By Summer 2018 M-VETS Student-Advisor Keith Bishop
16,000 Americans are expected to die of bladder cancer every year. Bladder cancer patients pass blood in their urine, suffer painful urination, and may suffer other symptoms such as back pain. Surgeries of various types are the most common treatment, including Transurethral Resection, in which a physician inserts a tube into the bladder and burns away the cancer with an electric current. “It’s not clear how many Vietnam veterans have suffered from bladder cancer.”
Vietnam veterans as a whole have been suffering from a pretty reliable host of conditions for decades. Because of this, Congress has made it easier for Vietnam veterans, whether soldier, sailor*, airman, or marine, to be approved for veteran’s disability compensation for many of these conditions. But despite all of these numbers of veterans with bladder cancer, neither Congress nor the Department of Veterans Affairs (“VA”) has added bladder cancer to that list that relaxes the burden of proof that veterans must meet to be awarded compensation.
Let’s unpack this a bit. Let’s look at the raw numbers, and the science and statistics behind this distinction between bladder cancer and other conditions closely associated with having served in Vietnam. Does it make sense for bladder cancer to be excluded, when so many conditions get special treatment?
BACKGROUND: Relative Risk
In basic civil law, a plaintiff—the person who feels aggrieved and is suing to win money or to have a court order the other party to do or not do something—needs to exceed a burden of proof called the “preponderance of the evidence” which essentially means ‘greater than 50%’ of the evidence is in his favor.  That could be 90%, or 60%, and some people say 51%, but that is almost 1% more than is required; 50.00000000000000000000001%, with an infinite number of zeros before that terminating 1, is sufficient to show a preponderance.
A lot of civil lawsuits, especially the ones against large corporations, are quite scientifically sophisticated, requiring teams of scientists, statisticians, and/or engineers to provide unbiased expert testimony. These are people unrelated to the parties with who by their “knowledge, skill, experience, training, or education” are in a position to explained to a jury their opinions and the basis for their opinions whether, for instance, an airplane wing was poorly designed or poorly manufactured, whether some scary-sounding chemical caused the victim to become sick, or whether a medicine was mistakenly released to the public or mistakenly prescribed. A lot of these kinds of lawsuits depend on a lot of mathematics and science.
When a plaintiff has met his burden of a preponderance, we can say that, all things being equal, defendant’s action or inaction was “more likely than not” the cause of plaintiff’s suffering. The most famous example in American law of this analysis was centered around Bendectin, a trade name for an anti-nausea medicine marketed to pregnant women from the 1950s until the early 1980s, and pulled from the market because it was suspected of causing birth defects. The Bendectin case, Daubert v. Merrell Dow Pharmaceuticals, reached the U.S. Supreme Court in 1993, and set the basis by which all federal courts evaluate scientific and expert evidence. The court below, whose decision was being appealed, had interpreted relative risk and the preponderance of the evidence standard mathematically: “plaintiffs must establish not just that their mothers’ ingestion of Bendectin increased somewhat the likelihood of birth defects, but that it more than doubled it—only then can it be said that Bendectin is more likely than not the source of their injury.” (emphasis added) Did you catch that? A suspected cause of injury must ‘more than double’ the risk of a naturally occurring problem. That sounds like a number. To a statistician who studies health, that is a very specific type of number, the relative risk.
To understand whether bladder cancer should be easier for Vietnam veterans to win disability benefits for, it is not necessary to be a statistician, or an epidemiologist, or even very scientifically inclined at all. But it is vital that you at least understand that relative risk is a conventional tool that statisticians use to determine whether we ought to consider a suspected cause as sufficiently linked to a suspected effect to feel comfortable concluding that they two are causally related.
So what is relative risk? For that, consider this well-constructed example from theanalysisfactor.com, which I will quote verbatim:
“Suppose you have a school that wants to test out a new tutoring program. At the start of the school year they impose the new tutoring program (treatment) for a group of students randomly selected from those who are failing at least 1 subject at the end of the 1st quarter. The remaining students receive the customary academic support (control group). At the end of the school year the number of students in each group who fail any of their classes is measured. Failing a class is considered the outcome event we’re interested in measuring. From these data we can construct a table that describes the frequency of two possible outcomes for each of the two groups.
