Antibiotics for Coughs and Colds

Healthcare · §605

The cough you brought to the clinic is almost certainly viral, and the five-day pill bottle does nothing to it. Across the four Cochrane reviews covering the common cold, acute bronchitis, sinusitis of under ten days, and undifferentiated sore throat, antibiotics produce no clinically meaningful symptom benefit — and adverse events at roughly the same magnitude as the benefit. What they do reliably hit is your gut, your one-year future as a carrier of drug-resistant bacteria, and the global resistance cascade now ranked among the leading causes of death worldwide. This entry covers when antibiotics genuinely belong in your hand, why most of the time they do not, and how to walk out of a clinic visit without the script you did not need.

Avoid · As-needed Evidence Strong Chapter Healthcare

The win here is what does not happen: the week of diarrhea or rash that comes with about one antibiotic course in five, the year your body spends carrying drug-resistant bacteria after each course, and the share of the global resistance problem that does not bend the wrong way because of you. The catch is social — saying no to a script in a fifteen-minute visit, accepting that a cold takes a week and a cough takes two. The evidence is high and mainstream medicine is aligned; the friction lives in the waiting room, not in the journals.

A cold is a virus. So is most of what gets called a sinus infection, the chest cough that comes with one, and the great majority of sore throats. Rhinovirus, coronavirus, RSV, parainfluenza, influenza, adenovirus, human metapneumovirus — these are what get into your nose, throat, sinuses, and bronchi when you catch "a bug." Antibiotics are drugs designed to kill bacteria. They have no mechanism — none — against viruses. Amoxicillin breaks bacterial cell walls; bacteria have cell walls, viruses do not, end of conversation. Azithromycin stops protein synthesis on bacterial ribosomes; viruses use your own cells' ribosomes, which the drug cannot touch. A drug aimed at a category of organism that is not present cannot help, no matter how miserable the symptoms feel.

The exceptions are a short list and they matter. Group A streptococcus accounts for roughly 5–15% of adult sore throats and 20–30% of those in children; a rapid antigen swab in the clinic confirms it in a few minutes Shulman 2012. Acute bacterial sinusitis is a small fraction of sinus presentations, and it has a clinical signature: symptoms persisting ten or more days without improvement, severe symptoms at onset (high fever with purulent discharge and face pain), or "double-sickening" — clear improvement followed by clear worsening within ten days Rosenfeld 2015. Pneumonia, whooping cough, and bacterial exacerbations of chronic lung disease are their own categories with their own decisions. Everything else — the standard cold-cough-snot-throat that fills clinic schedules through winter — is viral.

The colour of your snot is not a sorting tool. Yellow or green mucus comes from myeloperoxidase, an enzyme released by the neutrophils your immune system sends against anything inflammatory. Viral infections turn snot green as readily as bacterial ones. The pattern over time — how long the symptoms last, whether they came back worse after improving — is what tells the two apart, not the colour Harris 2016.

What the trials actually show

Four large independent reviews cover the cluster. They were each done by the same body — Cochrane, the organisation widely treated as the gold standard for synthesising clinical trials — and they say the same thing four different ways.

For the common cold, eleven trials in 1,047 people: no statistically significant difference in how long symptoms last between antibiotics and placebo, with adverse events in adults roughly two-and-a-half times more common on the drug Kenealy 2013.

For acute bronchitis — the cough that comes with or after a cold — seventeen trials in 5,099 people: about half a day less cough on average, and a 20% higher rate of adverse events. The authors' bottom line: "limited evidence of clinical benefit" Smith 2017.

For acute sinusitis in adults, fifteen trials in 3,057 people: eighteen people have to be treated for one extra clinical cure that can be credited to the antibiotic, against eight treated to produce one extra adverse event. Most untreated adults recover on their own within the same window Lemiengre 2018.

For sore throat, twenty-nine trials in 15,337 people (mostly children): an average of about sixteen hours less sore throat at day three across all participants, with the actual benefit concentrated entirely in the minority with confirmed strep. Side-effect rates ran at the same order of magnitude as the symptom benefit Spinks 2021.

Sixteen hours sounds like something. It is not. It is the difference between the sore throat ending Thursday afternoon and Thursday evening. And it is averaged across people who include those without strep at all — for whom the drug did precisely nothing for the throat, and only contributed to the side-effect column.

The official position of mainstream medicine matches the evidence. The American College of Physicians and the CDC jointly recommend against antibiotic prescribing for the common cold, acute uncomplicated bronchitis, and uncomplicated acute sinusitis at first presentation; and against empirical sore-throat prescribing without a positive rapid strep test or culture Harris 2016. The Infectious Diseases Society of America's pharyngitis guideline says the same Shulman 2012. NICE in the UK says the same. The American Academy of Otolaryngology – Head and Neck Surgery says the same for sinusitis Rosenfeld 2015.

Despite all of that, roughly a third of US outpatient antibiotic prescriptions are unnecessary, and the dominant category driving the unnecessary fraction is acute respiratory infection Fleming-Dutra 2016. The recommendation and the prescribing pad are in two different rooms.

