# Lesson 4: Alcohol and Other Depressants

*Companion-podcast transcript • Sarah & Kiffer*  
*Office Hours episode to listen to after working through the lesson*

---

**Sarah:** Welcome back to Office Hours. I'm Sarah.

**Kiffer:** And I'm Kiffer. Alcohol and the other depressants are the subject of Lesson four, and alcohol's pharmacology has a feature that sets it apart from most of what we've covered so far.

**Sarah:** Question one. Most drugs in this course leave the body according to a half-life. Why doesn't alcohol, and why does it matter?

**Kiffer:** Lesson three described a half-life as a constant fraction of a drug removed in each period of time. That pattern holds when the enzymes that break a drug down have plenty of spare capacity. Alcohol dehydrogenase, the main enzyme for alcohol in the liver, is saturated after the first few drinks. Once it's working at full capacity, it removes about the same amount each hour, however much alcohol is waiting. In moderate drinkers that's about fifteen milligrams per hundred millilitres of blood per hour, on average.

**Sarah:** So here's one for everyone listening. A hypothetical person's blood alcohol peaks at sixty milligrams per hundred millilitres. At fifteen an hour, how long until it's gone, and how long if the peak were twice as high, at one hundred twenty? Take a few seconds, and pause the audio if you'd like more time.

*(Pause)*

**Kiffer:** Sixty divided by fifteen is four hours. One hundred twenty divided by fifteen is eight hours. Double the peak and the wait doubles too. Compare a drug with a half-life. There, each doubling of the dose adds one more half-life to the time it takes to fall to any given level, so a dose four times as large adds only two half-lives. With alcohol, a peak four times as high takes about four times as long to clear.

**Sarah:** And fifteen an hour is an average. How much do people vary?

**Kiffer:** Quite a lot. A review of the research by Wayne Jones, a forensic toxicologist in Sweden, found rates of about ten to fifteen after drinking on an empty stomach, fifteen to twenty after a meal, and twenty-five to thirty-five in people with alcohol dependence, whose livers have stepped up a second enzyme, CYP2E1. At ten an hour, our peak of one hundred twenty takes twelve hours to clear.

**Sarah:** And the clock doesn't really start at the last drink, either. Alcohol that's still in the stomach and intestine keeps entering the blood.

**Kiffer:** Right. The United States National Institute on Alcohol Abuse and Alcoholism points out that blood alcohol can keep rising after a person stops drinking. There's one more wrinkle at the other end. Jones notes that the steady pace describes the decline well at levels above about twenty. Below that, the enzyme is less saturated, and the last of the alcohol leaves more slowly, so the times we just calculated are, if anything, a little short.

**Sarah:** Question two follows straight from that. How far can we trust a blood alcohol estimate from the Widmark formula?

**Kiffer:** Let's test it on the hypothetical person from the practice problem in the lesson: seventy kilograms, four standard drinks, three hours after they started drinking. With a Widmark factor of 0.68 and an elimination rate of fifteen, the estimate is about sixty-eight. Each input has a range, though. For people of average build, the Widmark factor runs from about 0.61 to about 0.72, and in moderate drinkers elimination runs from about ten to about twenty an hour. Sarah, what are the highest and lowest estimates those ranges allow? Everyone else, try it too. Take a few seconds.

*(Pause)*

**Sarah:** Okay. Four standard drinks is four times 13.45, or 53.8 grams. For the highest estimate I'll use the biggest numbers. A Widmark factor of 0.72 times seventy kilograms is 50.4. One hundred times 53.8 is five thousand three hundred eighty, and divided by 50.4 that's about one hundred six point seven. Then the elimination is twenty an hour for three hours, which is sixty. And one hundred six point seven minus sixty is about 46.7. So the highest estimate is about forty-seven.

**Kiffer:** Before you go on, compare that with the central estimate of sixty-eight. Does it make sense?

**Sarah:** No, the top of the range can't sit below the middle. I see what I did. The Widmark factor is in the denominator, and it stands for body water, so a bigger factor means the alcohol is diluted into more water and the level comes out lower. And a faster elimination takes more away. So the highest estimate needs the smallest factor and the slowest elimination.