“The probability of an event in the Treatment group is a/(a+b)= R1 . It’s the number of tutored students who experienced an event (failing a class) out of the total number of tutored students. You can think of it this way, if a student is tutored, what is the probability (or risk) of failing a class? Likewise, the probability of an event in the Control group is c/(c+d) = R2. Again, it’s just the number of untutored students experienced an event out of the total number of untutored students. Although each of these probabilities (i.e., risks) is itself a ratio, this isn’t the risk ratio. The risk of failing in the tutored students needs be compared to the risk in the untutored students to measure the effect of the tutoring. The ratio of these two probabilities R1/R2 is the relative risk or risk ratio. Pretty intuitive.
“If the program worked, the relative risk should be smaller than one, since the risk of failing should be smaller in the tutored group. If the relative risk is 1, the tutoring made no difference at all. If it’s above 1, then the tutored group actually had a higher risk of failing than the controls.” (emphasis added)
If you have read that more than once and still do not understand, it’s okay. The government decisionmakers in charge of adding disorders to the easier veterans’ lists rely on the advice of those who understand…and may not understand the mathematics involved themselves. The important point, is: in the example above, how badly would a tutor have to be before you could be pretty assuredly confidence he was alone, without much debate, making the students significantly worse? In basic epidemiology and toxic tort litigation, the standard is showing a relative risk of at least 2. RR = 2 is the same as saying that the defendant’s action or the effect of a chemical ‘more than doubled’ the risk. That is, a tutor more than doubled a student’s chance of failing.
What is the Relative Risk of death if you are shot in the chest? Probably at least 2—your chances of dying are at least twice as high as if you were not shot—unless you are only feet away from a well-run emergency room.
BASIC VETERANS DISABILITY LAW
Returning to veterans, typically, to be awarded disability compensation, the veteran must satisfy three conditions. He must (1) suffer a current disability, (2) have underwent an in-serve event or exposure that resulted in injury (or aggravation of a pre-existing condition), and (3) establish a nexus between the current disability and the in-service event or exposure. The word “nexus” is the tricky part. It basically means “was caused by.” How are veterans suffering from medical conditions in the 1980s, 1990s, 2000s, and 2010s supposed to show that their medical condition was caused by some experience in Vietnam decades earlier? Fortunately, Congress understood that almost no veteran suffering from any condition would be able to make that showing, even assuming that there was an absolute, definitive link between some environmental hazard, and some medical condition, because how many veterans were collecting soil samples while humping through the jungles, listening for the tell-tale sounds and smells of an impending ambush?
Because veterans of the same era and conflict often suffer and petition for disability compensation for many of the same conditions at the same time, and because there is often scientific evidence to support causal connections between hazards and diseases, Congress has eliminated, for certain classes of veterans, the need for each to establish that nexus. When this happens, we say that instead of a “direct-service connection” that the veteran can establish a “presumptive” service connection. Presumptive connection means that the condition is included in a list of disorders connected to certain categories of service, and can be set in law either by statute or by regulation. For Vietnam veterans, Congress in 1991 established an initial list of disorders commonly associated with service in Vietnam, believed to have been caused by their exposure to Agent Orange, an herbicide dropped in vast quantities throughout the conflict in order to assist troops maneuver there had previously been haltingly dense jungle, and related herbicides. Congress also directed the VA to promulgate new rules to add more conditions to the list of presumptive disorders based on the recommendation of the National Academy of Sciences that there existed sufficient statistical evidence that Agent Orange and other herbicides used in Vietnam were related to a given medical condition.
Even though bladder cancer is not included on either the statutory list of presumptive disorders, nor has been added to the VA’s regulatory list of presumptive disorders for service in Vietnam, veterans can still win claims for bladder cancer based on Agent Orange and herbicides by the conventional direct-service connection route. That is, the lack of a condition being presumptive does not mean that it is impossible to show it was causally connected to a disorder; it just means that neither Congress nor the VA believe there is a strong enough statistical connection to grant that service connection to all veterans with that exposure and a given disorder.
THE Presumptive CONDITIONS
Congress has specified, to simplify its language, eight conditions or sets of conditions as being presumptive tied to exposure to herbicides in Vietnam:
- Non-Hodgkin’s lymphoma
- Soft-tissue sarcoma (with exceptions)
- Chloracne / acneform disease
- Hodgkin’s disease
- Porphyria cutanea tarda (PCT), the most common subtype of porphyria
- Respiratory cancers (cancer of the lung, bronchus, larynx, or trachea)
- Multiple myeloma, sometimes known as plasma cell myeloma
- Diabetes Type-2
The VA has used its rulemaking authority, based on the scientific recommendations provided by the National Academy of Sciences, to add…six(?)…more:
- AL amyloidosis
- Ischemic heart disease
- Chronic B–cell leukemias
- Parkinson’s disease
- Early-onset peripheral neuropathy
- Prostate cancer
So….why not bladder cancer? Is bladder cancer sufficiently different from a scientific, statistical standpoint, that the conditions above are all presumptive service connected, and it is not?