What most people (and a lot of patients) get wrong

"Green snot means I need antibiotics." The colour comes from myeloperoxidase in your own white blood cells and is present in viral infections at the same intensity as bacterial. Pattern over time discriminates; colour does not Harris 2016.
"I got better right after starting the amoxicillin — proof it worked." Colds peak at day three to five and resolve by day seven to ten; coughs commonly hang on two to three weeks. Anyone who starts a placebo on day three of a cold will reliably feel better within forty-eight hours. That is the cold ending, not the drug starting.
"A five-day course is harmless." A single course rearranges your gut microbiome, with measurable changes for months and incomplete recovery in many adults; about a third of bacterial taxa fail to return to baseline six months later Dethlefsen 2011. In children, lifetime cumulative antibiotic exposure tracks with reduced microbial diversity and downstream metabolic and allergic patterns Korpela 2016.
"Antibiotics keep a cold from turning into pneumonia." For uncomplicated upper-respiratory illness in healthy adults and children, this is not what the trial data show. The reviews above tested for exactly that effect; the absolute reduction in complications is negligible Smith 2017; Kenealy 2013.
"If the doctor wrote the script, I must need it." Roughly a third of US outpatient antibiotic prescriptions are unnecessary Fleming-Dutra 2016, and prescribing rates rise across a clinic session as the day goes on — a decision-fatigue signal that has nothing to do with whether the next patient has a bacterial infection Linder 2014. The script and the indication are not as tightly coupled as the visit feels.

What three winter colds a year — with an antibiotic each — looks like ten years out

The first one is uneventful. You take the five-day course of amoxicillin for the sinus pressure that wasn't bacterial, you feel better on roughly the same timeline you would have felt better anyway, and the only thing you really notice is a few days of looser stools you put down to being run-down. The next time you have a real bacterial infection — a few years out, a kidney infection, a wound that gets angry — your first-line drug works fine, and you do not connect any of it to this course.

By the third or fourth course, the gut starts to talk back. The diarrhea during the drug stops being a surprise. A thrush flare in the week after, or a yeast infection if you are a woman; a couple of weeks where food doesn't sit right. The relationship between "taking a course" and "feeling off for a month afterwards" becomes hard to miss. The bacteria that have spent generations evolving to balance the populations that cause trouble are not back to baseline by the time the next bottle gets opened.

By the end of the decade, the bill for the cumulative pattern is harder to see but no less real. Your gut microbial diversity is lower than it was. You are more likely to be carrying a resistant strain of E. coli or Staphylococcus aureus for about a year after each exposure — the meta-analysis put the effect at twelve months Costelloe 2010. When you do get a real bacterial infection — a kidney infection in your forties, a serious chest infection in your sixties — the front-line drug fails more often than it would have for the version of you that declined the cold-courses, and the second-line drug is harsher, longer, and more expensive.

If you are a parent, the bill arrives faster, because the developing gut is more susceptible to lasting reorganisation than the adult one Korpela 2016. The three courses of amoxicillin your toddler took for what were always going to be viruses leave a microbiome that does not fully reset by the time she starts primary school. The asthma signal, the allergy signal, the BMI signal in the observational data — none of these are guaranteed for any individual child, but the dose-response curve is real, and the part of it that is avoidable was avoidable.

And the part that does not belong to any one person. Antimicrobial resistance, taken globally, was directly responsible for 1.27 million deaths in 2019, with resistance playing a role in roughly another 3.7 million — more than HIV and malaria combined Murray 2022; GBD AMR 2022. No single course of doxycycline you took for a sinus infection in 2018 caused any of those deaths. Roughly a third of US outpatient prescribing being unnecessary, cumulated across decades and continents, did Fleming-Dutra 2016; CDC 2019.

When antibiotics actually do belong, and how to leave the visit without one when they do not

The short list of times an antibiotic is the right answer for a respiratory presentation:

Strep throat confirmed by a rapid in-clinic swab test or throat culture, in a patient with the right clinical picture: sudden sore throat, fever, swollen and tender front-of-neck lymph nodes, no cough Shulman 2012; McIsaac 1998.
Sinus illness where symptoms have persisted ten or more days without any improvement, or where they got better and then clearly worse again within ten days ("double-sickening"), or where the onset was unusually severe (high fever with face pain and purulent discharge from the start) Rosenfeld 2015; Lemiengre 2018.
Suspected pneumonia — focal exam findings, low oxygen, abnormal vital signs.
Whooping cough (pertussis) in or near a household with infants or pregnant women.
People with significant immunocompromise (chemotherapy, transplant, advanced HIV, biologic-treated autoimmune disease) or established structural lung disease (COPD, bronchiectasis, cystic fibrosis) — these are not the readers this entry asks to push back; their decisions belong to the clinician who knows the rest of the picture.

Outside that list, the action is to ask for and accept the symptomatic plan: hydration, paracetamol or ibuprofen for fever and pain, saline nasal rinses for sinus pressure, honey for cough in anyone over a year old, lozenges and warm fluids for the throat, sleep. The cold takes a week. The cough hangs on for two to three. The sore throat without strep gets better within a few days. None of those timelines are improved by amoxicillin.

