Risk Factor Gauntlets

Each IARC-established risk factor is tested as a primary exposure across 12 cancer types using three methods (BYM2, long-difference, IV/2SLS). The pesticide covariate is included in every model to test whether it survives.

Cross-Gauntlet Summary

Four risk factors were tested as primary exposures. Smoking passes convincingly (7/8 IARC-linked cancers significant). The other three show mixed results. Crucially, the pesticide–kidney and pesticide–colorectal associations survive as covariates across all four gauntlets.

Gauntlet IARC Score Best Hit Pest → Kidney Pest → Colorectal IV F-stat
Smoking 7/8 PASS Larynx RR=1.205* 1.025* — 16.3
Obesity 4/8 MIXED Myeloma RR=1.081* 1.025* 1.015* 6.4 (weak)
Alcohol 2/7 MIXED Oral Cavity RR=1.044* 1.025* 1.015* 3.6 (weak)
Inactivity 2/7 MIXED Liver RR=1.085* 1.025* 1.015* 25.9 (strong)
Rate ratio stability across gauntlets showing pesticide associations survive all four risk factor gauntlets
Figure 4. Pesticide rate ratio stability across risk factor gauntlets. The kidney and colorectal pesticide associations remain significant regardless of which risk factor is modeled as the primary exposure, demonstrating independence from established carcinogenic pathways.

Individual Gauntlet Results

7/8 IARC PASS

Smoking Gauntlet

Smoking is the strongest positive control: 7 of 8 IARC smoking-linked cancers show significant BYM2 associations. Lung (RR=1.192), larynx (RR=1.205), esophagus (RR=1.139), oral cavity (RR=1.133), bladder (RR=1.117), and kidney (RR=1.067) are all significant. This validates the BYM2 pipeline’s ability to detect known carcinogens.

BYM2 forest plot for smoking gauntlet across 12 cancer types
Smoking BYM2 Forest Plot. Rate ratios for smoking as primary exposure across 12 cancer types. Strong, expected hits for lung and larynx.
Smoking vs pesticide head-to-head comparison
Head-to-Head: Smoking vs Pesticide. Pesticide associations for kidney and colorectal survive when smoking is the primary exposure.
  • BYM2: 12 models converged. Lung RR=1.192*, Larynx RR=1.205*, Esophagus RR=1.139*
  • Long-difference: Uses change in smoking prevalence. All 7 outputs generated.
  • IV/2SLS: Instruments = convenience stores per 1k, fast food per 1k. Mean F=16.3.
  • Negative controls: Breast RR=0.974* (inverse), Prostate RR=0.946* (inverse)

Key Takeaway

Smoking passes the gauntlet convincingly, validating the analytical pipeline. The pesticide–kidney association (RR=1.025*) survives in the same models.

4/8 IARC MIXED

Obesity Gauntlet

Obesity shows a mixed IARC scorecard: 4 of 8 linked cancers are significant. The strongest signal is myeloma (RR=1.082*). Kidney shows both obesity (RR=1.036*) and pesticide (RR=1.026*) significant in the same model, suggesting independent risk pathways. Colorectal shows pesticide-only (obesity NS).

BYM2 forest plot for obesity gauntlet across 12 cancer types
Obesity BYM2 Forest Plot. Rate ratios for obesity as primary exposure. Myeloma is strongest; kidney significant but smaller than smoking effects.
Obesity vs pesticide head-to-head comparison
Head-to-Head: Obesity vs Pesticide. Kidney is the only cancer where both obesity and pesticide are significant.
  • BYM2: 12 models, max Rhat=1.09. Kidney: Obesity RR=1.036* + Pesticide RR=1.026*
  • Long-difference: Only kidney significant (β=0.142, p<0.0001)
  • IV/2SLS: Food environment instruments (grocery stores, food swamp ratio, low food access). Mean F=6.4 — WEAK instruments
  • Key: Colorectal is pesticide-specific (obesity NS in BYM2)

Key Takeaway

Kidney cancer has independent associations with both obesity and pesticide exposure. Colorectal cancer is associated with pesticide only—obesity does not explain the colorectal signal.

2/7 IARC MIXED

Alcohol Gauntlet

Alcohol shows only 2 of 7 IARC-linked cancers significant in BYM2. Only oral cavity (RR=1.044*) is clearly positive. Long-difference results are stronger (8/12 significant), but IV instruments are very weak (F=3.6). The pesticide–colorectal association survives (alcohol is NS for colorectal in BYM2).

BYM2 forest plot for alcohol gauntlet across 12 cancer types
Alcohol BYM2 Forest Plot. Only oral cavity shows clear significant association. Colorectal NS.
Alcohol vs pesticide head-to-head comparison
Head-to-Head: Alcohol vs Pesticide. Pesticide–colorectal survives; alcohol is not a confounding explanation for the colorectal signal.
  • BYM2: Only Oral Cavity sig (RR=1.044*). Colorectal NS (RR=1.003), Liver NS (RR=1.017)
  • Long-difference (7yr): 8/12 sig — Breast β=0.177***, Oral, Larynx, Liver, Colorectal, Kidney, Prostate, All-Site. Lung NS (good negative control)
  • IV/2SLS: Mean F=3.6 — VERY WEAK instruments. Results unreliable.
  • Negative control failure: Lung RR=1.039* (unexpected positive in BYM2)

Key Takeaway

Alcohol does not explain the pesticide–colorectal signal. Alcohol is null for colorectal in BYM2 while pesticide remains significant.

2/7 IARC MIXED

Physical Inactivity Gauntlet

Physical inactivity shows only 2 of 7 IARC-linked cancers significant: colorectal (RR=1.031*) and liver (RR=1.085*). IV instruments are the strongest of any gauntlet (F=25.9, using RUCC + population density). Inactivity does not outperform obesity as a metabolic driver. Pesticide associations survive.

BYM2 forest plot for inactivity gauntlet across 12 cancer types
Inactivity BYM2 Forest Plot. Only colorectal and liver significant. Kidney NS (RR=1.016).
Inactivity vs pesticide head-to-head comparison
Head-to-Head: Inactivity vs Pesticide. Pesticide associations for kidney and colorectal are independent of inactivity effects.
  • BYM2: Only Colorectal sig (RR=1.031*). Kidney NS (RR=1.016). Bladder inverse (RR=0.951*)
  • Long-difference (7yr only): Colorectal sig (β=0.131***), also Stomach, Liver, Lung
  • IV/2SLS: RUCC + pop_density instruments. Mean F=25.9 (STRONG). Colorectal sig (p=0.025)
  • Key: Inactivity does NOT outperform obesity as metabolic driver

Key Takeaway

Inactivity has strong IV instruments but weak IARC scorecard. Pesticide–kidney and pesticide–colorectal both survive independently.

Synthesis

Pesticide Associations Survive All Four Gauntlets

The pesticide–kidney (RR=1.025) and pesticide–colorectal (RR=1.015) associations remain significant as covariates regardless of which established risk factor is modeled as the primary exposure. This rules out the hypothesis that the pesticide signal is merely a proxy for smoking, obesity, alcohol, or physical inactivity. The associations appear to represent an independent environmental risk pathway.

Gauntlet result CSVs available on the Downloads page. Full methodology described on the Methods page.