How Medicines Actually Work

TL;DR. A drug is a molecule shaped to stick to one particular protein and change what it does. That is the whole idea. Everything else (why you take it twice a day, why it takes six weeks to work, why it upsets your stomach, why grapefruit is a problem, why the dose is lower for your grandmother) follows from two questions: what does the drug do to the body, and what does the body do to the drug. Side effects are not accidents or impurities. They are the same mechanism acting where you did not want it, and understanding that turns a scary leaflet into a predictable list.

Key takeaways

• Most drugs work at one of four target types: receptors, enzymes, ion channels, or transporters. Anti-infectives are the exception, aiming at structures the pathogen has and you do not.
• Half-life determines dosing frequency. It takes about 4 to 5 half-lives to reach steady state, which is why some drugs work in an hour and others take weeks.
• Side effects have exactly three sources: the same target in the wrong tissue, binding to an unintended target, and downstream consequences of the intended effect.
• The therapeutic index (the gap between the effective dose and the toxic dose) is why some drugs need blood monitoring and most do not.
• The liver's CYP450 enzyme family metabolises most drugs, and it is where the majority of drug interactions happen, grapefruit included.
• Biologics (the drugs ending in -mab) are engineered antibodies. They are large, must be injected, cost far more than pills, and can hit targets small molecules cannot.

What a drug is

In short: Nearly every drug binds one of four things: a receptor, an enzyme, an ion channel, or a transporter.

A drug is a small molecule (usually 200 to 600 daltons, small enough to cross membranes and be swallowed) whose shape fits into a pocket on a protein. When it binds, it either mimics the natural signal, blocks it, or changes the protein's activity. Almost every drug in this book is one of the following.

That last row is the entire basis of selective toxicity, the reason antibiotics are safe and chemotherapy is not. Bacteria differ from us enormously, so bacterial-only targets are easy to find. Cancer cells are our own cells with a few changes, so classical chemotherapy attacks anything dividing fast and hits hair, gut lining, and bone marrow as collateral. Fungi are eukaryotes like us, which is why antifungals are comparatively harsh and few.

What the body does to the drug: ADME

In short: Absorption, distribution, metabolism, and excretion decide how much drug arrives and for how long, and half-life is what sets the dosing schedule.

Four processes decide how much drug reaches the target and for how long.

Absorption. A swallowed tablet must dissolve, survive stomach acid, and cross the gut wall. Then, crucially, all blood from the gut goes to the liver first, which may destroy much of the dose before it ever reaches the circulation. This is first-pass metabolism, and it is why some drugs are useless as tablets and must be injected, inhaled, or placed under the tongue. Nitroglycerin for angina goes under the tongue for exactly this reason. Bioavailability is the fraction of a dose that makes it into the bloodstream: 100 percent by definition for intravenous, often 20 to 80 percent by mouth.

Distribution. The drug spreads into tissues. Fat-soluble drugs accumulate in fat and linger; water-soluble ones stay in blood and fluid. The blood-brain barrier excludes most large or water-loving molecules, which is why some antihistamines make you drowsy (they cross) and newer ones do not (they do not), and why treating brain infections and brain tumours is so much harder than treating them anywhere else.

Metabolism. Mostly the liver, mostly a family of enzymes called cytochrome P450 (CYP). Their job is to make foreign molecules more water-soluble so kidneys can excrete them. Two consequences follow. First, drugs that induce or inhibit a CYP enzyme change the blood levels of every other drug using that enzyme, which is where most serious interactions come from. Grapefruit juice inhibits CYP3A4 in the gut wall, which can raise levels of certain statins and immunosuppressants enough to matter. Second, people carry different CYP variants, so identical doses give different blood levels, which is the practical basis of pharmacogenomics.

Excretion. Mostly kidneys, some in bile and stool. This is why kidney function governs the dosing of a huge range of drugs, and why an older person with half the kidney function of a 30-year-old needs a smaller dose of the same medicine to reach the same blood level.

Half-life, and why the schedule is what it is

Half-life is the time for blood concentration to fall by half. It determines two things.

Dosing interval: a drug with a 4-hour half-life dosed once daily would spend most of the day below the useful level, so it is dosed three or four times a day or reformulated as a slow-release tablet.

Time to steady state: with regular dosing, levels rise and plateau after about 4 to 5 half-lives. That plateau is when the full effect appears.

This is the honest explanation for something patients are often told without reason: antidepressants "take a few weeks to work." Part of that is pharmacokinetics and part is slower biological adaptation, but the practical rule stands. Judging an antidepressant after four days is judging it before it has arrived.

What the drug does to the body: dose and response

Plot effect against dose and you get an S-shaped curve. Below a threshold, nothing. Then a steep region where small changes matter a lot. Then a plateau where more drug adds effect only in side effects. Doubling a dose on the plateau buys toxicity, not benefit, which is the reason maximum doses exist.

Plot the toxicity curve alongside it and the gap between the two is the therapeutic index. A wide gap (penicillin, most modern drugs) means dosing can be approximate. A narrow one means small errors are dangerous, and those drugs get blood level monitoring: warfarin, lithium, digoxin, phenytoin, gentamicin, and many chemotherapies.