Where this goes wrong in practice

Confusing the natural recovery with the drug. Most people feel better within forty-eight hours of starting an antibiotic for a cold because the cold was going to peak and turn the corner on that day anyway. The improvement and the drug are decoupled; the brain refuses to believe it until it has watched the same pattern unfold three or four times.
Asking, and getting. The strongest predictor of an unnecessary prescription in pediatric respiratory illness is the clinician's read of whether the parent expects one; explicitly hinting at it roughly quadruples the odds of an inappropriate script Mangione-Smith 2015. The same dynamic operates in adult primary care. The visit feels like an exchange; the doctor reaches for the thing that ends the exchange.
The end-of-session script. Antibiotic prescribing for acute respiratory infection rises across a clinic block, peaking at the end of each morning and afternoon — the same doctor with the same chief complaint prescribes more often at five o'clock than at nine, with no change in case mix Linder 2014. If you can choose, book early in the day.
Reaching for the bigger gun. When a clinician does prescribe for a respiratory infection that did not need any antibiotic, picking azithromycin or a fluoroquinolone over a narrow-spectrum agent compounds the damage: broader spectrum means broader gut destruction and broader resistance selection — for no extra clinical gain. Azithromycin specifically carries an excess sudden-cardiac-death signal during treatment days Ray 2012; for a five-day course in a higher-risk patient, that becomes a non-trivial trade.
Generalising the "avoid" stance too far. The argument here is about cold, cough, sinus, and undifferentiated sore throat in otherwise healthy people. A real bacterial infection — confirmed strep, an actual pneumonia, a kidney infection, a serious cellulitis — is a different conversation, with antibiotics as the right answer.

What you get back when you stop reaching for the script

Within the same week, the most concrete thing: no diarrhea, no rash, no thrush, no yeast infection that was not going to happen if you had not taken the drug. Across the Cochrane reviews, somewhere between roughly one in five and one in eight patients on a course of antibiotics for these indications develops a notable side effect Lemiengre 2018; Kenealy 2013. If you take three URI antibiotic courses across the next decade, the odds are that at least one of them hits you with something visible. That visible something does not happen.

Across the year that follows, your gut is yours. The taxa that took a hit during your last course are not now taking another. The carriage of a drug-resistant strain in your nose, gut, or skin that comes with each exposure — the meta-analysis ran the effect out to about twelve months — is not running in the background Costelloe 2010. The risk of Clostridioides difficile infection in the weeks after a course (community-associated CDI is roughly three to four times more likely after recent antibiotic exposure) goes back to baseline Deshpande 2013. When you do get a real bacterial infection at some point in the next decade, the front-line drug is more likely to work the first time, at a normal dose, for a shorter course.

Across the longer arc, the microbiome you hand to whatever ages with you — the immune calibration that lives in those bacteria, the metabolic balance that depends on them — is not getting reset on cold-season cadence. The version of you at sixty whose gut diversity has been preserved through a few dozen winters of declined-amoxicillins is observably different from the version who took one for each. None of this is felt in any particular afternoon. All of it shows up later as the absence of something you do not have to know about to benefit from.

And the part nobody feels directly: the share of the global resistance cascade that does not bend the wrong way because of you. It is a probabilistic gift to people you will never meet — the woman whose post-surgical infection responds to a first-line antibiotic in 2040 because the population reservoir of resistance is one course smaller than it would otherwise have been. The arithmetic only works at scale; the action only works one decision at a time.

Adjacent topics worth knowing about: pneumonia (the bacterial respiratory infection that genuinely needs antibiotics, with its own decision tree); antibiotic use in livestock and food production (a larger contributor to the global resistance pool than human outpatient prescribing in many countries); pediatric ear infections (their own age-dependent watchful-waiting decision); urinary tract, skin, and sexually transmitted infections (different infections, different rules, often genuinely bacterial); and gut microbiome restoration (what a damaged gut can and cannot do to recover on its own). For the seasonal-cold version of the same question, see also the entries on chronic allergic rhinitis, indoor air quality, and the broader pattern of polypharmacy in older adults.

Substance and claimed effects

The substance covered by this entry is the act of taking — or requesting — an antibiotic course for an acute upper-respiratory illness that is, in the great majority of presentations, viral: the common cold (acute viral rhinosinusitis), acute bronchitis (the "chest cold" / acute cough illness in otherwise healthy adults), uncomplicated short-duration acute sinusitis (less than ~10 days, no double-sickening), and most undifferentiated sore throats. Antibiotics are bactericidal or bacteriostatic agents (β-lactams, macrolides, tetracyclines, fluoroquinolones, others) with no mechanism of action against respiratory viruses (rhinovirus, coronavirus, RSV, influenza, parainfluenza, adenovirus, human metapneumovirus). Across these indications the catalogue evaluates four sets of claims: (1) symptom-duration / complication-prevention benefit at the individual encounter; (2) population-level antimicrobial resistance (AMR) contribution and individual carriage of resistant organisms after exposure; (3) gut-microbiome disruption and its sequelae; (4) common adverse drug effects (gastrointestinal, allergic, cardiac, neuromusculoskeletal). The entry is keyed avoid: the reader's decision is whether to take, request, or pressure for an antibiotic in the common viral URI presentation, not whether to take one when a clinician identifies a clear bacterial indication (group A streptococcal pharyngitis with positive RADT/culture, community-acquired pneumonia, pertussis, complicated or prolonged bacterial rhinosinusitis). At least 30% of US outpatient antibiotic prescriptions are unnecessary, the largest single contributor being acute respiratory conditions Fleming-Dutra 2016.