To sidestep that question briefly, you should know that the National Academy of Sciences and the VA are currently working to add bladder cancer to the list. In March 2016, the National Academy of Science released its 2014 Update, in which it upgraded its assessment of the possible connection between Agent Orange and bladder cancer from “inadequate or insufficient” to “limited or suggestive” because, among other things, a new study of Korean and Vietnam veterans showed increased mortality caused by bladder cancer after higher exposure to herbicides. Based on this, the VA is reported to have lobbied the White House’s budget office to include bladder cancer as a presumptive disorder, but no announcements have been made in response to such a recommendation.
What is behind that choice, that change in position? How does NAS and the VA assess “increased mortality”? If you said Relative Risk, congratulations (…..somewhat).
In particular, one new study had some interesting results: “Yi et al. (2014b) reported a statistically significant two-fold increase in bladder cancer–specific mortality (RR = 2.04, 95% CI 1.17–3.55) comparing the high- and low-exposure groups without adjustment for smoking; these results were based on 42 deaths from bladder cancer in the high category.”
If you caught that RR in that statement was greater than 2, congratulations, you are ready to make national health policy.* But the NAS and VA don’t rely on just one study. The Update includes a list of dozens of studies, and most of them do not report RR of at least 2.
So let’s compare one condition already on the lists with bladder cancer, which is not, but possibly could be soon.
According to the National Academies, for Non-Hodgkin’s lymphoma, the first-listed disorder in the statute, “there is sufficient evidence of an association between exposure to at least one of the [herbicides] and NHL.” The Update is written by experts who go out of their way to prevent readers with only my level of expertise from misinterpreting their results; there are twenty pages of listings of scientific article summaries, which include individual RRs, but the narrative that explains the Academies’ assessment rarely refers to relative risks and does not base its ultimate decision on the result of a massively compounded single-number result…as amateurs may want them to. Reviewing the data tables, however, there does seem to be a clear difference in the numbers, in the typical RR numbers of the bladder cancer tables (pages 517-530) compared to the Non-Hodgkin’s lymphoma tables (pages 591-610). For any blog readers truly sophisticated in this level of research and mathematical aptitude, I now invite you to laugh at me.
A satisfying understanding of why the Update treats bladder cancer one way and Non-Hodgkin’s lymphoma another requires serious study. There are four dense pages of discussion and conclusions about bladder cancer, and six dense pages of conclusions about Non-Hodgkin’s lymphoma What are the scientists looking for, if not just some oversimplifying statistical tool? They have a series of categories of analysis they consider for each disorder: studies of Vietnam veterans themselves, occupational and environmental studies, case-control studies, environmental studies, and biological plausibility. After examining those categories of analyses in isolation, the National Academies synthesizes those conclusions to identify trends and themes, and then provides a conclusion. Looked at from that perspective, is bladder cancer truly different from the other, established, conditions, or at least Non-Hodgkin’s lymphoma?
Comparing the bladder cancer synthesis and conclusion with that of Non-Hodgkin’s lymphoma, it is apparent why Non-Hodgkin’s lymphoma has long been presumptively service connected and bladder cancer is only now under serious consideration to become so. The Non-Hodgkin’s lymphoma synthesis and conclusion indicates that “The first [such] committee found the evidence to be sufficient to support an association” between the suspected herbicides and Non-Hodgkin’s lymphoma. It explains further, and I apologize for not being able to adequately clean up this density of technical language:
“Results of some high-quality studies show that exposure to 2,4-D and 2,4,5-T appears to be associated with [chronic lymphocytic leukemia (now regarded as same disease as small lymphocytic leukemia [SLL] and designated by some as CLL/SLL)], including the incidence study of Australian veterans (ADVA, 2005a), the case-control study by Hertzman et al. (1997) of British Columbia sawmill workers who were exposed to chlorophenates, the Danish-gardener study (Hansen et al., 1992), and the population-based case control study in two US states by Brown et al. (1990) that showed increased risks associated with any herbicide use and specifically the use of 2,4,5-T for at least 20 years before the interview.”
For my purposes, the almost-lay perspective, it suffices that these experts believe that they have “high-quality studies” that show “increased risks… at least 20 years before the interview.”
