When errors become etched in stone: The U.S. Navy's deeply flawed,never retracted World War II asbestos epidemiology study

Introduction

After Germany invaded Poland in September 1939, the United States started preparing for war. These efforts were greatly accelerated by the attack at Pearl Harbor by the Japanese on 7 December 1941. The country embarked on a massive shipbuilding program, with the 11 Naval shipyards and selected private shipyards turning out warships, and other private shipyards producing merchant vessels including the famous Victory and Liberty ships. This was a huge effort, employing as many as 4.5 million men and women over the duration of the war. ^{1 p 109}

Given the huge scale, heavy industrial work, and the inexperienced workforce, it is not surprising that there were numerous occupational injuries and illnesses. This was the case across all “war industries” but was particularly acute in shipbuilding. At the Navy and Maritime Commission contract shipyards, as many as 350 people were killed with 86,000 lost time accidents in some years. ² Workers were exposed to a wide range of toxic gases, vapors, fumes and dusts, including asbestos. The Navy was aware of this, and after fretting over the potential hazard posed by asbestos dust for several years, it commissioned a study in 1945 to evaluate the issue. The resulting cross-sectional epidemiological study “A Health Survey of Pipe Covering Operations in Constructing Naval Vessels” was published in January 1946. ³ It has been so often referenced, particularly by the legal community, that it is widely known as the “Fleischer/Drinker” report, named for the first and last authors, U.S. Naval Reserve physician Walter Fleischer and Harvard industrial hygiene professor and U.S. Maritime Commission consultant Philip Drinker.

This would prove to be one of the most consequential published papers in the history of industrial hygiene and occupational medicine. Its outsized significance derives from the fact that it concluded that work with asbestos insulation at shipyards was “not a dangerous occupation.” Of course, over the following 40 years, this would prove to be a catastrophically inaccurate judgment, as tens of thousands of ex-shipyard workers developed asbestosis, as well as lung cancer and mesothelioma. ⁴

Review of the 1946 article reveals that this erroneous conclusion was in large part based on the use of a non-standard and unvalidated air sampling and analysis method and a one-off exposure limit selected to look like the prevailing United States Public Health Service (USPHS) recommended asbestos exposure limit. Further, the vast majority of the workers studied had not been exposed to asbestos long enough to develop asbestosis, and this limitation was either unrecognized or intentionally ignored. Despite these problems, this deeply flawed paper has never been retracted and is still periodically referred to in a positive light by asbestos litigation defense attorneys and their expert witnesses.

Epidemiological studies of asbestos exposure

Evidence that asbestos inhalation could cause lung disease first appeared in the United Kingdom at the very end of the nineteenth century. ^{5 p 11} By 1925, the danger was confirmed and the name for the disease was selected: “asbestosis.” ⁶ The first published epidemiological study of the health impacts to workers exposed to asbestos dust was performed in England by E.R.A. Merewether and C. W. Price. ⁷ These authors chose to study asbestos textile workers, as their dust exposure was almost exclusively to asbestos. This cohort included 363 workers, of which 95 were diagnosed with pulmonary fibrosis (26%). An additional 21 workers were found to have “precursor signs of the disease.” In the subfraction of 133 employees X-rayed, 62 (47%) presented radiographic signs of a diffuse fibrosis and another 25 had changes that were not definitely fibrosis, presumably changes such as pleural plaques. Thus, 65 percent of the workers X-rayed showed signs of damage caused by asbestos inhalation.

Since this was a selection of workers with more than the average years of exposure to asbestos, the results could not be generalized to the whole worker population, which was dominated by workers with < 5 years of asbestos exposure. After correcting for this, the authors concluded “…it appears that the general incidence rate of fibrosis of the lungs amongst those employed in these sections of the industry is rather less than 1 in 8, or, excluding those employed under 5 years, rather less than 1 in 3.”

The workers with the lowest exposure and the lowest rate of asbestosis were the “spinners.” The spinners took longer to develop asbestosis—18.7 years on average—versus about 13 years for workers employed in dustier tasks. The take-home messages from this study were several: (1) asbestos inhalation causes lung scarring, (2) the higher the dust level, the quicker workers became ill, (3) the longer the exposure, the higher the incidence of illness, (4) asbestosis could have a very long latency period—at the lowest exposure level, it required almost 20 years on average for workers to develop disease.

