How this instrument works
PLCOm2012 is a logistic-regression model published by Tammemagi, Katki, Hocking, and colleagues in the New England Journal of Medicine in 2013, built from the control arm of the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial. It predicts the probability that a current or former smoker will be diagnosed with lung cancer within the next six years, combining age, race/ethnicity, education, body mass index, personal history of COPD, personal history of any cancer, family history of lung cancer, current-versus-former smoking status, smoking intensity, smoking duration, and years since quitting into one logit equation. The model has been widely studied as an alternative to simple age/pack-year screening rules because it weighs several risk factors at once rather than a single threshold.
The smoking-intensity term deserves its own explanation because it is easy to mistake for pack-years and it is not. The published Table 2 footnote defines intensity as a reciprocal transform of cigarettes per day: -1.822606 multiplied by [(10 divided by cigarettes-per-day) minus 0.4021541613]. That shape means each additional cigarette per day adds progressively less to the score at high intensities than at low ones, unlike a simple linear term. Duration, by contrast, IS a plain linear term in this model — 0.0317321 times (years smoked minus 27) — so the two smoking variables behave quite differently even though both describe smoking history.
Race and ethnicity enter as a categorical coefficient rather than a continuous adjustment: this calculator implements Table 2 of the original paper exactly as published — White and American Indian/Alaskan Native are both anchored at 0, Black adds +0.3944778, Hispanic subtracts 0.7434744, Asian subtracts 0.466585, and Native Hawaiian/Pacific Islander adds +1.027152. It is worth being direct about a documented issue with that table: lead author Tammemagi's own 2018 follow-up paper states that the original Table 2 had the labels for 'Native Hawaiian or Pacific Islander' and 'American Indian or Alaskan Native' reversed, and that the corrected American Indian/Alaskan Native odds ratio is close to the value this table currently attaches to Native Hawaiian/Pacific Islander. Because this calculator reproduces the coefficients exactly as published in the cited NEJM reference, that labeling question is carried forward unresolved rather than silently reassigned here.
PLCOm2012 was developed and validated in current and former smokers aged 55 to 74, and it does not apply to people who have never smoked — the source paper's own study population excluded never-smokers entirely. It is also a risk-prediction tool, not a diagnostic test, and it is a distinct instrument from the U.S. Preventive Services Task Force's age-and-pack-year lung cancer screening eligibility criteria; the two are often discussed together in screening-program design but are not the same calculation, and this calculator does not reproduce the USPSTF rule.
- Enter age (age), race/ethnicity (race), and education (education) — the formula is centered on age 62 and 'some college' education, so values away from those shift the score.
- Enter height (height) in meters and weight (weight) in kilograms; the calculator derives body mass index (bmi) automatically from those two fields.
- Set history of COPD (copd), personal history of cancer (cancerHist), and family history of lung cancer (familyHist) to Yes or No.
- Choose smoking status (current) as current or former smoker, then enter average cigarettes per day (cpd) and total years smoked (duration).
- If a former smoker, enter years since quitting (quitYears); leave it at 0 for a current smoker.
- Review the smoking-intensity contribution (intensityTerm), the logistic linear predictor (logit), and the predicted 6-year lung cancer risk (riskPercent).
Worked example — the source paper's own reference case
A 62-year-old White man with some college education, height 1.70 m and weight 78.03 kg (BMI 27.0), no history of COPD, no personal cancer history, and no family history of lung cancer quit smoking 10 years ago after smoking 80 cigarettes a day for 27 years. Every one of those held-constant values — age, race, education, BMI, COPD, personal history, family history, smoking status, duration, and quit-years — matches the fixed covariates in the original paper's own Figure 1; cigarettes-per-day is not one of Figure 1's fixed values, since intensity is the variable Figure 1 plots risk against, and 80 is simply one point read off that curve. Age, education, BMI, duration, and quitYears all sit exactly at the formula's centering values (62, some college, BMI 27, 27 years, 10 years), so each of those five terms contributes exactly zero to the logit, and the race term is also zero because White is the reference category.
That leaves only the constant and the smoking-intensity term active. At 80 cigarettes per day, the reciprocal transform gives intensityTerm = -1.822606 x [(10/80) - 0.4021541613] = -1.822606 x (0.125 - 0.4021541613) = -1.822606 x (-0.2771541613), which works out to about 0.5051. Adding that to the base constant of -4.532506 gives a logit of roughly -4.0274.
Converting that logit to a probability: exp(-4.0274) is about 0.01782, so risk% = 0.01782 / (1 + 0.01782) x 100 works out to about 1.75%. This calculator returns 1.7509% for these exact inputs, matching the hand-worked math above digit for digit, and it agrees with the documented example call in the resplab/PLCOm2012 R package (cigarettes-per-day = 80, which returns prob = 0.01750922, i.e. 1.7509%) — a useful check that the reciprocal intensity transform, and not a linear pack-years shortcut, is what actually gets applied.