Evidence by addressing question

Mechanism

Viral and bacterial respiratory infections share a clinical phenotype (rhinorrhea, cough, sore throat, fever, malaise, purulent secretions) but diverge entirely in their susceptibility to antibacterial drugs. Rhinovirus accounts for the majority of common-cold presentations, with coronavirus, RSV, influenza, parainfluenza, adenovirus, and human metapneumovirus making up most of the rest. In acute bronchitis in immunocompetent adults, viral pathogens are identified in the great majority of cases where a pathogen can be identified at all; bacterial etiology (e.g., Mycoplasma pneumoniae, Bordetella pertussis, Chlamydia pneumoniae) accounts for less than ~10%, and routine antibacterial therapy targets none of them effectively in the typical presentation Smith 2017; Harris 2016. Acute rhinosinusitis is overwhelmingly viral; bacterial superinfection complicates roughly 0.5–2% of viral upper-respiratory infections and is suggested clinically by symptom duration ≥ 10 days without improvement, severe symptoms (fever ≥39°C with purulent nasal discharge or facial pain ≥ 3–4 days at illness onset), or "double-sickening" — initial improvement followed by worsening within 10 days Rosenfeld 2015; Lemiengre 2018. Group A β-haemolytic streptococcus (GAS) is identified in approximately 5–15% of adult sore throats and 20–30% of pediatric sore throats; the remainder are viral or, less commonly, due to other bacteria (group C/G strep, Fusobacterium necrophorum) Shulman 2012. Yellow, green, or purulent nasal discharge reflects neutrophil myeloperoxidase activity and is not a discriminator between viral and bacterial illness Harris 2016; Rosenfeld 2015.

Evidence

Four Cochrane systematic reviews — the most rigorous independent syntheses available — converge on the same conclusion: antibiotics for viral URI either fail to shorten symptoms, shorten them by clinically insignificant margins, or trade a small symptom benefit against a similar-magnitude adverse-event burden.

Common cold. Eleven trials (n = 1,047 participants with rhinopharyngitis or purulent rhinitis): antibiotics produced no statistically significant difference in persistence of general symptoms versus placebo (risk ratio 0.95) in adults or in children. Adverse events were significantly more common in the antibiotic arm in adults (RR 2.62) Kenealy 2013.
Acute bronchitis. Seventeen trials, n = 5,099. Antibiotics conferred a small reduction in cough at follow-up (RR 0.91) and shortened symptomatic time by about half a day on average; adverse events were significantly increased (RR 1.20). Authors conclude there is "limited evidence of clinical benefit to support the use of antibiotics in acute bronchitis" Smith 2017.
Acute rhinosinusitis (adults). Fifteen trials, n = 3,057. Pooled cure at 7–14 days favoured antibiotics modestly (OR 1.25), corresponding to NNTB ≈ 18 — eighteen people treated for one additional clinical cure attributable to the drug — against NNTH ≈ 8 for any adverse event. Most untreated adults are cured or improved without antibiotics. Authors note benefit is modest and harm is common; clinicians should weigh both Lemiengre 2018.
Sore throat. Twenty-nine trials, n = 15,337 (mostly children). Antibiotics reduce symptom duration by approximately 16 hours at day three across all-comers; the absolute benefit is concentrated in GAS-positive patients. Acute rheumatic fever — historically the principal complication-prevention rationale — has fallen to <1 per 100,000 in high-income settings, reducing the modern absolute benefit of empiric treatment further. Peritonsillar abscess and otitis media incidence are reduced in pooled trials, but absolute risks were very low. Side-effect rates were the same magnitude as the symptom benefit Spinks 2021.

Clinical practice guidance is aligned. The American College of Physicians and the CDC jointly recommend against antibiotic prescribing for the common cold, acute uncomplicated bronchitis, and uncomplicated acute rhinosinusitis at presentation, and recommend testing-driven prescribing (rapid antigen detection test / throat culture using a McIsaac or Centor score of ≥2–3) for GAS pharyngitis Harris 2016; Shulman 2012; McIsaac 1998. NICE and the Otolaryngology–Head and Neck Surgery Foundation reach the same operational conclusions Rosenfeld 2015. Despite this convergence, US ambulatory data through 2011 found ≥30% of outpatient antibiotic prescriptions to be inappropriate, with acute respiratory conditions as the predominant clinical category driving the misuse Fleming-Dutra 2016. UK general-practice data show similar overall volumes and substantial inter-practice variability driven by patient demand, time pressure, and consultation length rather than clinical mix Palin 2019. US pediatric prescribing has fallen since the 1990s but remains heavy: antibiotics accounted for roughly one in five pediatric outpatient visits in the late 2000s, with broad-spectrum agents prescribed for largely viral conditions Hersh 2013.

