代写POLITICAL SCIENCE 2120: “‘FOLLOW THE SCIENCE?’ THE POLITICS OF HEALTH” Midterm Exam代写Processing

日期：2024-09-15 06:40

POLITICAL SCIENCE 2120:

“‘FOLLOW THE SCIENCE?’ THE POLITICS OF HEALTH”

Midterm Exam — Take-Home Questions

Colorectal cancer is the third-leading cause of cancer death among both men and women in the United States. And it appears to have become more common among younger cohorts since the mid-1990s.

There are several procedures currently used to screen asymptomatic people to try to identity this disease in its earliest stages, when it might be more treatable, with colonoscopy being the most common.

Colonoscopies can be both diagnostic (used to detect disease) and therapeutic (used to treat) — if precancerous polyps are found, they can be removed during the procedure, potentially preventing cancer from developing later.

The United States Preventive Services Task Force (USPSTF) is a volunteer, government-sponsored panel that synthesizes the best available medical research and aims to provide evidence-based recommendations for health procedures designed to prevent serious illness. USPSTF does not consider costs in its recommendations — it only looks to see whether the research suggests that a particular procedure is likely to benefit patients, regardless of the magnitude of the benefit or the financial cost.

By law, services recommended by the USPSTF must be covered by health insurance without any charge to patients.

Currently, USPTSF currently recommends that all adults aged 45 and older undergo colorectal cancer screening. While several other cancer screening methods have been examined using randomized controlled trials (RCTs), colonoscopy had not — until now. Last October, the New England Journal of Medicine published the first RCT of a colonoscopy screening program for the general population. (Note: We are not talking about screening for high-risk populations, such as people with family history of particular cancer.) That study is posted on Carmen. You will need to read it before answering the  questions below. You are not expected to read the additional documents hyperlinked in this file; I included them as background for interested students.

After you have read the study, answer all of the questions described below. Note you can tackle each   part separately — there is no need to try to combine your answer into a single, coherent essay. Do not write more (or less) than is necessary to answer the questions. The goal is to have clear and concise answers, without fluff or throat-clearing.

Part A: In your own words, briefly summarize the design of the study and how it was carried out. Also briefly summarize its overallfindings about the effects ofbeing randomly assigned to receive a referralfor screening. Note that not everyone assigned to get screened actually followed through, a major criticism of this study you’ll consider below. Because it is the invitation to get screened — rather than actually getting screened — that is randomly assigned, the experiment is able to estimate what is known as the “intent-to-treat” effect, often abbreviated as ITT.

Part B: When evaluating medical interventions, it is important to understand the difference between absolute and relative effects. For example, suppose that 2 percent of people who get Covid-19 end up passing away. (This is not the true estimate, I’m just using it as an example.) Suppose further that we find a medication that reduces the death rate from 2 percent to 1 percent. That represents a relative reduction in death of 50% (1 is 50% smaller than 2), but an absolute reduction of only 1 percentage point.

A quick note: In research, including the study you are reading, relative risk is often reported as a “risk ratio.” The baseline risk (e.g., for the control group) is normalized to 1, and then the increase or decrease is measured relative to this baseline. As an example, a risk ratio of 1.15 corresponds to a 15% increase above the baseline (1.15-1=0.15 as a decimal, or 15% as a percent). A risk ratio of 0.85 corresponds to a 15% decrease (0.85-1=-0.15, or -15%).

Using the intent-to-treat estimates from the colonoscopy study, what are absolute and relative effects of colonoscopy screening on the 10y-ear risk of (1) being diagnosed with colorectal cancer; (2) death from colorectal cancer; and (3) death from any causes. You can present these estimates in table form, if that is easier. Hint: All of the numbers you need  to calculate these effects are found in Table 2 of the study.

Part C: There is much debate among medical researchers about the most relevant endpoint, or outcome  of interest, for measuring the impacts of screening and other preventative procedures. Some argue that   we should focus on overall death, regardless of the cause, since ultimately that is what people care about and this end point can capture unintended health harms of screening and treatment. Others argue that we should focus on disease-specific death, since everyone will die of something at some point. Showing that a treatment prevents you from dying from cancer only to die from a heart attack a few years later still means there is a meaningful impact of screening and meaningful gains in life expectancy.

This debate is particularly heated in the context of cancer screening. For breast cancer, for example, a    number of studies have found that mammography reduces deaths due to breast cancer — but does not appear to affect overall mortality from all causes.

From the perspective ofpolicy — whether screenings should be recommendedfor the whole population by the government, and whether the government and private insurers should be required to pay for them — rather than the perspective of an individualpatient, which endpoint should we use to make policy decisions?

Part D: Some might be surprised that cancer screenings don’t always prolong people’s lives. This could happen for three reasons. First, diagnosing cancer early may not actually help — either because no effective treatments exist, or because the cancer that was found was unlikely to kill you. (Prostate cancer, for example, is usually very slow-growing. Many men over the age of 70 who die from some   other cause are found, upon autopsy, to have had a cancer growing in their prostate but likely had no symptoms of the disease.)

Second, the screening procedure may itself harm health, offsetting the benefits. Colonoscopy, for example, can cause bowel perforation and infection, bleeding, severe abdominal symptoms, and even  heart attacks. These unintended side effects are quite rare but may outweigh the benefits of proactive screening at the population level if the cancer is also rare.

Third, the value of such tests depends on their accuracy, and all clinical tests are imperfect. Clinical tests are evaluated using two main metrics: sensitivity and specificity. Sensitivity is the percent of actual cancers that the test accurately detects (in other words, the probability that the test will accurately detect true cases of disease). Specificity is the percent of patients without cancer that the test accurately identifies as being healthy (rather than a false-positive). Both sensitivity and specificity for colonoscopy are around 90 percent.

