9/11 and cancer: what 25 years of surveillance actually shows

Twenty-five years of cancer surveillance among the people exposed to the World Trade Center dust and smoke shows a real but narrow signal: rates of a handful of specific cancers — most consistently prostate and thyroid — run above what would be expected, while overall cancer incidence in the largest group of responders studied sits at or slightly below the general-population rate [s2] [s3]. The evidence for those specific excesses is moderately solid and has held up across separate cohorts, but it is complicated enough by increased screening, short latency and small case counts that "certified as a covered condition" — a compensation decision — is not the same thing as "proven to be caused by the exposure."

Why cancer is a covered condition

That certification came in 2012. Title I of the James Zadroga 9/11 Health and Compensation Act of 2010 created the World Trade Center Health Program, run by the National Institute for Occupational Safety and Health within the CDC, and a final rule published on 12 September 2012 added a broad list of cancers to the program's covered conditions, effective 12 October 2012 [s4]. The list is striking in its breadth: most solid-organ cancers, mesothelioma, blood and lymphoid cancers, melanoma and non-melanoma skin cancers, and blanket categories for childhood cancers and rare cancers [s4].

The reasoning behind that list is where the distinction between policy and proof lives. For most cancers, the program did not require proof that 9/11 exposure causes that specific cancer. Under the rule's dominant test, a cancer qualifies if a "9/11 agent" was documented in a peer-reviewed exposure assessment of responders or survivors, and if that agent is classed by the National Toxicology Program as a known or reasonably anticipated human carcinogen and by the International Agency for Research on Cancer as having sufficient or limited evidence of causing that cancer type [s4]. In plain terms: the dust and smoke are known to have contained carcinogens, so cancers those carcinogens can cause were made eligible for monitoring and treatment. That is a precautionary, administrative standard, deliberately set below the bar of epidemiological proof.

What the incidence studies found

The epidemiology is a separate question, and it has been tracked since the first years after the attacks. An early cohort of New York City firefighters, published in The Lancet in 2011, followed 9,853 men employed as firefighters on 1 January 1996 and compared cancer rates in the roughly seven years after 9/11 [s1]. Against the general US male population, the standardised incidence ratio — observed cancers divided by expected — was 1.10 (95% confidence interval 0.98 to 1.25), a modest excess whose interval crossed the no-effect line [s1]. Compared with firefighters who were not WTC-exposed, it was 1.32 (1.07 to 1.62) before any correction, falling to 1.19 (0.96 to 1.47) once the authors corrected for possible surveillance bias [s1]. Their own interpretation was cautious: the excess was modest, spread across organ types rather than concentrated, and observed over a span they called short for cancer [s1].

The next year, a study in JAMA looked at 55,778 people in the WTC Health Registry, including 21,850 rescue and recovery workers, followed through the end of 2008 [s2]. For all cancer sites combined in 2007–2008, the incidence ratio among rescue and recovery workers was 1.14 (0.99 to 1.30) — not statistically significant — and among enrollees not involved in rescue and recovery it was 0.92 (0.83 to 1.03), if anything below expected [s2]. But three sites stood out among the workers: prostate cancer at 1.43 (1.11 to 1.82), thyroid cancer at 2.02 (1.07 to 3.45) and multiple myeloma at 2.85 (1.15 to 5.88) [s2]. Those ratios rest on small numbers — 67 prostate, 13 thyroid and 7 myeloma cases — and the authors flagged that limitation and the problem of testing many sites at once [s2].

The most comprehensive read comes from a 2021 pooling of the three main responder cohorts, published in the Journal of the National Cancer Institute, covering 57,402 rescue and recovery workers followed through 2015 [s3]. Its headline number cuts against intuition: incidence for all first-time primary cancers combined was below expectation, at 0.96 (0.93 to 0.99) [s3]. Yet specific sites were significantly elevated — melanoma at 1.43 (1.24 to 1.64), prostate at 1.19 (1.11 to 1.26), thyroid at 1.81 (1.57 to 2.09) and tonsil cancer at 1.40 (1.00 to 1.91) [s3]. Crucially, the study found a dose-response signal: workers who arrived on 11 September itself had higher adjusted hazard ratios than those arriving after 17 September for prostate cancer (1.61, 1.33 to 1.95) and thyroid cancer (1.77, 1.11 to 2.81), with a statistically significant exposure-response trend for both [s3]. An exposure gradient of that kind is one of the stronger arguments that a signal is real rather than an artefact of watching this group more closely.

The caveats that matter

Several problems keep the honest verdict short of "9/11 causes cancer" as a blanket statement. The first is surveillance bias: responders receive free, intensive medical monitoring, which finds more cancers — especially slow-growing prostate and thyroid tumours that screening turns up — than would be diagnosed in a less-watched population. The firefighter study showed how much this matters, its estimate dropping from 1.32 to 1.19 once corrected for it [s1].

The second is the choice of yardstick. A 2025 analysis in Scientific Reports of 65,691 responders showed that the incidence ratio for a given cancer, and even whether it reaches significance, shifts with the reference population used — city, state or national [s5]. Prostate and thyroid cancer were consistently elevated across all three comparisons, but colon and lung cancer came out significantly lower than every reference group, and results for other sites depended on which population the responders were measured against [s5]. That both anchors the robust findings and warns against reading too much into the shakier ones.

The remaining caveats are latency and numbers. Many solid tumours take decades to appear, so 25 years is still early for some of them, and the rarer the cancer, the fewer the cases and the wider the confidence interval — the myeloma estimate spanning 1.15 to 5.88 is a reminder that a handful of extra cases can look dramatic without being certain [s2].

Where that leaves the question

The defensible summary is neither reassurance nor alarm. Overall cancer rates in this population are not markedly raised, and for the biggest responder cohort they are slightly below expected [s3]. A specific set of cancers — prostate, thyroid, melanoma among them — is genuinely elevated, the prostate and thyroid signals survive different reference groups, and the exposure gradient for both strengthens the causal case [s3] [s5]. The broad list of covered cancers reflects a policy choice to insure against plausible carcinogenic exposure, not a finding that every listed cancer has been shown to be caused by 9/11 [s4]. As the cohorts age into the latency window for long-latency tumours, the next decade of surveillance will do more to settle the open sites than the first quarter-century could.