The question in one sentence

In critically ill adults without renal replacement therapy, given a continuous-infusion style regimen, how often do virtual subjects jointly clear a free ceftazidime exposure rule against MIC and an avibactam threshold rule?

Notice what that sentence already fixed: population, infusion idea, free-drug thinking, and joint success.

Step 1: keep the components separate

Ceftazidime and avibactam get their own concentration-time profiles. Binding, clearance, and variability are component-specific. If published clearances are correlated, that correlation belongs in the draw. Collapsing both drugs into one fake “combo level” throws away the reason the pair exists.

Step 2: choose targets you can defend

A common teaching pattern is:

• ceftazidime judged on a free steady-state concentration relative to MIC,
• avibactam judged on staying above a fixed free threshold,
• joint success only when both pass.

Whatever thresholds you use, cite them and keep free/total conversion explicit.

Step 3: put kidney function in the population

Continuous infusion steady levels move with clearance. eGFR strata are not optional decoration for this story. Draw or assign renal classes, map them to clearance, and report attainment by stratum as well as overall. If the labeled regimen adjusts by renal function, say whether you modeled that adjustment.

Step 4: build PTA, then maybe CFR

Compute joint PTA across an MIC grid for the beta-lactam rule while the inhibitor rule stays active in every subject. If you have a dated MIC distribution for the relevant Gram-negative ecology, weight the PTA curve to get joint CFR. Then re-run under an alternate MIC prior if plasmid-driven shifts are a live concern.

Step 5: say what the case did not claim

A plasma free-drug joint PTA is not automatically ELF adequacy. A standard-inoculum MIC grid is not automatically a high-burden abscess story. Healthy-volunteer donor pieces are not automatically ICU truth. Write those limits in ordinary sentences. Readers forgive uncertainty. They distrust invisible upgrades.

Closing checklist you can read aloud

1. Are both components simulated, with free/total stated?
   
2. Is success defined as an intersection, not an average?
   
3. Are eGFR strata in the population, with dosing rules named?
   
4. Are donor PK and MIC sources labeled and dated?
   
5. Did you show joint PTA across MIC, and CFR only under an explicit ecology prior?
   
6. Did you name what you did not model: effect site, inoculum, RRT, and so on?

If you can answer those six without flipping through hidden notes, the teaching case did its job. A full research package will add more calibration and sensitivity detail, but the skeleton stays the same.

Takeaway

The CAZ-AVI style case is a checklist in narrative form: separate components, free-drug honesty, joint intersection, renal structure, provenance, then PTA/CFR. If you can explain those choices out loud without slides, the simulation is ready to be read by another scientist.

Related lessons: Joint targets, Free drug, eGFR strata, Provenance. Project page: CAZ-AVI evidence-composite PK/PD.

What people mean by inoculum effect

In plain terms: the same drug may need a higher concentration, or may kill more slowly, when the starting bacterial count is much higher. Enzymes can be more abundant. Binding or local drug depletion can matter more. Quorum-related physiology can change. The assay MIC at standard inoculum can then understate the challenge of a high-burden infection.

That does not mean every PTA grid is wrong. It means a high-inoculum pneumonia, endocarditis vegetation, or undrained focus may deserve a harsher PD scenario than a bloodstream model with lower density.

Where this collides with PTA habits

Classic PTA uses an MIC distribution from standard testing. If your clinical question is a high-burden site with a known inoculum effect for that class, a second scenario with worsened PD inputs is more honest than a single optimistic curve.

This also connects to tolerance and persistence. High density can change early kill even when categorical susceptibility looks unchanged. Keep the mechanism language careful: inoculum effect is not automatically “resistance,” though both can raise the practical exposure need.

Practical reporting

State the inoculum assumed by your PD target. If you borrowed in vitro high-inoculum data, mark them as donor evidence and check scenario compatibility. High-inoculum lab conditions are not automatically the same as an ICU pneumonia densitiy, but they are often closer to the worry you are trying to name.