Questions
Does this calculator apply to people who have never smoked?
No. PLCOm2012 was developed and validated only in current and former smokers aged 55 to 74 — the source study's population excluded never-smokers entirely, so the model has no coefficients estimated for that group and cannot honestly extrapolate to it. If you have never smoked, this particular tool is not built for you; a different, separately published model would be needed to estimate lung cancer risk in never-smokers, and that is outside what this calculator covers.
Why is smoking intensity a reciprocal transform instead of pack-years?
Because that is exactly how the original study specified it. Table 2 of Tammemagi et al. 2013 defines the smoking-intensity term as -1.822606 multiplied by [(10 divided by cigarettes per day) minus 0.4021541613], a nonlinear reciprocal shape rather than a simple linear cigarettes-per-day or pack-years term. This means the effect of one additional cigarette per day is larger at low daily intensities than at high ones. Duration (years smoked) is a separate, ordinary linear term in the same model, so the two smoking variables are combined quite differently even though both describe smoking history.
Why does the calculator treat American Indian/Alaskan Native the same as White?
This calculator reproduces Table 2 of the 2013 paper exactly, where American Indian/Alaskan Native and White both carry a coefficient of 0. That is not a settled small-sample decision, though — it is a known labeling issue. Tammemagi's 2018 follow-up paper states Table 2 had the labels for 'Native Hawaiian/Pacific Islander' and 'American Indian/Alaskan Native' reversed, reporting a corrected Am.Indian/Alaskan Native odds ratio (about 2.79) close to what this table assigns Native Hawaiian/Pacific Islander (+1.027152). This calculator implements the coefficients as the NEJM reference published them, carrying that unresolved labeling question forward rather than silently reassigning values.
Is this the same as the USPSTF lung cancer screening eligibility criteria?
No, and the two get confused often enough that it is worth stating plainly. The U.S. Preventive Services Task Force's screening eligibility recommendation uses a simple age-and-pack-year rule, not the PLCOm2012 logistic model this calculator implements. PLCOm2012 is a multivariable risk-prediction score studied as one possible alternative or complement to that simpler eligibility rule, and the two produce different numbers by different methods. This calculator computes PLCOm2012 only; it is not a USPSTF eligibility checker.
Can this calculator diagnose lung cancer or tell me I need a scan?
No. It performs the published PLCOm2012 arithmetic on the values entered and returns a population-derived probability, nothing more — it cannot see a chest CT, a biopsy, or any finding specific to you. A calculated percentage is meant to inform a conversation with a clinician about screening or risk-reduction, including quitting smoking, not to stand in for medical evaluation or a screening referral decision on its own.
What does it mean that age, education, BMI, duration, and quit-years are 'centered' in the formula?
Each of those five variables is written as (your value minus a fixed reference point) — age minus 62, education minus 4 ('some college'), BMI minus 27, duration minus 27 years, and years-since-quitting minus 10. When your value equals that reference point, the subtraction is zero and the term contributes nothing to the logit that step. It is a modeling convenience from the original regression, not a claim that those specific numbers are 'normal' or risk-free — someone at exactly BMI 27 still has every other term in the equation counted normally.
Where do the coefficients behind this calculator come from?
Every coefficient — the constant, the age/education/BMI/duration/quit-year slopes, the race/ethnicity categories, the COPD/cancer-history/family-history/current-smoker terms, and the smoking-intensity transform — comes directly from Table 2 of Tammemagi MC, Katki HA, Hocking WG, et al., 'Selection criteria for lung-cancer screening,' New England Journal of Medicine, 2013;368(8):728-736. The worked example on this page matches every fixed covariate in the paper's own Figure 1 except cigarettes-per-day, which Figure 1 plots as a variable rather than fixing; the resulting 1.7509% figure also matches the documented example output of the resplab/PLCOm2012 R package for the same inputs.
What are the six education levels used in this calculator?
Education is entered on a six-level ordinal scale drawn from the original study's questionnaire: 1 = less than high-school graduate, 2 = high-school graduate, 3 = some training after high school, 4 = some college, 5 = college graduate, and 6 = postgraduate or professional degree. The formula centers on level 4 ('some college'), subtracting 4 from whatever level you select and multiplying by -0.0812744, so higher levels lower the logit slightly and lower levels raise it. That is simply the coefficient the source paper estimated for this variable alongside ten other predictors, not a claim that any one education level is inherently higher or lower risk on its own.
References
- Tammemagi MC, et al. — Selection criteria for lung-cancer screening, N Engl J Med. 2013;368(8):728-736
- US Preventive Services Task Force — Screening for Lung Cancer: Recommendation Statement, JAMA. 2021;325(10):962-970
Read this first: This instrument computes a screening figure from population formulas — it is not a diagnosis, and it cannot see the whole picture a clinician can. Use it to inform a conversation, not to replace one.