**Kiffer:** Good catch. Try it again with the smallest factor and the slowest rate.

**Sarah:** Okay, 0.61 times seventy is 42.7. Five thousand three hundred eighty divided by 42.7 is about one hundred twenty-six. Ten an hour for three hours is thirty, so one hundred twenty-six minus thirty is ninety-six. And the forty-seven I got first is really the lowest estimate, from the largest factor and the fastest elimination.

**Kiffer:** So the same four drinks give anywhere from about forty-seven to about ninety-six, roughly a twofold range. And that's before asking whether all the alcohol had been absorbed, or whether the drinks were really standard. The Canadian Institute for Substance Use Research points out that a nine-ounce glass of fourteen percent wine holds more than two standard drinks.

**Sarah:** Which is why a calculation like this is good for learning how the body handles alcohol, and can't tell anyone whether a particular person is impaired or able to drive.

**Kiffer:** Yes. Measurable effects begin at around twenty-eight milligrams per hundred millilitres, far below the Criminal Code driving limit of eighty, and no formula can see a real person's body water, liver or stomach contents. What the formula teaches well is the direction each factor pushes. More body water lowers the level, food delays the peak, and the liver keeps its own pace.

**Sarah:** Question three. If someone takes a benzodiazepine sleeping pill every night, exactly as prescribed, for years, are they addicted?

**Kiffer:** Most likely they're physically dependent, which is a different thing. Physical dependence means the body has adapted to the drug, so stopping brings withdrawal. Addiction means compulsive use despite harm, which Lesson eight develops. A person can be dependent on a medicine they take exactly as directed and show no sign of addiction. Mixing up the two words can attach stigma to people who have simply followed their prescription.

**Sarah:** And after years of nightly use, is the pill still helping them sleep?

**Kiffer:** Possibly much less than it did. Professor Heather Ashton, whose manual on benzodiazepine withdrawal is still widely cited, reported that tolerance to the sleep effect develops within a few weeks of regular use. After that, much of what the nightly pill does may be holding off withdrawal symptoms such as insomnia and anxiety. So the person's experience is that they can't sleep without it, which feels like proof that it works.

**Sarah:** Which is also why stopping suddenly is dangerous. The brain has adjusted to a boosted GABA signal, and taking the drug away all at once leaves too little inhibition. After long-term, high-dose use, that can mean seizures.

**Kiffer:** And this is where half-life comes back, in an example that goes a step past the lesson. Take two drugs from the half-life table. For alprazolam, let's use six hours, the short end of its range. For diazepam, let's use twenty-four hours, which sits near the short end of its range. How much of each is left a day after a single dose?

**Sarah:** Diazepam is easy. Twenty-four hours is one half-life, so half is still there, and its active metabolite lasts even longer. For alprazolam, a day is four half-lives. A half, then a quarter, an eighth and a sixteenth, so only about six percent is left.

**Kiffer:** So with the shorter-acting drug, the level rises and falls steeply around each dose, and any cut in the dose shows up within a day. With diazepam, the level drifts down gently over several days. That's the pharmacology behind Ashton's view that switching from drugs like alprazolam to a slowly eliminated drug like diazepam can smooth withdrawal, and behind her advice to reduce slowly, usually over months, at a pace the person controls. In any real case, that plan is worked out with a prescriber.

**Sarah:** Then there's the trial in Quebec pharmacies led by Cara Tannenbaum. Older adults who had taken benzodiazepines for a long time received printed information about the risks and a step-by-step tapering plan. After six months, twenty-seven percent had stopped, compared with five percent of those in usual care.

**Kiffer:** The adjusted risk difference was twenty-three percentage points, which gives a number needed to treat of about four. The trial also reported a ninety-five percent confidence interval for that difference, a range of plausible values, running from fourteen to thirty-two points. So, for everyone: what range of numbers needed to treat goes with that interval? Take a few seconds.

*(Pause)*

**Sarah:** One divided by 0.32 is about 3.1, and one divided by 0.14 is about 7.1. So somewhere between about three and about seven. And the ends swap over, because the biggest difference gives the smallest number needed to treat.

