisher →
- 3.Gillison ML, et al. Distinct risk factor profiles for human papillomavirus type 16–positive and human papillomavirus type 16–negative head and neck cancers. JAMA. 2008;168(3):294–305.View on PubMed / publisher →
- 4.International Agency for Research on Cancer. Betel-quid and Areca-nut Chewing and Some Areca-nut Derived Nitrosamines. IARC Monograph 85. Lyon, France: IARC; 2004.View on PubMed / publisher →
- 5.Napier SS, Speight PM. Natural history of potentially malignant oral lesions and conditions: an overview of the literature. Journal of Oral Pathology & Medicine. 2008;37(1):1–10.Leukoplakia 5–17% and erythroplakia 14–50% malignant transformation ratesView on PubMed / publisher →
- 6.Pavia M, et al. Evidence-based medicine on the relationship between diet and cancers of the oral cavity and pharynx. Oral Oncology. 2006;42(1):15–25.View on PubMed / publisher →
- 7.National Cancer Institute. SEER Cancer Statistics Review 1975–2021. Surveillance, Epidemiology, and End Results Program.View on PubMed / publisher →
- 8.Hashibe M, Brennan P, Chuang SC, et al. Interaction between tobacco and alcohol use and the risk of head and neck cancer: pooled analysis in the International Head and Neck Cancer Epidemiology Consortium. Cancer Epidemiology, Biomarkers & Prevention. 2009;18(2):541–550.Heavy combined users: OR 35.8× vs non-users of both, supporting supra-multiplicative interaction termView on PubMed / publisher →
- 9.Negri E, Boffetta P, Berthiller J, et al. Family history of cancer: pooled analysis in the International Head and Neck Cancer Epidemiology Consortium. International Journal of Cancer. 2009;124(2):394–401.OR ~2.0 for first-degree relative with head and neck cancerView on PubMed / publisher →
- 10.D'Souza G, Kreimer AR, Viscidi R, et al. Case-control study of human papillomavirus and oropharyngeal cancer. New England Journal of Medicine. 2007;356(19):1944–1956.HPV-16 seropositivity OR 32.2× for oropharyngeal cancer; behavioral exposure proxies used in screenerView on PubMed / publisher →
- 11.Perea-Milla López E, Minarro-Del Moral RM, Martinez-Garcia C, et al. Lifestyles, environmental and phenotypic factors associated with lip cancer: a case-control study in southern Spain. British Journal of Cancer. 2003;88(11):1702–1707.Sun exposure OR 2–3× for lower lip squamous cell carcinoma in outdoor vs indoor workersView on PubMed / publisher →
- 12.Engels EA, Pfeiffer RM, Fraumeni JF Jr, et al. Spectrum of cancer risk among US solid organ transplant recipients. JAMA. 2011;306(17):1891–1901.Transplant Cancer Match Study, 175,732 recipients; elevated incidence across many sites including the oral cavity and pharynx, and markedly elevated for lipView on PubMed / publisher →
- 13.Grulich AE, van Leeuwen MT, Falster MO, Vajdic CM. Incidence of cancers in people with HIV/AIDS compared with immunosuppressed transplant recipients: a meta-analysis. Lancet. 2007;370(9581):59–67.Establishes that the two immunosuppressed populations share a similar pattern of raised incidence, which is the basis for grouping them into one questionView on PubMed / publisher →
Review and Maintenance
The factor list, weights, thresholds, and sources on this page were last re-checked against current literature in August 2026, and revised following external review by a practising oral and maxillofacial pathologist. That review is what added the immune and radiation history question, added gum growth and recent tooth mobility to the symptom guidance, and prompted the caveat that a substantial share of oral cancers occur with no conventional risk factors at all.
This page is reviewed at least annually, and whenever a question is added, removed, or reweighted. If you spot an error in a citation or a weight, it is worth reporting: corrections are made directly and the review date is updated.