The first U.S. epidemiological study to relate exposure to asbestos dust with the development of asbestosis was initiated in 1929 but only published in 1935 after it had been edited by the asbestos industry sponsors. ⁸ The Metropolitan Life Insurance Company worked with two asbestos companies they insured, Johns-Manville and Raybestos-Manhattan, which were “…desirous of ascertaining whether asbestos dust was an occupational hazard in their establishments and, if so, what was the nature of this hazard and what should be done to prevent or control it.” Metropolitan's medical directors Anthony Lanza and William McConnell and industrial hygienist J. William Fehnel measured airborne dust and examined workers for signs and symptoms of asbestosis in several unidentified factories.

The primary tool they used to collect airborne dust was the Bureau of Mines/USPHS standard Greenburg-Smith “impinger,” with analysis of collected dust by counting under a 100 x microscope. This tool and analytical procedure were developed in the early-1920s ⁹ and rapidly became the standard in the United States for measuring airborne dust. ¹⁰ Dust concentrations, expressed in units of “millions of particles per cubic foot of air” (mppcf), were found to vary widely among asbestos plants, ranging from 0.5 to 4 mppcf in one plant and 43 to 82 mppcf in two others. The investigators found a direct relationship between how long the workers had been in the asbestos industry and the prevalence of disease. In the longest exposed group (>15 years) 87 percent of the workers had clinical and X-ray evidence of asbestosis. In the shortest exposed group (<5 years), 43 percent were found to have some degree of asbestosis. The authors were unable to suggest a “safe” exposure limit for asbestos workers, since the disease was so widespread among the workers.

The next U.S. asbestos epidemiological study evaluated four plants in Pennsylvania in 1934. Warren Fulton et al. from the Industrial Hygiene Section of the Bureau of Industrial Standards of the State of Pennsylvania led the effort. ¹¹ These plants produced asbestos-containing textiles, brake pads, cements, insulating tape, yarn, rope, shingles, and block insulation. Airborne dust levels measured with the standard USPHS/Bureau of Mines impinger method were exceedingly high, particularly in the locations where the asbestos was prepared for further uses—opening bags of asbestos, mixing, and screening asbestos. In these locations, concentrations ran from 10.7 to 123 mppcf (mean 36 mppcf). Grinding asbestos clutch facings generated a dense cloud of dust, 434-­­515 mppcf. Some of the other operations were less dusty, with average dust levels of 7–14 mppcf. The physicians on the team diagnosed asbestosis in one-fourth of the plant's employees. Again, the researchers were unable to recommend a safe exposure limit, as 8 percent of workers exposed to dust at the lowest concentration measured (below 5 mppcf) were diagnosed with asbestosis.

The third published American asbestos dust epidemiology study resulted in the publication of a “safe” exposure limit for asbestos-containing dust. The USPHS Division of Industrial Hygiene, led by physician Waldemar C. Dreessen, and staffed by J. M. DellaValle (industrial hygienist), Thomas I. Edwards (technical editor), J. W. Miller (pathologist) and Royd R. Sayers (physician/industrial hygienist), evaluated asbestos exposure and health status of workers in three North Carolina asbestos textile plants. They published their findings in 1938 in a government document titled “A Study of Asbestosis in the Asbestos Textile Industry.” ¹² The investigators collected and analyzed 242 air samples, again primarily using the USPHS/Bureau of Mines standard Greenburg-Smith impinger/low power microscope collection and counting methods. Dust levels in the plants were similar to those found in the Metropolitan Life study, but somewhat lower than found in the Pennsylvania study, ranging from 1.2 to 74 mppcf.