**Kiffer:** Right. Even the cautious end, about seven people receiving the printed material for one more person to stop, is a large effect for patient education. I'd add two cautions, though. The follow-up was six months, so the trial can't tell us how many people stayed off the drug after that. And the participants were older adults in one province, so the result may not carry over to younger people who take benzodiazepines for anxiety.

**Sarah:** Before question four, a short content note, because this part discusses overdose deaths. If you'd like support, you can call or text nine eight eight, the Suicide Crisis Helpline, at any time. The National Overdose Response Service, for people who use drugs alone, is at one, eight eight eight, six eight eight, six six seven seven. The Hope for Wellness Help Line for Indigenous people is at one, eight five five, two four two, three three one zero. So, question four. When a benzodiazepine turns up in an overdose death, what does that tell us?

**Kiffer:** Let's start with a real figure. Researchers in Ontario reviewed all eleven thousand thirty-three death investigations from 2017 to 2021 in which toxicology testing found an opioid. A benzodiazepine was found in fifty-four point five percent of them. The most common was etizolam, a benzodiazepine-type drug. It's tempting to read that as benzodiazepines explaining more than half of those deaths.

**Sarah:** And that reading has the same problem as the nitrous oxide case reports. Eighty percent of the hundred published patients had tingling or numbness, yet every one of them was already ill. Death investigations tell us what was in the blood of people who died. They can't tell us how common the same drugs were among everyone who used opioids, including the many people who survived.

**Kiffer:** So let's see why that missing number matters, with a hypothetical example. Imagine a group of people who use opioids, and suppose half of them are also exposed to a benzodiazepine-type drug. Now suppose, just for the exercise, that the benzodiazepine adds no risk at all, so each half dies at the same rate. What share of the deaths would show both drugs? Take a few seconds.

*(Pause)*

**Sarah:** Half. Two groups of the same size dying at the same rate contribute the same number of deaths. So fifty percent of the deaths would show both drugs, even though the benzodiazepine added nothing.

**Kiffer:** Now the harder version. Keep the deaths at fifty-four point five percent, the real Ontario figure. Then suppose, again hypothetically, that only one person in five in our group is exposed to a benzodiazepine, counting everyone who used opioids, whether they died or survived. How many times higher is the risk of dying among the people who are exposed?

**Sarah:** Let me divide each group's share of the deaths by its share of the people. The exposed are twenty percent of the group and account for fifty-four point five percent of the deaths. Fifty-four point five divided by twenty is about 2.7. The unexposed are eighty percent of the group and account for forty-five point five percent of the deaths. Forty-five point five divided by eighty is about 0.57. Then 2.7 divided by 0.57 is about 4.8. So in this made-up group, the risk of dying among the exposed would be nearly five times the risk among the unexposed.

**Kiffer:** And if half the group were exposed, the same fifty-four point five percent would mean a risk only about 1.2 times as high. Both of those exposure shares are invented, so neither answer is an estimate for Ontario. One real figure fits anything from a small difference in risk to nearly five times the risk, depending on a number that death investigations don't collect.

**Sarah:** Then how do we know the combination is dangerous at all?

**Kiffer:** From evidence that supplies what the death counts can't. The study of United States veterans receiving opioid pain medicines compared the two thousand four hundred who died of an overdose with a random sample of more than four hundred thousand veterans who also received opioids. That sample is the missing denominator. Veterans with a current benzodiazepine prescription died of overdose at nearly four times the rate of those without one, after adjustment, and the risk rose with the benzodiazepine dose.

**Sarah:** That's still observational, though. People prescribed both drugs might have had more severe pain or mental illness to begin with.

**Kiffer:** Confounding is possible, yes, and that's where the experiment in volunteers helps. When alcohol was added to oxycodone under close monitoring, breathing fell further than with oxycodone alone, especially in the older volunteers. It used alcohol where the veterans study looked at benzodiazepines, but both converge on the same drive to breathe. The experiment shows the mechanism, the veterans study shows a dose-response pattern in real practice, and the Ontario data shows how often the combination appears among people who died. Each covers a gap in the others.

**Sarah:** And the Ontario figure matters for prescribing rules too. British Columbia's standard can limit what physicians co-prescribe, but it can't reach a combination that arrives already mixed in the unregulated supply.