How to keep the wording clean

Say “high-inoculum scenario” when you worsened PD inputs based on cited in vitro density effects. Do not say “resistant” unless genetics or MIC category actually changed. Do not say “persistent” unless you modeled a refractory subpopulation. Inoculum effect can interact with those ideas, but the labels should stay separate so collaborators know which lever you moved.

If the literature for your drug-bug pair is thin, it is better to list inoculum as an unquantified limitation than to invent a severity factor with no source. Empty honesty beats decorative precision.

Takeaway

Bacterial density can change the pharmacodynamic hill the drug must climb. When the literature flags an inoculum effect, put density into the scenario list beside MIC, binding, and renal function. Otherwise your clean PTA may describe a lighter fight than the infection.

Previous: Renal function and eGFR strata · Next: CAZ-AVI mini case.

Why kidney function changes the plot

If clearance rises as eGFR rises, the concentration-time curve sits lower for better renal function, all else equal. For time-driven targets, that can erase trough coverage first. For continuous infusion, steady free concentration moves almost directly with clearance. For AUC-driven targets, the whole area shrinks.

So renal strata are not decorative subgroups. They are different pharmacokinetic machines.

Building strata without clutter

A practical pattern:

1. Define eGFR classes that match the covariate model you cite.
   
2. Draw subjects inside each class, or draw eGFR then map to clearance.
   
3. Report PTA within strata and overall.
   
4. Keep dosing rules explicit if the labeled regimen itself adjusts by kidney function.

If the clinical regimen reduces dose in low eGFR, your simulation should not compare “full dose in everyone” unless that is the deliberate what-if.

What to report so readers can trust you

Name the equation used for renal function, the class boundaries, the clearance relationship, and whether dialysis or renal replacement scenarios were excluded. Mixed populations without labels are hard to reuse and easy to over-interpret.

What changes in the figure when you stratify

Overall PTA can hide a split world: strong attainment in low eGFR and fragile attainment in high eGFR on the same nominal dose. Or the reverse after renal dose reductions that overshoot. Stratified plots make that visible. A single pooled PTA can average the tension away.

When you present, show at least one stratum contrast beside the overall number. Readers who treat ICU kidneys as one bin will misuse your result. Readers who see the bins can argue with you productively about class boundaries and clearance equations.

If renal replacement therapy is excluded, say that in the first methods paragraph. Mixed RRT and non-RRT evidence is a classic provenance trap.

Also write down which equation produced eGFR or creatinine clearance in your covariate chain. Switching formulas mid-project is a quiet way to invent a new population without noticing.

If body size and renal function both enter clearance, keep that structure intact in the virtual draws. Dropping weight while keeping eGFR, or the reverse, can warp the tails that decide PTA at higher MIC values.

Takeaway

For renally cleared antibiotics, eGFR is not a demographic decoration. It reshapes clearance, exposure geometry, and attainment. Put kidney function into the virtual population on purpose, then say whether dosing was fixed or adjusted.

Previous: Effect site versus plasma · Next: Inoculum effect.

Plasma is a proxy, sometimes a good one

If free plasma exposure tracks free exposure at the effect site closely enough for your question, plasma-based PTA can be a fair first tool. If the site is protected, poorly perfused, or governed by transporters, plasma can flatter the regimen.

A simple mental split helps:

• Plasma-centric question: given systemic PK, how often do we clear a plasma free-drug target?
• Site-aware question: given systemic PK plus a site model or site observations, how often does the effect-site exposure clear the relevant target?

Those are different claims. Both can be useful. They should not share one sentence in the abstract as if they were identical.

Common site stories modelers meet

Epithelial lining fluid (ELF) for pneumonia work: ratios to plasma vary by drug and sampling time.
CSF for meningitis: barrier and inflammation change penetration.
Tissue or abscess: averages may hide necrotic cores where drug arrives late.

In each case you need either direct site observations, a justified transfer model, or an honest plasma-only limitation statement.