445 current workers received physical examinations, including a full occupational history; 27.5 percent were diagnosed with asbestosis. In their analysis, the authors considered both the concentration of dust to which employees were exposed and their duration of exposure, the product expressed in integrated units of “mppcf-years.” “Asbestosis 4,” the most severe version of the disease, was found in 7 employees, all of whom had >100 mppcf-years of asbestos exposure. “Asbestosis 3” was found in 16 workers all of whom had >50 mppcf-years of exposure. There were 26 “Asbestosis 2” cases, one of whom had a lifetime exposure of only 25–49 mppcf-years. Asbestosis 1, the mildest form of the disease, was diagnosed in 25 workers, 5 of whom had less than 25 mppcf-years of lifetime exposure, and 2 of which had 25–49 mppcf-years exposure. They noted that 70 cases of asbestosis occurred in jobs where dust exceeded 5 mppcf, none where levels were below 2.5 mppcf and only three “doubtful” cases among workers exposed between 2.5 and 5 mppcf. They proposed that “5 million particles per cubic foot may be regarded tentatively as the threshold value for asbestos dust exposure until better data are available.”

There were problems with this study. 150 out of the 445 (34%) plant employees had been fired 15 months before, part of the industry-wide purge of suspected asbestotics in North Carolina. As a result, the group of active employees examined had an unusually low average duration of dust exposure: “There was an abnormally large percentage of workers with less than 5 years of employment in the asbestos textile industry and an abnormally small percentage of persons who had worked 10 years or more in this industry.” The average age of the employees was only 32, less than in 14 other similar factory studies that the USPHS had completed. Only 47 employees had worked in the asbestos industry for 10–15 years and only 16 had worked 15–20 years. Critically, only five workers in the lowest exposure group (<5 mppcf) had more than 10 years of exposure. This was a very small number of subjects to hang their hat on when setting the exposure limit given the long latency period for development of asbestosis at lower exposure levels.

Despite these and other substantial limitations, the proposed 5 mppcf “threshold value” stuck, and remained unchallenged as the “safe” exposure limit for asbestos dust in the United States for 30 years. The American Conference of Governmental Industrial Hygienists (ACGIH) adopted this value in 1946 as a Maximum Allowable Concentration. ^{13 p 11} The ACGIH did not reconsider the 5 mppcf value until 1968 ¹⁴ and only changed it in 1974. Nationally, this value was only abandoned when OSHA came into existence in May of 1971, as it adopted the draft ACGIH standard of 2 mppcf or 12 fibers per cubic centimeter of air and then very quickly reduced the limit to five fibers per cubic centimeter of air, while completely abandoning the limit as measured in mppcf.

As in the Metropolitan and Pennsylvania studies, this 5 mppcf value was based on counting all visible particles under the microscope, not just fibers, in accordance with the standard USPHS/Bureau of Mines procedure. It was not the case that the USPHS team had found only asbestos fibers in air samples they collected in the textile mills, quite the opposite. In some instances, as much as 99 percent of the dust in the air they measured was nonfibrous—lumps, chunks, shards, fragments, some of it certainly asbestos, much of it arising from serpentine rock contaminants in the asbestos, the cotton blended into the fabric, or background sources. One to 8 percent of the dust was fibrous in air samples of crushing, twisting, carding, and picking operations, while 12–26 percent of particles were fibrous when measured in weaving operations. Overall, only 9 percent of the dust measured was fibrous. These values were derived from parallel samples taken with the Owens Dust Jet device and counted under high magnification. Far fewer fibers were seen in the impinger samples used to establish the exposure limit. Nevertheless, the authors confirmed that “Dust counts included both fibrous and nonfibrous particles, irrespective of size.” ^{12 p 23} This study represented the state of the art until the 1960s.

The U.S. Navy and asbestos

At the dawn of WWII, the risk of asbestosis associated with inhalation of dust in asbestos processing factories had been well established by the aforementioned studies and others in the United Kingdom, Germany and France. Several documents show that the U.S. Navy was aware that asbestos insulating work performed at its shipyards could be hazardous and it had set certain expectations to protect its employees.^15,16 At that time, the U.S. Navy may have been the only organization in the United States taking any steps to protect its “pipecoverers and insulators” from asbestos dust exposure, at least on paper.

In early 1942, the Navy and U.S. Maritime Commission, concerned about the high lost time accident rate at the private shipyards operating under contract to the Commission, set up a joint project to carry out safety and health surveys at these contract shipyards. Daniel Ring, Director of the Division of Shipyard Labor Relations at the Maritime Commission, led the project. Two health and safety specialists were appointed to oversee the survey—John M. Roche of the Safety Engineering Staff of the American Safety Council, and Dr. Philip Drinker, professor and head of the industrial hygiene program at Harvard. ⁴ On 7 July 1942, Drinker was officially appointed “Chief Health Consultant to the U.S. Maritime Commission.” Aided by a number of subordinates, the Drinker-Roche teams visited 20 private shipyards over the next few months.