**Kiffer:** That's right, and it's why the same evidence leads in two directions. For prescribers, it supports limiting co-prescribing while tapering slowly and keeping patients in care, which is what the British Columbia standard requires. For the unregulated supply, it raises questions about what's in the drugs themselves, which Lesson five takes up, and about the harm reduction services that HSCI 402 develops.

**Sarah:** Okay. Now the part where we disagree. In 2011, Canada's drinking guidelines set the low-risk limit at ten drinks a week for women and fifteen for men. In January 2023, Canada's Guidance on Alcohol and Health put low risk at one to two drinks a week. Its expert panel made an explicit decision about how much risk should count as low, and the line it drew comes down to one in a thousand.

**Kiffer:** Let me set out what the expert panel did. Advice about voluntary activities generally uses a lifetime risk of premature death of one in a thousand, while several countries have based alcohol guidelines on a higher risk of one in a hundred. The panel's modelling found that in Canada, one in a thousand corresponded to two standard drinks a week, and one in a hundred to six.

**Sarah:** Australia is an example of the one in a hundred approach. Its 2020 national guidelines say healthy adults should drink no more than ten standard drinks a week, which sounds far more generous than six, until you check what a standard drink is there.

**Kiffer:** That makes a nice worked example. Can you convert it into Canadian standard drinks?

**Sarah:** An Australian standard drink is ten grams of alcohol, so ten of them is a hundred grams. A Canadian standard drink is 13.45 grams. A hundred divided by 13.45 is about 7.4. So Australia's limit is about seven and a half Canadian drinks a week, compared with Canada's six at a similar level of risk.

**Kiffer:** So two panels, with their own models and data, landed less than one and a half Canadian drinks apart. The label of ten drinks hides that, because the drinks are different sizes. On the bigger choice, here's my view. I find the case for one in a thousand persuasive, because it's the level generally used for other voluntary risks, which holds alcohol to the same standard. And the continuum frames every drink as adding a little risk, so the message is useful wherever someone starts. A person at ten drinks a week learns that cutting to six lowers their risk.

**Sarah:** Consistency is a good argument. My worry is what the message does in practice. Someone who has one drink with dinner twice a week and two at a weekend gathering is at four drinks, which is moderate risk. If ordinary drinking sounds risky, some people may tune out the whole message, including the advice on more than two drinks per occasion. And the threshold is a value judgement. Experts can estimate the risk at each level of drinking, but how much risk is acceptable is a question the public should have a say in.

**Kiffer:** That's a fair point, and I think the guidance handles it reasonably well by stating its risk level openly. Anyone who prefers one in a hundred can see where that falls, because the boundaries of the continuum match the two risk levels. Where I'd agree with you is that the guidance only works if people can count standard drinks, which is why I think the panel's proposal to put standard-drink counts on containers is one of its most useful parts.

**Sarah:** I'm more doubtful about the threshold than Kiffer is, but we agree on the practical rule. Before you compare a drinking limit with any other, from another country or another decade, find the level of risk it was set to and the size of the standard drink it counts in.

**Kiffer:** Let's pull it together with three things to take away.

**Sarah:** First, once the liver enzyme is saturated, alcohol leaves the body at a roughly fixed pace, so a higher peak means a proportionally longer wait, and that pace varies from about ten to thirty-five an hour between people.

**Kiffer:** Second, before trusting an estimate, work out the range its inputs allow. The same four drinks gave estimates from about forty-seven to about ninety-six, so a blood alcohol estimate teaches how the body works and can't tell anyone whether a person is impaired or safe to drive.

**Sarah:** And third, a share counted only among people who died, or only among people who were harmed, can't measure risk by itself. You also need to know how common the exposure is among everyone at risk, survivors included, which is what the veterans study's random sample supplied, and what the nitrous oxide case reports lacked.

**Kiffer:** If you'd like more practice, rework the blood alcohol range problem from this episode for a hypothetical person weighing sixty kilograms, and check that your highest estimate sits above the central one.

**Sarah:** Next time, it's Lesson five, Opioids and the Toxic Unregulated Drug Supply, where the story of breathing from today continues at the opioid receptor, along with what naloxone can and can't reverse.

**Kiffer:** Take care, everyone.

**Sarah:** See you in Lesson five.