How to write this without hand-waving

In methods, name the compartment your target refers to. In limitations, say what you did not model. If donor ELF or CSF data come from a different population than your primary PK, treat them with the same provenance care you would use for a borrowed clearance equation.

Two abstracts, only one is honest

Honest: “We evaluated free plasma target attainment for regimen X in a virtual ICU population. Site-of-infection concentrations were not modeled.”

Overstated: “We showed regimen X provides adequate lung exposure,” after only plasma PTA.

The first abstract can still be scientifically valuable. The second asks readers to invent a penetration model you never wrote down. If you do have ELF or CSF data, bring them in with dates, methods, and sensitivity to lower penetration. If you do not, keep the claim inside plasma and sleep better.

Timing matters too. A site-to-plasma ratio from a late sampling window may not describe early therapy. If your PTA question includes the first hours, a single steady ratio can be the wrong proxy.

Takeaway

Bacteria live at an effect site. Plasma is often the ladder we climb to estimate that site, not the destination itself. Keep the proxy relationship visible, or your PTA will answer a cleaner question than the infection deserves.

Previous: HGT and MIC ecology · Next: Renal function and eGFR strata.

From gene traffic to P(MIC)

CFR needs weights: how common is each MIC category in the population you care about? That is an ecological claim. It includes species, region, year, specimen mix, and which mechanisms are circulating.

If a plasmid carrying a clinically important enzyme expands in a hospital network, the right-hand side of the MIC histogram can fatten within seasons. A CFR computed on last year’s weights can remain numerically tidy and scientifically late.

Habits that survive contact with mobile resistance

1. Date-stamp every MIC source used in CFR.
   
2. Separate mechanism strata when the data allow (enzyme class versus target mutation).
   
3. Recompute CFR under alternate local distributions as sensitivity analysis.
   
4. Avoid promoting one ward antibiogram into a universal prior.
   
5. Keep the chromosome-versus-plasmid distinction nearby when you talk about stability.

A concrete CFR story

Your PTA curve looks reassuring through MIC 4 mg/L and then falls. Under last year’s local distribution, most isolates sit at MIC 1 or 2, so CFR looks high. This year a plasmid-associated enzyme becomes common and the mass shifts toward MIC 4 and 8. The PTA curve did not change. CFR did, because the weights changed.

That is why dating the MIC prior is not bureaucracy. It is part of naming the ecological world your summary number lives in. If you only have an old table, say so and show a harsher alternate prior beside it.

Pair this habit with the chromosome-versus-plasmid lesson: mobility tells you how nervous to be about calendar age.

When surveillance is sparse, you can still be useful. Report PTA as the main figure and treat CFR as exploratory under two or three stated priors. That is better than one confident CFR built on an undated convenience sample.

Ask the microbiology lab what changed recently: outbreak notes, new carbapenemase detections, or shifted species mix. Informal context is not a substitute for surveillance tables, but it tells you whether yesterday’s prior deserves extra suspicion.

Takeaway

Treat MIC ecology as perishable when mobile elements are in play. Your PTA curve can stay the same while CFR moves, simply because the weights under the curve changed.

Previous: Chromosome vs plasmid resistance · PK/PD: Uncertainty and provenance · Next: Effect site versus plasma.

Label every input like it might be audited

Useful labels look boring. That is why they work.

Missing data should stay missing until a compatible donor is justified. Silent median fills create false completeness.

Uncertainty comes in layers

1. Parameter uncertainty: IIV, residual error, covariate effects.
   
2. Structural uncertainty: compartment choice, infusion constraints, correlation.
   
3. Evidence uncertainty: wrong population, wrong matrix, stale MIC ecology.
   
4. Decision-rule uncertainty: which threshold counts as “enough.”

Monte Carlo on layer 1 alone can look precise while layer 3 quietly dominates.