No air samples for asbestos dust were obtained during these cursory visits—these were walkthroughs, not industrial hygiene surveys. Along the way, inspectors periodically noted concerns with exposure to asbestos. It seemed to be a relatively minor problem overall. Further, there were few workers involved—pipecoverers and insulators only accounted for 0.2 percent of the national shipyard workforce during the war (0.1% on the west coast). In contrast, 15 percent of shipyard workers were involved with welding, which presented acute exposure hazards. ^{17 p 24} Still, the hazard posed by asbestos dust inhalation was on the inspection team's radar, and they found problems.

The team visited the Oregon Shipbuilding Corporation at Portland, Oregon in September 1942. The inspectors noted significant concerns in the pipe covering and insulating shop. Their report stated: “Recognizing the fact that asbestos dust can cause permanent lung damage to persons exposed to an excessive amount over a period of time, conditions in this shop can only be regarded as alarming, and immediate steps should be taken to see that the shop is cleaned. Demonstrating the urgency of the matter is a discussion we had with the asbestos wool cutter in which he stated that it is his intention to resign his job as his chest feels tight in the morning, and he is coughing up blood occasionally.” ¹⁸ Also in September 1942, the inspection team visited the Bath Iron Works Shipyard in Bath, Maine. Here the team found troubling dust exposures in the insulating shop, reporting: “The conditions in this shop present a very real asbestosis hazard and immediate steps should be taken to segregate the most dusty processes into a well-ventilated area. Local exhaust systems of proper design should be installed…” ¹⁹

Overall, the inspectors on this first round of evaluations found a lot of poorly controlled hazards, including asbestos work. Drinker commented “…asbestosis is caused in connection with the handling of asbestos…we rather expect it to occur in shipyards, because we have seen asbestos being handled in installation work with little or no precautions.” ^{20 p 22} In this era, there were no federal safety regulations to guide the operators of the shipyards, only scattered and variable state regulations. In response, Drinker and Roche drafted a set of proposed safety rules titled “Minimum requirements for safety and industrial health in contract shipyards.” The final details were hammered out in a 2-day meeting in Chicago on 7–8 December 1942. The resulting set of rules was approved by the U.S. Navy on 20 January 1943 and by the Maritime Commission on 9 February 1943. ²¹

The Minimum Requirements outlined “Eight common types of disease and the methods for their preventions…” The list consisted of “Flashburns and foreign bodies in the eye, Lead Poisoning, Solvent Vapors, Zinc fume fever, Fiberglas [sic], Asbestosis, Silicosis and Dermatitis.” The asbestos rules required segregation of dusty work, effective ventilation in the shop, or in the alternative, the “wearing of special respirators.” Periodic medical examinations were specified. Ventilation shipboard was not required for asbestos work as it was for welding.

After promulgation of the Minimum Requirements, the Maritime Commission continued to send inspection teams to private contract shipyards to check compliance. Pushed by Drinker, in December 1944, a medical survey was conducted of pipecoverers and insulators at the Bath Iron Works Shipyard. The effort was led by physician Waldemar Dreessen of the USPHS, the same physician who had led the North Carolina study that produced the 5 mppcf asbestos dust standard. He was aided by W. E. Fleischer, a Naval reserve physician. ²² There were 120 such workers at the Bath Iron Works Shipyard, 38 of the long-term employees were examined. 12 of these workers, with only 2–9 years of exposure, showed signs and symptoms of lung damage, with six having frank asbestosis. Their exposures must have been very intense to develop asbestosis so quickly.