Now, many people think that 90 percent sensitivity and specificity imply that, if you have a colonoscopy that comes back positive, it means you have 90 percent chance of actually having cancer. But that’s incorrect! To interpret clinical test results, you also need to consider the prevalence of the disease in the population (also known the pre-test probability that a person has a disease).

For example, suppose that only 1.5 percent of the population actually has colorectal cancer — a ballpark estimate based on the RCT study you read for this exam. Among this subset of people, screening will accurately find 90 percent of the cancers (sensitivity=90%), or about 1.35 percent of the  population (1.5 percent of population with cancer * 90% sensitivity). But among the other 98.5 percent of the population that is cancer-free, the colonoscopy is going to indicate cancer 10 percent of the time (100% – 90% specificity = 10% false-positive rate). That’s 9.85 percent of the total population!

Add up those number — 1.35 percent of the population that is true-positive plus 9.85 percent of the population that is false-positive — and you can estimate that just 12% of the positive colonoscopies are real cancer cases (1.35+9.85/1.35), while the remaining 88% are false positives among people who don’t have   cancer. Even with really accurate tests, there will always be many false positives for conditions as rare as colorectal cancer. And these false positives can have real harms, including unnecessary anxiety and worse. For example, both suicides and heart attack rates have been found to increase significantly among men who are told that their prostate screening indicates they have cancer, even though most of these diagnoses turn out to be false positives upon further testing.

Are you surprised by the true vs.false positive calculation above? How should these figures be incorporated into the policy decision about whether to recommend cancer screening for the whole population?

Part E: Many vocal supporters of colonoscopy — including professional associations for doctors who make a living performing them — have been very critical of the population colonoscopy screening RCT you read. They have made two arguments.

First, they argue that the results from the experiment are not consistent with earlier observational  (non-

RCT) research, which seemed to show much bigger benefits of colonoscopy. For example, a 2018 study found that colonoscopy reduced colorectal cancer deaths by 67 percent over 10 years (this is a relative risk reduction). This study used the case-control design we have covered in class: Patients who died from colorectal cancer were matched based on observable characteristics, including sex, age, and geography, to people who didn’t die from cancer, and then the rate of colonoscopies between the two groups was compared. Drawing on whatyou learnedfrom the other examples we have studies so far this semester,

explain why this observational design is likely to overstate the benefits of colonoscopy, and why should we prefer to calculate benefits using an RCT. Be sure to discuss specific readings or examples from class in your answer.

Second, only 42 percent of the participants randomly assigned an invitation to get a colonoscopy in the RCT actually received one. “A colonoscopy will only work if a patient gets one,” a gastroenterologist highly critical of the study told NPR. In addition to the main intent-to-treat results, the authors of the   study present a supplemental “per-protocol” analysis. In this analysis, only the 42 percent of the treatment group that actually chose (i.e., self-selected) to get a colonoscopy were matched to observationally similar participants in the control group, and the authors found much larger benefits of screening in this per-protocol analysis than in their overall estimates. Should we believe the results of this analysis? How could they be biased in ways that the intent-to-treat analysis is unlikely to be? Again, your answer should mention specific examples from readings or class.

Part F: Policymaking is driven in part by evidence, but also many other political considerations,

including advocacy from interest groups — like colonoscopy providers, who responded “swiftly and

unequivocally to the media coverage” of this study — and well-meaning policy activists. The latter

group includes cancer survivors. People who have an asymptomatic cancer discovered and removed as a result of undergoing a routine screening will naturally believe that the screening saved their lives and become powerful public advocates. (They assume, of course, that the cancer would’ve killed them had it not been found; as we saw in the prostate cancer example above, that may not be true. And they ignore the health harms created by the false positives for other people who were screened and didn’t end up having cancer.)

Another group of activists is family members and loved ones of people who have died from cancer.

They also advocate for universal asymptomatic screening in the hope that it will catch disease early

enough for treatment and help other families avoid the pain they have suffered. (They assume that early detection will indeed lead to effective treatment, which is not always the case, and also ignore the harms resulting from false positives.)

One such advocacy group, the Colon Cancer Coalition, put out a statement in response to the study you read suggesting that it “may be misleading” and arguing that “getting screened for CRC does save lives.”

What role should these kinds ofadvocacy and considerations have in the decision about whether to recommend cancer

screening? What should policymakers do when the anecdotal stories from cancer survivors andfamilies conflict with

evidence provided by rigorous RCTs? Overall, how do you think these kinds ofpolitical — rather than evidence-based — considerations affect government policy?

Part G: As noted in the introduction, the USPSTF does not consider the cost of treatments. But these costs are real. For the average patient, a colonoscopy has been estimated to cost $1,200. When paid for by private insurance, this will increase the cost of health insurance for everyone and make it unaffordable for some. When paid for by the government (e.g., through Medicare and Medicaid), it could take away resources from other government programs, including programs that could improve people’s health and extend their lives in other ways.

Although USPSTF is not allowed to consider costs and tradeoffs, policymakers can do so. Suppose you are working as an adviser to public officials and helping policymakers decide whether the government should recommend that everyone gets screenedfor colorectal cancer by undergoing a colonoscopy. Incorporating whatyou learnedfrom reading the RCT and the analysisyou didfor the questions above, as well as any other relevant consideration that we may have notyet taken into account, what advice wouldyou give to policymakers? Be sure to clearly articulate and explain the reasoning behindyour recommendation.