Protocol

The action is structural avoidance with two operational components: (a) understanding when an antibiotic is and is not warranted, and (b) navigating the clinical encounter to receive non-antibiotic guidance without leaving worse off. Trigger criteria for legitimate antibiotic consideration (per ACP/CDC, IDSA, AAO-HNSF): GAS pharyngitis confirmed by RADT or culture in a patient with McIsaac score ≥2 (fever >38°C, absence of cough, tender anterior cervical lymphadenopathy, tonsillar exudate, age 3–14 contributes; age >45 subtracts) Shulman 2012; McIsaac 1998; acute bacterial rhinosinusitis defined by symptoms ≥ 10 days without improvement, or severe symptoms at presentation, or double-sickening Rosenfeld 2015; Lemiengre 2018; suspected community-acquired pneumonia (focal exam findings, hypoxia, abnormal vitals); pertussis (paroxysmal cough > 2 weeks with whoop or post-tussive emesis); patients with significant immunocompromise, structural lung disease (COPD exacerbation with purulent sputum), or other comorbidity raising the bacterial pre-test probability. The "watchful waiting" / delayed-prescription protocol — the patient is given a script but instructed to fill it only if symptoms persist or worsen at day 3–5 — reduces actual antibiotic consumption by ~50% across trials without increasing complications and is now mainstream practice in UK and northern European guidelines.

Symptomatic care has its own evidence base. Acetaminophen and NSAIDs reduce fever and pain; saline nasal irrigation reduces sinus symptom severity in adults and is recommended as adjunctive therapy Rosenfeld 2015; honey (≥ 1 year of age) modestly reduces cough frequency in children compared to placebo or no treatment; throat lozenges, hydration, and rest cover most of the remaining ground. Mucolytics (acetylcysteine, carbocysteine) show modest benefit in pediatric ARI with no safety signal in children >2 years Chalumeau 2013. Oseltamivir or baloxavir within 48 hours of symptom onset for laboratory-confirmed influenza is a distinct decision and not addressed by this entry's scope.

Contraindications

There is no contraindication to the action itself (declining or deferring an antibiotic prescription for a non-bacterial URI). Where the action would be unsafe is in cases where genuine bacterial infection is plausible and the patient is high-risk: immunocompromised hosts, structural lung disease with purulent exacerbation, suspected pneumonia, suspected GAS pharyngitis with positive testing in a patient with prior acute rheumatic fever, suspected pertussis exposure in a pregnant or infant household. In those settings, withholding antibiotics is medically wrong; the article's framing must not encourage refusal of clinician-determined bacterial-indication therapy.

Misconceptions

(1) "Coloured mucus = bacterial infection." Neutrophil myeloperoxidase turns nasal/sputum discharge green or yellow in viral infections as readily as bacterial ones; the colour is a feature of any inflammatory response, not a diagnostic discriminator Harris 2016. (2) "A 5-day course is harmless." A single course induces gut-microbial perturbations lasting weeks to years, with incomplete recovery in many adults and individualised long-term shifts Dethlefsen 2011; Korpela 2016. (3) "Antibiotics prevent the cold turning into something worse." For uncomplicated URI in immunocompetent patients, antibiotics do not reduce progression to pneumonia or other bacterial superinfection at clinically meaningful absolute rates Smith 2017; Kenealy 2013. (4) "I got better right after starting the amoxicillin — proof it worked." Viral URIs naturally peak at day 3–5 and resolve by day 7–10; cough commonly persists 2–3 weeks. Regression to the natural course, not the drug, accounts for most felt benefit. (5) "If the doctor prescribed it, it must be necessary." 30%+ of US outpatient antibiotic Rx are unnecessary, and time-of-day analyses show prescribing rises across a clinic session — a decision-fatigue signal independent of clinical indication Fleming-Dutra 2016; Linder 2014.

Audience

The action applies across the adult lifespan and to parents of children. Pediatric exposure is the highest-impact subgroup for two reasons: pediatric outpatient prescribing volume is large (≈20% of pediatric visits include an antibiotic; ARI is the dominant indication) Hersh 2013, and the developing microbiome is more susceptible to lasting disturbance Korpela 2016. Older adults face higher absolute risk for Clostridioides difficile infection and for QT-related cardiovascular events from macrolides and fluoroquinolones. Immunocompromised and structurally vulnerable patients are out of scope for the "avoid" action — antibiotic decisions in those populations belong to the clinician.

Alternatives

Symptomatic care (hydration, rest, analgesics/antipyretics, nasal saline, honey-based cough relief in children > 1 year, throat lozenges) is the dominant alternative for the common viral URI. "Delayed prescribing" — receiving a script that is held in reserve and filled only if criteria are met at day 3–5 — is a hybrid approach with strong supporting evidence for both reducing total antibiotic exposure and preserving patient autonomy and reassurance. Patient-centred communication strategies that explicitly name the viral diagnosis and provide a contingency plan reduce prescribing without reducing visit satisfaction Mangione-Smith 2015.