In contrast, the Update examines the possibility of bladder cancer quite more skeptically: “Many of the available analyses of an association between exposure to the [herbicides] and bladder cancer risk are characterized by low precision because of the small numbers of exposed cases, low exposure specificity, and a lack of ability to control for confounding, particularly cigarette smoking, which is a major risk factor for bladder cancer.” Less overwhelmingly confident than the Non-Hodgkin’s lymphoma synthesis, though, the bladder cancer synthesis does have a theme of positive connection. In analyzing one large study, the Update concludes “indicating that the results for bladder cancer mortality are unlikely to have been majorly confounded by smoking.” Similarly: “Subsequently, follow-up reports on mortality after 1992 in several of the IARC subcohorts found elevations in bladder cancer mortality,” and then it described the results of multiple reports with promising* results, but then had to tamper that optimism with explanations of an equal number of reports with minimal increases in mortality, or exposure that led to no mortality at all.*
What should we make of this? Well, trust in the National Academies of Science to look at what really matters in science—replicate-ability of results, plausibility, and an intelligent look at confounding variables—is being earned, while mindless worship of numbers is thankfully not. And after looking through a fair amount of the Update, I for one have been convinced that
No, not bladder cancer. Not yet.
Unfortunately, maybe bladder cancer is, hiding under all this data, under all the limitations of conducting research, including the time it takes to replicate study after study, very well caused or at least strongly influenced by the herbicides and other chemicals to which so many of our veterans were exposed in Vietnam. If so, we may prove that beyond a reasonable doubt, and hopefully do so before it is too late to provide them with benefits and government-sponsored treatment.
On the other hand, though, it is certainly possible that these suspected connections, the sometimes fit of the data, are the result of pure chance, and smoking, and other problems of research design, or analysis. Note that none of this discussion has mentioned experiment, the best way to truly tease out causality, because of course we cannot conduct experiments with herbicides and human beings. All we are left with is math. Math and the patience to apply it correctly. Society, though, and these veterans in particular, have little use for patience.
 Keith R. Bishop earned a Bachelor’s of Arts degree in Psychology in 2003 from The George Washington University, focusing on cognitive neuroscience. He earned his Juris Doctor degree with a National Security Law and Policy concentration in 2018 from Antonin Scalia Law School, George Mason University, where he participated in the Mason Veterans and Servicemembers Legal Clinic (M-VETS). He has served as a U.S. Army Chemical officer since 2003, including two deployments to Iraq. He currently serves as commander, 231st Chemical Company, Maryland Army National Guard. The views and opinions expressed in this article are the author’s alone and do not represent the views of the Department of Defense, the U.S. Army, the Maryland Army National Guard, or any other governmental agency.
 National Academies of Science, Veterans and Agent Orange: Update 2014 (“Update”), 515.
 Charles Ornstein and Terry Parris Jr., “Vietnam Vets Push VA to Link Bladder Cancer to Agent Orange”, Pro Publica, Inc., April 27, 2016, https://www.propublica.org/article/vietnam-vets-push-va-to-link-bladder-cancer-to-agent-orange.
 There is ongoing political and legal debates as to how close to Vietnam’s shores sailors must have been in order to draw benefits. See Leo Shane III, “Senate plans path ahead for ‘blue water Navy’ benefits fix”, Military Times, July 17, 2018, https://www.militarytimes.com/veterans/2018/07/17/senate-plans-path-ahead-for-blue-water-navy-benefits-fix/.
 I preface this section with an important caution. I do not profess to be a statistician or a scientist of any kind. I am only as knowledgeable about these specialized, very technical subjects, as anyone who has studied them casually. One of my main purposes for this blog entry is to be farcical: if decisionmakers with no more than my level of expertise in these subjects are satisfied with and making important policy choices based on only raw relative risk numbers and other statistical and mathematical results, without understanding the inherent flaws or limitations behind each type of measure, as I can only partially grasp myself, then their decisions will be at least as untethered from the true causalities involved as random luck, poor experimental design, and bad science allow.
 I will use male pronouns simply to simplify this article, and do not intent to convey that there aren’t many women plaintiffs.
 Federal Rule of Evidence 702(a).
 The scientific consensus today is that it was pulled from shelves prematurely, due to justifiable fears of legal liability. See, e.g. Bexis, “The Ghost of Bendectin – Exorcized by the FDA?”, Drug & Device Law (website), November 17, 2015, https://www.druganddevicelawblog.com/2015/11/the-ghost-of-bendectin-exorcized-by-fda.html.
 Daubert v. Merrell Dow Pharmaceuticals, 509 U.S. 579 (1993).
 See, e.g. Standards and Procedures for Determining the Admissibility of Expert Evidence After Daubert, 157 F.R.D. 571, 571 (1994).