The physicians visited the insulating shop and noted that it appeared to be very dusty. No asbestos dust air samples were collected but their report referred to an upcoming industrial hygiene survey to be conducted by the shipyard's insurance carrier. A month later the air sampling data were reported by the American Mutual Liability Insurance company of Boston. Dust counts were high, ranging from 34 mppcf to 52 mppcf. The industrial hygienist, E. Ward Thompson, noted that in most of the samples examined, other than those taken while “cutting asbestos,” the majority of the dust was nonfibrous. However, following the USPHS/Bureau of Mines standard dust quantitation protocol, he apparently counted all of it. ²³ In response, Drinker sounded the alarm, noting that dust counts were “…very much higher than anyone would recommend.” ²⁴ He pitched a bigger study to inspect “…at least two Navy Yards and two Navy Contract yards.” ²⁵

The deeply flawed Fleischer/Drinker study

Drinker was given the go-ahead to start his epidemiological study of asbestos exposure and asbestosis at four shipyards. His team included Navy physician Walter E. Fleischer and Naval reserve industrial hygienists Fredrick J. Viles Jr. and Robert L. Gade. The investigators measured dust exposure and X-rayed and examined 1047 pipecoverers and insulators. They were permitted to publish their findings in the Journal of Industrial Hygiene and Toxicology in 1946: “A Health Survey of Pipe Covering Operations in Constructing Naval Vessels.” ³

The four shipyards examined in this study were using enormous quantities of asbestos-containing materials, particularly the two Navy-operated facilities. Every month the two navy yard workers installed: 108,000 square feet of Amosite insulation (two football fields), 155,000 linear feet of asbestos-containing pipe insulation (29 miles), 184,000 square feet of asbestos cloth (3 + football fields), and 50,000 pounds of asbestos cement. The scale of work can also be realized by considering the large number of pipecoverers and insulators employed. “Navy Yard A” employed 84 workers in the shop and 467 on board ships. In “Navy Yard B,” there were 50 employees working in the shop and 700 working on ships. Contract yards C and D were smaller, but still employed 168 and 174 pipecoverers and insulators, respectively.

Ongoing consequences

For at least 45 years, asbestos industry defense attorneys have wielded Fleischer/Drinker's anodyne final conclusion to argue that asbestos insulation manufacturers could not possibly have known that asbestos insulation work (as contrasted with asbestos factory work) was dangerous before the publications by Marr, Selikoff, Harries, Stumphius, and others in the 1960s. Fleischer/Drinker is still being misused in court cases—as recently as 2018 it was mischaracterized by an “expert witness” as a “…well designed study” in the docket of a case argued before the U.S. Supreme Court. The same expert was convinced that the 5 mppcf value used by Fleischer/Drinker was equivalent to the value set by the USPHS in 1938. It was not a well-designed study, and the exposure limits were not equivalent. ⁴³

Retraction issues

The integrity of the published biomedical and occupational health literature is critical to accurate and efficient scientific progress and implementation of effective occupational health programs. Retraction is one mechanism by which a defective paper published in a science journal is disavowed so that its results and conclusions are (it is hoped) no longer relied upon. The purpose of retraction is to correct the literature and ensure its integrity, not to punish the authors. ⁴⁴ Retraction of a publication is generally considered when it is found to be unreliable, the data fabricated or manipulated, unethical, duplicate, plagiarized, to infringe copyright, or subject to compromised peer review or undisclosed competing interest(s). Generally, editors and publishers have the final say on retracting a disputed article. Retracted articles are not typically removed from the published literature but are marked as retracted.

With the growth of predatory journals, hijacked journals, and scientific paper mills producing low-quality, stolen, or fake science papers for a fee, the rate of retractions from the science literature has ramped up dramatically, due to greater awareness, greater accountability, and better tools of detection. ⁴⁵ According to Ivan Oransky, founder of the retraction tracking organization Retraction Watch, in 2002, science journals retracted about 40 papers worldwide, or roughly 0.002 percent of all published articles. In 2014, that figure was closer to 0.02 percent, and in 2024, it reached a remarkable 0.2 percent, or 1 in 500 papers. ⁴⁶ There were over 10,000 retractions of articles from global science journals published in 2023 alone, so far the standing record. ⁴⁷ The recent overall increase in retractions is driven, in part, by an increase in retractions by researchers and journals located in non-western countries. Several of the first identified “paper mills” were based in China, Russia, and Iran. ⁴⁷ The spike in retractions in 2023 was extraordinary, ultimately reaching over 13,000 articles, inflated by a purge of thousands of defective articles published by the Egyptian company Hindawi, which had been acquired 2 years prior by the publishing giant Wiley. ⁴⁸