Compatibility is a scientific decision

ICU continuous-infusion parameters are not automatically interchangeable with healthy-volunteer bolus data, CSF studies, or in vitro kill curves. Those sources can still help as donors or anchors. They should not be pooled as if they were one population. Write the compatibility rule in human language: what was allowed to enter the primary scenario, and what stayed outside with a label.

What good provenance feels like in a meeting

Someone asks why ICU PTA used a clearance relationship from a continuous-infusion critically ill study and an MIC table from a regional surveillance year. You can answer in one breath: primary PK from source A for scenario match, MIC weights from source B dated to that year, fu transformed from total using source C, inhibitor threshold from source D with a compatibility note. No scavenger hunt through forgotten spreadsheets.

If you cannot give that breath-length answer, stop polishing the figure and rebuild the input table first. Most “mysterious” PTA sensitivity traces back to one unlabeled donor.

Keep a one-page input ledger beside the code: parameter, value or distribution, source, transformation, scenario role. It is unglamorous. It is also what makes the project reusable six months later.

Takeaway

Reproducible PK/PD is less about fancier random draws and more about refusing to launder uncertain evidence into false certainty. If the trail of sources is clear, other people can challenge your result productively. That is a feature.

Previous: Joint targets for BL/BLI pairs · Next: HGT and MIC ecology · Then: Effect site versus plasma.

Two conditions, one success flag

Suppose drug A must meet a free-exposure target against MIC, and drug B must stay above a threshold that preserves A’s activity. For each simulated person:

[ = {A} {B} ]

Joint PTA is the fraction of people for whom that intersection is true. It is not the average of PTA(A) and PTA(B). It is not “A looks fine, so ignore B.”

Mistakes that look sophisticated

1. Simulating only the beta-lactam because “that is the antibiotic.”
   
2. Reporting the mean of two PTAs as if it were joint attainment.
   
3. Borrowing an inhibitor threshold from another matrix or population without labeling.
   
4. Collapsing both drugs into one fake combined concentration.

Why this shows up in CAZ-AVI style work

Ceftazidime and avibactam are a teaching example because the beta-lactam exposure and the inhibitor exposure are not interchangeable. A regimen can clear a ceftazidime rule while avibactam is the limiting partner, or the reverse depending on the scenario. Component-level evidence reconstruction exists partly because papers report pieces of that system separately, and the modeler has to keep those pieces labeled until a joint rule is applied on purpose.

A tiny counting example

Imagine 100 virtual subjects.

• 90 pass the beta-lactam rule.
  
• 80 pass the inhibitor rule.
  
• Only 70 pass both.

Joint PTA is 70%. The average of 90% and 80% is 85%. That 15-point gap is exactly why intersection logic exists. If clearances are positively correlated, joint failures may cluster; if not, the overlap changes again. Either way, report the joint indicator you actually coded.

When you write results, show component PTA only as supporting context. Lead with joint PTA if the clinical object is the combination.

If you need a debugging view, plot the fraction failing A only, B only, and both. That split tells you whether the regimen is beta-lactam limited, inhibitor limited, or jointly fragile. A single joint number hides that diagnosis.

Takeaway

For BL/BLI pairs, protection failure is regimen failure. Joint PTA is an intersection inside each simulated person. Build it that way, then be transparent about thresholds, correlation, and which MIC the beta-lactam rule refers to.

Previous: Intermittent vs continuous infusion · Next: Uncertainty and provenance · Case study: CAZ-AVI end-to-end.

Same daily dose, different shape

Hold clearance, volume, fu, and MIC fixed. Change only the input rate.

• Intermittent short infusion: oscillating free levels. Success depends on how long each cycle stays above MIC.
  
• Extended infusion: softer peaks, higher troughs.
  
• Continuous infusion: approach to steady state. Many analyses then ask whether steady free concentration stays above MIC, or above a multiple of MIC.