 Daubert v. Merrell Dow Pharm., Inc., 43 F.3d 1311, 1320 (9th Cir. 1995)
 I will not address the inherent mathematical arbitrariness of the “more than double” or “p < 0.05” standards in this bog post. Those scientific conventions have been set for one hundred years. See, for example: Gerard E. Dallal, Why P=0.05?, http://www.jerrydallal.com/lhsp/p05.htm (Last modified: 05/22/2012 22:52:42).
 Audrey Schnell, “The Difference Between Relative Risk and Odds Ratios”, https://www.theanalysisfactor.com/the-difference-between-relative-risk-and-odds-ratios/ (accessed August 15, 2018).
 Earlier I explained what “preponderance of the evidence” means. In veterans law, in contrast, the veteran wins at the “as likely as not”, or 50%, burden of proof. 38 C.F.R. § 3.102.
 Some sources cite 38 CFR 3.303 for these elements, but the language of that section is far less concise. Many cases cite those elements to Shedden v. Principi, 381 F.3d 1163, 1166 (Fed. Cir. 2004), which itself cites Hansen v. Principi, 16 Vet. App. 110, which cites Caluza v. Brown, 7 Vet. App. 498, 505 (1995), aff’d, 78 F.3d 604 (Fed. Cir. 1996). Caluza does not use those terms, but does indicate that its own explanation is a holistic reading of “38 U.S.C. §§ 1110, 1112(a), 1131, 1137; 38 C.F.R. §§ 3.303(a), 3.306, 3.307 (1994).” Other opinions cite Davidson v. Shinseki, 581 F.3d 1313 (Fed. Cir. 2009), which also cites Caluza.
 PL 102–4, February 6, 1991, 105 Stat 11.
 Jeanne Mager Stellman, et. al. “The extent and patterns of usage of Agent Orange and other herbicides in Vietnam,” Nature, 17 APRIL 2003, http://epiville.ccnmtl.columbia.edu/assets/pdfs/Stellman3.pdf.
 38 U.S. Code § 1116(a)(1)(b).
 38 U.S.C. 501(a) and 1116.
 Bladder cancer is presumptive for some other categories of veterans. See, e.g. 38 C.F.R. § 3.309(e) (radiation-exposed veterans) and Federal Register /Vol. 82, No. 9 / Friday, January 13, 2017 (Final Rule – Diseases Associated with Exposure to Contaminants in the Water Supply at Camp Lejeune).
 Combee v. Brown, 34 F.3d 1039, 1044 (Fed. Cir. 1994), citing 38 U.S.C.A. §§ 1113(b) and 1116, and 38 C.F.R. § 3.303. “To permit the denial of service connection for a disease on the basis that it is not likely there is any nexus to service solely because the statistical analysis does not support presumptive service connection, would, in effect, permit the denial of direct service connection simply because there is no presumptive service connection. This is contrary to the recognition in Stefl that ‘[t]he existence of presumptive service connection for a condition based on exposure to Agent Orange presupposes that it is possible for medical evidence to prove such a link before the National Academy of Sciences recognizes a positive association.’ Stefl, 21 Vet.App. [120,] 124 [(2007].” Polovick v. Shinseki, 23 Vet. App. 48, 55 (2009).
 38 U.S.C. § 1116(a)(2)
 One must be deftly familiar with oncological terms to properly count the distinct conditions references in the regulations, as the regulations define many conditions, but indicate that many are sub-categories of other already-specified conditions.
 38 C.F.R. § 3.309(e).
 Veterans and Agent Orange: Update 2014, 10.
 Id. at 516.
 Wentling, “Vets with bladder cancer could wait years for government to recognize Agent Orange link”.
 Returning to my point about this being a farcical article, this parenthetical term is a measurement of the Confidence Interval. That term indicates that although the statistical average Relative Risk of this study is slightly above two, it is very possible that the true Relative Risk was only 1.17, which would fall far short of “doubles the risk”. On the other hand, this study is as equally as likely to indicate that the true relative risk is 3.55, which would mean that Agent Orange exposure more than tripled the risk of mortality in the population.
 Veterans and Agent Orange: Update 2014, 516.
 Id. at 517-26.
 Id. at 618.
 Id. at 591-610.
 Id. at 590, 614-18.
 Id. at 516, 527-29.
 Id. at 590, 614-18.
 Id. at 618.
 Id. at 527-28.
 Id. at 528.
 Promising for the veterans’ legal cases, not promising for the patients’ prognoses.
 The opposite of the previous footnote, of course.
 Note that that is not a term commonly used in science.