Failure modes

(1) Conflating the recovery curve with the drug. URIs follow a roughly self-limited 7–14-day course; people who start antibiotics on day 3 will reliably feel better in 2–4 days regardless of whether the drug did anything. (2) Patient-driven prescribing pressure. The strongest predictor of antibiotic prescribing in pediatric ARI is the clinician's perception that the parent expects one; explicit expectation suggestions ("non-recommended treatment suggestions") quadruple the odds of an inappropriate prescription Mangione-Smith 2015. (3) Decision fatigue. Within a single primary-care session, antibiotic prescribing for ARI rises through the morning and again through the afternoon, peaking at the end of each block — a pattern consistent with cognitive depletion rather than changing case mix Linder 2014. (4) Broad-spectrum substitution for an indication that does not need any antibiotic at all (e.g., azithromycin for "bronchitis") — adds disproportionate AMR pressure for no clinical gain. (5) Failing to complete a legitimately needed course (e.g., true GAS pharyngitis) on the assumption that the catalogue's "avoid" stance applies universally.

Practicalities

The financial cost of declining an antibiotic is negative (the patient does not buy the drug). The non-financial costs are social — clinician interaction friction, family pressure, time off work that the patient hoped a "real" treatment would shorten. The asymmetry between effort to ask and effort to push back creates the persistent prescribing volume documented across systems. UK and northern European systems with longer consultation times and structured delayed-prescription pathways achieve lower per-capita prescribing for the same disease burden Palin 2019.

Stakes

The "stakes" of taking antibiotics for viral URI are split across an individual axis and a population axis.

Individual adverse effects. Across Cochrane reviews, adverse events occur at rates roughly NNH ≈ 5–25 depending on drug and indication (diarrhea, nausea, rash, vaginal candidiasis, headache) Kenealy 2013; Lemiengre 2018. Anaphylaxis is rare (~1–5 per 10,000 doses for β-lactams) but lethal in a fraction of those. Severe cutaneous adverse reactions (Stevens-Johnson syndrome, toxic epidermal necrolysis) are rarer still but documented across most major classes.
Cardiovascular. Five-day courses of azithromycin (one of the most prescribed agents for ARI) were associated with an excess of ~47 sudden cardiovascular deaths per million courses versus amoxicillin during treatment days in a large Tennessee Medicaid cohort, concentrated in higher cardiovascular risk patients (HR 2.49 vs amoxicillin) Ray 2012. The signal led the FDA to update azithromycin labelling.
C. difficile. Community-associated C. difficile infection is strongly associated with prior outpatient antibiotic use, OR ≈ 3.55 in meta-analysis, with the highest signal for clindamycin, fluoroquinolones, and cephalosporins Deshpande 2013. CDI carries 5–10% mortality in older adults.
Individual carriage of resistant organisms. A single course of antibiotics in primary care increases the individual's probability of harboring resistant bacteria for up to 12 months after exposure, with the effect largest in the first month and tapering thereafter Costelloe 2010. The reader-relevant version: the next bacterial infection the patient does have is more likely to be drug-resistant — at the patient level, not just the population level.
Gut microbiome. Repeated ciprofloxacin courses produced incomplete and individualised recovery of the gut microbiota; ~1/3 of taxa failed to return to baseline at 6 months post-exposure Dethlefsen 2011. In a population of Finnish pre-school children, lifetime antibiotic use was associated with reduced microbial diversity, altered composition (depletion of Bifidobacterium and Lactobacillus; expansion of selected proteobacteria), and persistent shifts associated with increased BMI and asthma Korpela 2016. Early-life exposure has the largest long-term consequence; adult exposure recovers more reliably but not fully.
Population AMR. CDC estimates 2.8 million antibiotic-resistant infections and ≈35,000 directly attributable deaths annually in the US CDC 2019. Globally, the Lancet AMR analysis estimated 1.27 million deaths directly attributable to bacterial AMR in 2019, with 4.95 million deaths associated — placing AMR alongside the leading global causes of death and exceeding HIV/AIDS and malaria combined Murray 2022; GBD AMR 2022. The relevant individual contribution of any single inappropriate outpatient course is infinitesimal; the cumulative contribution of 30%+ of outpatient antibiotic Rx being unnecessary is structural.

Payoff

The payoff of declining inappropriate antibiotics is the inverse of the stakes list — avoided GI adverse events at NNH 5–25, no exposure to azithromycin's cardiovascular signal, no increased CDI risk, no 12-month carriage of resistant organisms, no gut-microbiome perturbation, and no contribution to the AMR cascade. The benefits accrue at three timescales: within days (no diarrhea/nausea/rash from a course not taken), within months (microbiome integrity preserved, resistant-organism carriage avoided), and over decades (cumulative gut and resistance debt lower for the individual and their household). Symptom resolution is unchanged by declining — the natural course of the viral illness is unaffected by the absence of an inactive drug.

History

Penicillin entered clinical use in the 1940s and broad-spectrum agents through the 1950s–60s. Outpatient antibiotic prescribing for URI rose through the 1970s–90s, with US per-capita rates peaking in the early 1990s. The CDC launched the "Get Smart: Know When Antibiotics Work" campaign in 1995 (rebranded "Be Antibiotics Aware" in 2018) to address public and prescriber demand drivers. Pediatric prescribing has fallen substantially since 2000, but adult ARI prescribing has been more stable, and the COVID-19 pandemic produced a transient bolus of azithromycin and other ARI prescribing despite lack of antibacterial indication.