What the simulator must keep honest

1. Loading dose status.
   
2. Renal function strata, because steady concentration moves with clearance.
   
3. Administration constraints that your scenario ignores (stability, line access). Label them.
   
4. Target form: percent fT>MIC versus a Css/MIC ratio rule. They are related ideas with different bookkeeping.

A classroom example without fake precision

Suppose two regimens deliver the same daily free-drug amount, one as intermittent doses and one as continuous infusion after a load. At low MIC both may look fine. As MIC rises, the intermittent troughs fail first while continuous Css may still clear a Css/MIC rule. That pattern is exactly why people study infusion mode. It is also why you must keep daily dose and population fixed when you make the comparison.

How to talk about this without overselling

A fair sentence sounds like this: “Under the same daily dose and the same virtual population, continuous infusion improved PTA at higher MIC values because trough undershoot disappeared; early coverage still depended on the loading dose.”

An unfair sentence sounds like this: “Continuous infusion is better,” with no fixed daily dose, no population lock, and no target definition.

Also remember pharmacy and line realities. A simulation can assume perfect continuous delivery. A ward may not. If your scenario ignores stability windows or access constraints, write that limitation in ordinary language. The PK/PD geometry lesson remains useful either way.

One more practical tip: when two papers disagree about continuous infusion benefit, check whether they locked daily dose, population, free/total handling, and target definition. Many apparent contradictions dissolve once those four locks are compared side by side.

Takeaway

Infusion mode redraws the exposure path around the MIC. For time-driven beta-lactam targets, that redraw is part of the science. Compare shapes under controlled conditions, and say whether your endpoint was interval fT>MIC or steady-state coverage.

Previous: Free drug and protein binding · Next: Joint targets for BL/BLI pairs.

What free means in plain language

In plasma, some drug molecules stick to proteins such as albumin. The usual modeling story says the unbound molecules are the ones available to act. The MIC assay is generally interpreted on that unbound-compatible scale. So the simulator should put free concentration versus time next to MIC.

A simple linear correction is:

[ C_{}(t)=f_uC_{}(t) ]

where (f_u) is the unbound fraction. This is a starting point, not a religion. Binding can change with albumin, pregnancy, burns, or critical illness. For a highly bound drug, a small error in (f_u) becomes a large error in free exposure.

Where people get hurt in practice

1. Target says free, plotted series is total.
   
2. fu comes from healthy volunteers, applied unchanged to hypoalbuminemic ICU patients.
   
3. The infection is not in plasma (ELF, CSF, tissue), but the model never leaves the plasma compartment.
   
4. Combination regimens use one fu for both components without checking.

A short checklist before you run PTA

• Label every series: free or total.
  
• Cite fu and whether it is fixed, sampled, or covariate-dependent.
  
• Re-run with lower and higher fu as a one-way sensitivity check.
  
• For BL/BLI pairs, bind each component on its own terms.
  
• If the effect site is not plasma, say so and move to the effect-site lesson with honest humility.

Walk through one messy case

You fit a model to total plasma concentrations. The target you want is 100% fT>MIC. Literature fu is 0.25 in healthy volunteers, but your population is critically ill and often hypoalbuminemic. A first run uses fu = 0.25. A second run uses fu = 0.40 as a what-if for lower binding. A third run keeps total concentrations and “forgets” the f, which wrongly inflates attainment.

Only the first two runs belong in a serious report, and both should be shown if binding is uncertain. The third run is a unit error dressed as optimism.

When you present results, say the free/total conversion out loud before you show the PTA plot. If listeners cannot tell whether the curve is free or total, the figure is not ready.

Takeaway

Binding is not cosmetic formatting. It decides which curve is allowed to talk to the MIC. Align the binding state of the PK output with the binding state of the PD target before you interpret attainment.

Previous: Resistance, tolerance, and persistence · Next: Intermittent vs continuous infusion.

Two homes for resistance genes

Resistance can live on the chromosome or on a plasmid (and related mobile elements). Both can raise an MIC. They do not raise it with the same epidemiological behavior.