Why side effects exist

In short: There are only three causes: the intended target in the wrong tissue, an unintended target, or a logical consequence of the effect you wanted.

There are exactly three reasons, and every leaflet entry is one of them.

1. The intended target, in an unintended tissue. Receptors are not confined to the organ you care about. Beta blockers slow the heart via beta-1 receptors, and also hit beta-2 receptors in the airways, so they can worsen asthma. Antihistamines block histamine in the nose and also in the brain, so old ones cause sedation. Anticholinergics used for bladder or Parkinson's symptoms dry the mouth, blur vision, and cloud thinking, because the same receptor type runs all of it.

2. Off-target binding. No molecule is perfectly selective. Many antipsychotics bind histamine and muscarinic receptors as well as the dopamine receptors they are aimed at, producing sedation and weight gain that have nothing to do with the therapeutic effect.

3. Downstream consequences of the intended effect. This is the most logical category and the easiest to predict. Aspirin blocks COX-1, which reduces pain and clot formation, and also reduces the prostaglandins that protect the stomach lining, hence ulcers and bleeding. SGLT2 inhibitors push glucose into the urine to lower blood sugar, and sugary urine promotes genital yeast infections. Immunosuppressants for rheumatoid arthritis quiet the immune attack on your joints and also on tuberculosis.

Don't be confused: a side effect and an allergy are not the same thing. Nausea from an antibiotic is a side effect: predictable, dose-related, usually manageable. A true drug allergy is an immune reaction (hives, swelling, wheeze, anaphylaxis), is not dose-related, and can be worse on re-exposure. The distinction is not pedantry. Being wrongly labelled "penicillin allergic" pushes patients toward broader, more toxic, less effective antibiotics for life, and most such labels turn out to be wrong when properly tested.

The other kinds of medicine

In short: Beyond small molecules there are engineered antibodies, peptides, vaccines, gene therapies, and living cells, and the -mab ending marks an antibody.

Not everything is a small molecule.

The -mab suffix is worth learning because it appears in nearly every modern treatment chapter: adalimumab for autoimmune disease, trastuzumab for HER2 breast cancer, pembrolizumab for cancer immunotherapy, omalizumab for severe asthma. Each is an antibody engineered to grip one human protein.

Tolerance, dependence, and stopping

Tolerance is needing more drug for the same effect, because the body adapts (for example by reducing receptor numbers). Physical dependence is the body having adjusted so that abrupt withdrawal causes symptoms. Neither is the same as addiction, which is compulsive use despite harm (Chapter 43). People taking opioids or benzodiazepines correctly for months develop dependence without addiction, and taper slowly for that reason.

Several drug classes are dangerous to stop abruptly for reasons that are not addiction at all. Beta blockers can cause a rebound surge in heart rate and blood pressure. Long term corticosteroids suppress the body's own cortisol production, so stopping suddenly can precipitate an adrenal crisis. Antiepileptics stopped abruptly can trigger seizures. The general rule is that if a drug has been taken daily for months, stopping is a plan, not a decision.

Placebo, nocebo, and why trials need controls

People improve after taking an inert pill: natural recovery, regression to the mean, expectation, and the ritual of care all contribute. Placebo response is largest for subjective outcomes (pain, mood, nausea, fatigue) and negligible for objective ones (tumour size, blood glucose, mortality). Nocebo is the mirror image: told about possible side effects, people report them, including in placebo groups. In statin trials, muscle aches are reported at nearly identical rates on statin and on placebo, which does not mean any individual's pain is imaginary, but does explain why the drug gets blamed more often than it deserves.

This is the entire reason Chapter 17 exists: with effects this large and this ordinary, no one can tell whether a treatment works by trying it on a few people and watching.

Two practical problems that dominate real prescribing

Adherence. Roughly half of patients on long-term medication do not take it as prescribed, and the rate is worst exactly where the disease is silent: blood pressure, cholesterol, early diabetes. Adherence beats drug choice as a determinant of outcome for these conditions. The fixes are unglamorous: fewer pills, once-daily dosing, combination tablets, no-cost supply, and an explanation of what the pill is for.

Polypharmacy. Treating five diseases by five guidelines gives an older patient ten or more medicines, with interactions no single guideline anticipated. The countermove is deprescribing: periodically reviewing whether each drug still has a purpose, and stopping what does not. Some of the clearest benefit in geriatric medicine comes from subtraction.

Sources and notes

Standard pharmacology (Rang and Dale, Pharmacology; Katzung, Basic and Clinical Pharmacology; Goodman & Gilman). Half-life values are typical adult figures and vary with age, kidney and liver function, and formulation. The 4 to 5 half-lives rule to steady state is a mathematical property of first-order elimination. Nocebo and statin muscle symptoms: the SAMSON n-of-1 trial (Howard et al., 2021, JACC) and blinded re-challenge studies. Penicillin allergy delabelling: multiple studies find that around 90 percent of patients labelled penicillin-allergic tolerate it on formal testing. Adherence estimates are WHO figures for chronic therapy in developed countries.

Open questions. How much of the placebo response is genuine physiology versus measurement artefact remains debated, as does the clinical value of routine pharmacogenomic testing outside specific drugs.

Next: how anyone knows whether any of this works. 👉