Out-of-scope

Antibiotics for genuinely bacterial illness (confirmed GAS pharyngitis, community-acquired pneumonia, pertussis, complicated or prolonged bacterial sinusitis), antibiotic decisions for urinary-tract infections, skin and soft-tissue infections, sexually transmitted infections, dental infections, perioperative prophylaxis, and antibiotic use in livestock and aquaculture are all outside the scope of this entry. Pediatric acute otitis media has its own decision tree (immediate treatment vs. observation by age and severity) and is a candidate for a separate entry.

The credibility range

Optimist case

Antibiotics are among the most effective drugs ever developed; for true bacterial infection they are lifesaving. Within the URI cluster, several edge cases plausibly benefit: confirmed GAS pharyngitis where symptom reduction of ~16 hours is real (and transmission to susceptible contacts is reduced); acute bacterial sinusitis with double-sickening, where Cochrane NNTB ≈ 18 is modest but non-zero; high-risk patients (immunocompromised, structural lung disease) where empirical coverage may be justified; settings where rheumatic-heart-disease incidence remains high. Clinicians may also argue that population-level guidance under-weights the rare patient with masked bacterial infection, and that watchful waiting carries its own risk of complication. The pro-prescribing position is not absurd; it is misapplied at the population scale.

Skeptic case

For the typical adult or pediatric patient presenting with a cold, cough, sinus pressure of less than ~10 days, or undifferentiated sore throat, the Cochrane data are unambiguous: no clinically meaningful benefit on symptom duration, no demonstrated reduction in complications, harm rates of similar magnitude to any small symptom benefit. The mechanism case (no antibacterial drug works against a virus) is airtight. The population evidence on AMR — CDC, GBD AMR, Lancet — places resistance among the leading global health burdens, with outpatient ARI prescribing as the largest single contributor to the unnecessary fraction. The individual evidence on personal carriage (Costelloe 2010) and on gut microbiome (Dethlefsen, Korpela) closes the "but my course is harmless" defense. The driver of inappropriate prescribing is documented to be patient demand and clinician decision fatigue, not clinical judgment in disguise (Linder 2014; Fleming-Dutra 2016).

Author's call

The catalogue lands solidly on the skeptic side for the cold/cough/sinus/sore-throat cluster: for the vast majority of presentations, antibiotics are unhelpful at the individual level and harmful at the population level, and the action the reader should take is to know when antibiotics are warranted, accept symptomatic care for the rest, and refuse social or self-imposed pressure to consume them otherwise. Edge cases (GAS-positive pharyngitis, bacterial sinusitis criteria met, pneumonia suspected, immunocompromise) remain firmly in the clinician's hands; the article must not conflate the population-level "don't" with the clinical-indication "do." Evidence is high (Cochrane × 4 + ACP/CDC + IDSA + AAO-HNSF + Lancet AMR); controversy is low at the level of mainstream medicine, though prescribing behavior remains contested by patient demand and time pressure in practice.

Stakeholder and incentive map

Patients / parents. Want symptom relief, return to work or school, reassurance, and "real treatment." The strongest driver of unnecessary prescribing in pediatric ARI is parental expectation as perceived by the clinician Mangione-Smith 2015.
Clinicians. Face short consultation times, satisfaction-tied performance metrics, decision fatigue across the workday Linder 2014, and diagnostic uncertainty without rapid bacterial confirmation. The path of least resistance during a 15-minute visit is often the prescription pad.
Pharma. Most outpatient antibiotics are off-patent and low-margin; direct commercial promotion is minor compared with prescription stimulant or anti-inflammatory categories. There is no significant industry counter-incentive to stewardship.
Public-health bodies. CDC, WHO, NICE, IDSA, ACP, AAO-HNSF, AAP all push stewardship; the "Be Antibiotics Aware" / antimicrobial-stewardship infrastructure is well-funded but consumer awareness lags clinician awareness.
Pharmacists. Increasingly positioned as stewardship gatekeepers in community settings (UK, Australia); their decision support to clinicians and direct patient education are an under-leveraged channel.
Veterinary and agricultural sector. Account for a majority of total antibiotic tonnage in many countries; out of scope for this entry but materially influence the AMR ecosystem.

Population variability

Highest-stakes subgroup: children, where developing microbiome exposure has the largest lifetime trajectory effect and prescribing volumes are highest Hersh 2013; Korpela 2016. Older adults are at elevated absolute risk for CDI mortality, QT-related macrolide and fluoroquinolone cardiac events Ray 2012, and drug-drug interactions (warfarin/macrolides, statins/macrolides). Immunocompromised patients (chemotherapy, transplant, advanced HIV, biologic-treated autoimmune disease) are an explicit exception — empiric broader-spectrum coverage may be appropriate even when bacterial infection is uncertain. Patients with structural lung disease (COPD, bronchiectasis, cystic fibrosis) have an elevated bacterial pretest probability for any acute respiratory exacerbation and follow different guidelines. Geographic variability in prescribing is substantial: US Southeast prescribes ~2× the US Northwest per capita; intra-country UK general-practice variability is similar in magnitude Palin 2019. Patients with prior anaphylaxis, severe cutaneous adverse reactions, or tendinopathy from fluoroquinolones carry class-specific contraindications independent of indication.