Retraction is often challenging and opaque and varies widely

Convincing publishers, editors, and/or authors to retract defective articles is often a difficult, drawn out process, even without the added complications imposed by the passage of several decades since publication. In some cases, an investigation by the author's employer is required, or data may have to be reanalyzed, it may even be necessary to repeat an experiment. One group of research reliability investigators submitted information to 77 journals representing 29 publishers, over a timeframe of March 2012 through February, 2020 supporting the retraction of 292 articles authored by one group of investigators. By October of 2020, only 115 publications (39%) had been acted upon by correction (3), publication of expressions of concern (18), or retraction (94). The median time from submission of concerns to the first journal correction was 22.1 months, only 25% acted within a 1-year period. ⁴⁹ Success in getting articles retracted varied widely among publishers with 10 or more challenged papers, ranging from 0% to 42%. These researchers concluded: “The current findings reinforce concerns about the inefficiency, inaccuracy, incompleteness and opacity of systems for ensuring publication integrity. While science espouses rigor, consistency and transparency, the mechanisms to provide quality control of scientific publications currently fail to achieve these goals.”

In a 2025 paper, this group of integrity researchers expanded on these observations: “Existing evidence reports that slow, incomplete, and opaque responses from publishers to integrity concerns are common, in both clinical and preclinical disciplines…Publishers’ responses to notification of concerns about the integrity of publications in their journals are markedly inconsistent, both in their timing and the nature of their editorial decisions. The reasons for these inconsistencies are unknown…” ⁵⁰

Lack of robust guidance for retracting very old articles

The refusal to retract the Fleischer/Drinker articles is driven by the age of the paper, published 80 years ago. This is not unique, in other cases publishers have apparently refused to retract old articles just because they were old. ⁵¹ When an editor or publisher makes this decision, it establishes a de facto statute of limitations on correcting old work, much as graffiti on public lands may become protected as an antiquity under the National Historic Preservation Act of 1966 if it survives more than 50 years. It is generally assumed that successor publishers are responsible for the management of the content they inherited from prior publishers, but it is not always clear what their responsibilities are for purging these archives of defective work. ⁵² Major scientific publishers (Taylor and Francis, Sage, Wiley, Nature Portfolio, Elsevier, Springer Nature, etc.) have established retraction policies, many of which are available online. In all cases, the publishers commit to correct work that was seriously technically flawed when published, not merely made obsolete by advances in science. However, none of these publishers’ policies explicitly address the retraction of old, legacy articles inherited from prior publishers, or cases where the authors are deceased, or both, as in the Fleischer/Drinker article.

Criteria and processes for retracting defective articles have been developed by a number of bodies including the International Committee of Medical Journal Editors, Council of Science Editors, and the World Association of Medical Editors, but none of these specifically address the issues that arise in retracting older articles.^53–55 A 20 person working group formed at a 2018 conference titled “Keeping the Pool Clean–Prevention and Management of Misconduct Related Retractions” likewise published a retraction guidance document that does not clearly address the questions that arise when considering retraction of very old articles. ⁵⁶

By far the most influential publishing industry guidance pertaining to the retraction of defective papers is provided by the Committee on Publication Ethics (COPE), a voluntary body with over 10,000 members that strives to provide “…leadership and an unbiased expert voice in publication ethics.” COPE has published a range of guidance documents to assist editors and publishers in fulfilling their duties. As of February, 2026, this includes a base retraction guidance document and 37 supporting “retraction flowcharts,” all of which are essentially silent on the issue of retracting very old articles in cases such as Fleischer/Drinker.^57,58 However, in August of 2024 COPE published a brief position statement online titled “Handling Retractions and Expressions of Concern for Old Articles.” Therein COPE advises ⁵⁹ : “Journals should always consider concerns about the integrity of research presented in articles that arise at any time and should not avoid issuing a retraction or an expression of concern simply on the basis of the age of the article. However, in some cases, it might be difficult to thoroughly investigate a past concern because the data are no longer available, are of poor quality, or because editors feel uncomfortable in applying current expectations and practices to articles published in the distant past. In these situations, editors need to decide which concerns are worth investigating and, in recognition of valid limitations, decide on a case-by-case basis which concerns to pursue.”

This fuzzy statement leaves editors and publishers to struggle with case-by-case decisions without the benefit of clear policy or a flowchart to use in assessing the feasibility and advisability of reviewing and potentially retracting a legacy article. Further, it does not directly address the specific issues cited by the editor of Archives of Occupational and Environmental Health in refusing to publish a retraction of the Fleischer/Drinker piece–the switch in publishers and death of the authors. When queried regarding this apparent gap in guidance, a representative from COPE offered that the organization did not make any recommendation regarding a statute of limitations for retracting defective articles, but also indicated that COPE is aware of the gap in its guidance on retracting old papers and is working on a revised policy, without suggesting when this policy will be made available. ⁶⁰

Like other science publishers, Taylor and Francis states that it follows the COPE guidelines in managing corrections and retractions. ⁶¹ When queried specifically about its corporate policy on retracting very old articles, a representative of the publisher essentially repeated the COPE guidance and provided no additional written procedures for handling these cases. ⁶²

Other publishers were also contacted and asked about their policy on retracting very old articles. The Research Integrity group at Sage Journals offered: “Because many variables factor into retraction decisions, beyond our retraction policy, we’re not able to provide a general answer. Retraction decisions are weighed very carefully and consider the full context of each individual case.” ⁶³ The Director of Research Integrity & Publishing Ethics at Elsevier offered a similar but more detailed response: “Beyond our publicly available policies, which align with the COPE guidelines, Elsevier does not maintain separate or supplemental policies that specifically address historic articles or cases in which authors are no longer contactable (for example, due to death or having moved institutions)…In practice, concerns raised about older articles are investigated using the same principles and processes that apply to more recent publications. We do, however, recognize that the passage of time can present additional challenges, particularly where authors are unable to respond to or comment on the issues raised. Depending on the circumstances, the journal editor — or an appropriate editorial designate — may be able to reach a decision based on the available evidence. In other cases, where the issues cannot be fully resolved due to the age of the article or limitations in obtaining further information, the journal may issue an Expression of Concern to alert readers to the unresolved nature of the matter.” ⁶⁴

Precedents

There are few precedents for retracting articles published as far back as the Fleischer/Drinker piece. In 2003 and 2004, a European journal retracted two papers dating to 1923 when it became clear that fraud was involved. ⁶⁵ These 80-year-old papers were retracted despite the fact that the author was long deceased, and they were, at worst, historical curiosities, with little potential for misuse by modern readers. In another case, a large number of papers by psychologist Hans Eysenck and collaborators published as far back as 1950 have been retracted or flagged with “Expressions of Concern” indicating their potential unreliability. ⁶⁶ Much of Eysenck's work had been funded by the tobacco giant, Philip Morris International, his published work suggested that other lifestyle factors were much more important in cancer susceptibility than smoking cigarettes. Eysenck died over 20 years prior to initiation of corrective actions related to his published work. ⁶⁷

In March of 2026, The Lancet retracted a 49-year-old unsigned commentary on the safety of cosmetic talc ⁶⁸ after it was discovered the author was a paid consultant to Johnson & Johnson, then a leading producer of talc products. At that time, the U.S. Food and Drug Administration was considering requiring monitoring talc for asbestos contamination, the commentary argued that this testing was unnecessary. As with the Fleischer/Drinker article, this exculpatory blurb had been used for decades by attorneys defending talc producers in legal matters.^69,70

The Retraction Watch database (https://retractiondatabase.org/) list a total of 18 papers published between 1923 and 1959 that were either retracted (8), corrected (1), or annotated with an expression of concern (9). ⁷¹ All but 5 were acted on since 2003, 10 in 2020 alone. The number of retractions/annotations increased to 38 among articles published from 1960 to 1970 and 105 among those published from 1971 to 1980. A few of the flaws in this latter decade were flagged shortly after publication, but the majority were only identified 24 or more years after they appeared in the literature. Publisher Taylor and Francis has retracted/corrected/annotated 1956 papers in its history, the longest gap between publication and retraction listed in the database is 30 years (published 1989, retracted 2019). Other articles were retracted by Taylor and Francis journals after gaps of 21 and 24 years.