Knowledge gaps

Causal certainty linking early-life antibiotic exposure to later-life asthma, atopy, IBD, and obesity remains observational; randomised data are ethically and practically unobtainable at scale, and confounding by indication is hard to rule out. Long-term randomised trials of restrictive versus permissive prescribing strategies at the population level are challenging; the field relies on quasi-experimental data (regional stewardship programs) and modeling. Patient-clinician communication interventions reduce prescribing in trials but scaling them to short-visit primary care has been uneven. Whether the precise within-class differences in microbiome impact (narrow vs broad spectrum, β-lactam vs macrolide vs fluoroquinolone) translate to differential long-term outcomes in healthy adults is an active research area. The reader-facing question — "is the one course I'm contemplating going to harm me individually?" — does not have an individual-level answer; the answer is statistical at the cohort level.

Scope vs brief. The brief named viral-vs-bacterial cause, symptom-duration effect, AMR, gut microbiome / side-effect harms, and prescribing pressures. All five are covered. Symptom-duration sits in evidence; viral-vs-bacterial in mechanism; AMR and microbiome harms split between stakes and payoff; prescribing pressures in failure-modes (patient demand, decision fatigue) and the protocol callout (delayed prescription). No part of the brief was dropped.

Action choice — avoid, not know. The natural read of the brief is heads-up about a common mistake; the operational read is "don't take, don't request." Avoid is the right verb because there is a decision the reader makes, repeatedly, with a defensible default. Know would have left the article without teeth.

Score calibration notes.

longevity = 2. The individual mortality reduction from declining one cold-course is statistically minute; what earns 2 is the cumulative individual carriage signal (Costelloe 2010), C. difficile risk reduction (Deshpande 2013), the azithromycin sudden-cardiac-death signal in higher-risk patients (Ray 2012), and the cumulative-over-lifetime AMR exposure. A purist individual-only reading would round to 1; the catalogue's holistic / lifetime read holds at 2.
health_short_term = 2. The NNH range (5–25 across reviews) is real but not transformative on any single encounter. 2 reflects "small but real."
effort_burden = 1, not 2. The action is occasional (twice a year for most adults), not daily. The friction is social, not effort-cost; 1 is more honest than 2.
mood and beauty_cumulative scored 0. The gut–mood and gut–skin axes are real but indirect at the depth this entry treats; scoring them would have required a paragraph each on speculative downstream effects. Per the spec, score-and-body track each other — kept clean.
evidence = 5. Four Cochrane reviews + ACP/CDC + IDSA + AAO-HNSF + Lancet AMR. The "name 2+ rigorous trials" bar is comfortably cleared.
controversy = 1. Mainstream medical bodies are aligned; residual disagreement is at the edges (immunocompromised cases, exact sinusitis threshold). The visible "controversy" is patient–doctor friction at the prescribing pad, not field-level dispute.

Excluded topics, deliberately.

Antibiotic use in livestock and aquaculture. A larger contributor to global AMR than outpatient human prescribing in many countries; out of scope because this entry is about the reader's individual decision in a clinic. Flagged in out-of-scope; a separate entry on agricultural antibiotic use is warranted.
Pediatric acute otitis media. Has its own decision tree (age, severity, observation vs immediate treatment). Separate-entry candidate.
UTI, skin, dental, STI antibiotics. Different indications, often genuinely bacterial. Separate categories.
Antibiotic-associated long-term outcomes (asthma, IBD, obesity, allergy). Mentioned briefly via Korpela 2016 in stakes; the causal certainty is not strong enough to put weight on individually. Future-link candidate when those entries exist.
Specific antibiotic-class adverse effects beyond azithromycin SCD. Fluoroquinolone tendinopathy / aortic aneurysm, severe cutaneous adverse reactions across classes, etc. Considered for inclusion in failure-modes; cut to keep that section about decision drivers rather than a class-by-class adverse-event tour.

Future link candidates. Pneumonia (decision tree), bacterial vs viral sinusitis (a tighter slice of this entry, if warranted), pediatric ear infections, agricultural antibiotic use, gut microbiome restoration after antibiotics, delayed-prescribing protocols as a clinician-side topic.

Hard calls during the write. The stakes section uses three-courses-a-year as the "typical reader" anchor; some readers will be at zero or one. Anchoring to three keeps the cumulative arithmetic visible without straying into the heavy-prescribing-only frame. The decision is consistent with the §5c rule against extreme-case anchoring — the typical adult in a high-prescribing country averages roughly that exposure across a decade.

Voice note. The article ends with a probabilistic / population-scale frame in the closing paragraph of payoff. That is intentional — the individual case for declining is real but small, and the honest framing of the bigger lever is the collective one. Felt-experience voice carries it inside ("the woman whose post-surgical infection responds in 2040"); the data layer hedges at sentence end via Murray 2022 and GBD 2022 in stakes.