The Fleischer/Drinker paper has been used by asbestos industry defense attorneys to defend the actions of their clients since third-party asbestos injury suits first arose in the mid-1970s; it was used as recently as 2018 in a case argued before the U.S. Supreme Court. There is precedent for retracting scientific articles subsequent to their use in litigation. Sage Publications retracted two articles related to the abortion medication mifepristone that were cited by a Texas federal court judge when he briefly reversed the FDA approval of this medication.^72,73 Presumably, this retraction removed these articles from the arsenal of attorneys and judges mounting future efforts to ban this abortifacient. The retractions were made only weeks before the case went before the U.S. Supreme Court for review, which ironically overturned the lower court's adverse ruling on a technicality, not on its merits.

Can an editor/publisher be compelled to retract a defective article?

Lawsuits to stop or reverse retractions have occurred from time to time. Sage was promptly sued by the authors in the mifepristone case discussed above, claiming the decision was a political rather than technical one and it had caused “enormous and incalculable harm” to their professional reputations, and “they are now being treated as pariahs.”^74,75 In contrast, few lawsuits have been pursued to force journals to retract published articles, and most of those have been unsuccessful, even in high profile, high stakes cases. Such efforts are prone to be extremely expensive and drag on for years, making them practical only in cases involving very deep-pocketed plaintiffs where there is a great deal of money on the line. Recently in the United States, this has largely been dominated by suits filed by pharmaceutical companies against authors questioning the effectiveness or safety of their pharmaceutical products.

For example, in 2020, the Johnson & Johnson Company went to court in an attempt to force retraction of an article alleging that cases of mesothelioma were linked to exposure to cosmetic talc purportedly contaminated with asbestos. ⁷⁶ While the court refused to force the journal to retract the article, ⁷⁷ the author submitted an erratum notice in 2023 reporting that one subject in the study should not have been included due to asbestos exposure other than via use of talc. ⁷⁸ This did not alter the authors’ conclusions regarding the association between talc exposure and mesothelioma. Subsequently, a Johnson & Johnson subsidiary sued the article's authors in New Jersey federal court, claiming product disparagement, fraud and making false statements in advertising. In June, 2024, the judge dismissed the case.

Short of legal action, publicity and peer pressure may have some influence on retraction behavior by editors. Attention is key to retraction—one research group found that significant predictors of retraction are “indicators of prominence”: (1) the corresponding author's affiliation with a top 25 U.S.-based university and (2) the number of citations an article receives in its first year after publication. ⁷⁹ Attention in the lay press may also provide the impetus to retract articles—the journal Science only retracted a long-controversial article asserting that certain bacteria can incorporate arsenic atoms in place of phosphorus 15 years after publication, shortly after the New York Times, read by millions of people, profiled the first author and her struggles to have her research accepted.^80,81

Conclusion

A Health Survey of Pipe Covering Operations in Constructing Naval Vessels, the so called “Fleischer/Drinker” article, was technically unsound when it was published in 1946. Peer review—if any—failed miserably. Sufficient information to judge the article unsound is provided in the text and tables of the article itself, thus the unavailability of the authors, original data sets, or other records is largely moot. The ongoing accessibility of this paper to asbestos industry defense attorneys provides justification for retracting it now, even at this belated date. Attorneys and their experts may still be able to cite the paper in court even after it is retracted, but they could not characterize it as a “…well designed study” as has occurred in the recent past.

The refusal to consider retraction of the Fleischer/Drinker paper may be inconsistent with the COPE guideline for retracting old articles, but that policy is so vague that it is impossible to say if this is the case with any degree of certainty. There is no apparent way to force the editor/publisher of Archives of Occupational and Environmental Health to retract the Fleischer/Drinker article if they do not want to do so for any reason(s) at all. While the authors are long deceased and cannot contest a retraction, if the editor/publisher did decide to retract the paper, they might face resistance from the people who still may make use of it—asbestos defense attorneys.

The sole publishing industry guideline for handling very old articles is very general and does not provide a clear process or an algorithm to guide editors in making retraction decisions, a weakness that has been recognized but not yet rectified. More explicit guidelines, perhaps with a flowchart, would be beneficial to the industry. As it is, we may end up with an ever-growing list of defective yet immutable zombie articles, preserved for all time as if they are etched in stone.