Chromosomal changes often travel mostly with the lineage: mother cell to daughter cell. Plasmid-borne resistance can move horizontally as well, sometimes across lineages or species. Plasmid copy number can also amplify expression, so “same gene” is not always “same burden.”

Same MIC number, different story

Imagine two isolates with MIC = 8 mg/L to the same drug.

In isolate A, a chromosomal target mutation reduced binding. In isolate B, a plasmid beta-lactamase is chewing the drug. A PTA grid may treat them identically if it only sees “8”. A combination with an inhibitor may help B and do little for A. Your joint-target lesson later depends on knowing which world you are in.

What to change in a PK/PD workflow

1. Label MIC provenance with date, region, species, and method.
   
2. Ask whether the mechanism is enzyme-mediated before you lean on inhibitor partners.
   
3. Refresh local MIC weights more aggressively when plasmid epidemics are active.
   
4. Keep genetics in the annotation layer unless you are explicitly building a genotype-to-phenotype model.

A short example conversation

Someone hands you an MIC table and asks for CFR. Before you run anything, ask three genetics-aware questions:

1. Are these isolates dominated by one clone, or is the lab seeing mixed plasmid traffic?
   
2. How old is the table, and has the hospital seen a recent enzyme-gene surge?
   
3. If we later evaluate an inhibitor combination, does the mechanism even match that idea?

You may still proceed with a standard CFR. The difference is that your methods paragraph will say what kind of phenotype ecology you assumed, instead of pretending every MIC cell is interchangeable biology.

If the answer to those questions is “we do not know,” that is still useful. Unknown mobility is itself a reason to run sensitivity CFR under a harsher MIC prior.

Takeaway

Do not ask genetics to replace PK/PD. Ask it to annotate the MIC layer: how mobile is this phenotype, how old is the ecology prior, and which mechanism your target rule quietly assumes?

Next genetics lesson: HGT and MIC ecology · PK/PD spine: Resistance, tolerance, and persistence.

Three words that should not be synonyms

Resistance usually means a heritable drop in susceptibility. In practice that often shows up as a higher MIC or a shifted pharmacodynamic curve. Genetics can sit underneath (target mutation, enzyme, efflux), but the PK/PD layer mainly notices the phenotype change.

Tolerance means a population can endure drug pressure longer without a large MIC shift. Time-kill curves look sluggish. The organism is not “fully susceptible in behavior” just because the MIC table looks polite.

Persistence points to a small subpopulation that survives much longer. The bulk may fall, then a refractory tail remains. Relapse risk lives in that tail more than in the first steep decline.

What the curves look like

Why a strong PTA can still disappoint

PTA asks whether exposure clears a threshold rule. That is a fair question. It is not every question.

A regimen can look excellent against the measured MIC distribution while:

• killing is too slow for the planned duration,
• a persistent fraction remains,
• or the assay MIC understates the real challenge at the infection site.

None of that makes PTA useless. It means you should label what your simulation tested. “High PTA vs MIC grid” is not the same claim as “sterilizing cure in every biology.”

Pathogen contrast, briefly

Different organisms lean on different survival styles. In comparative modeling work, Staphylococcus aureus and Mycobacterium tuberculosis are a useful pair: one often shows a faster tolerance-like pattern, the other a longer persistence window. The lesson for the modeler is simple. Do not copy a kill structure from one species into another because the antibiotic name matches.

Takeaway

Resistance moves the threshold. Tolerance slows the population kill. Persistence leaves a long tail. Say which one your simulation is actually testing, especially when the clinical story is failure despite a tidy MIC.

Previous: PTA, CFR, and virtual populations · Next: Free drug and protein binding · Genetics: Chromosome vs plasmid resistance.

Virtual populations are controlled variation

A virtual population is a set of simulated individuals. For each person you draw pharmacokinetic parameters from distributions that are supposed to reflect interindividual variability. If the literature reports correlated clearances, or clearances that depend on eGFR, that structure belongs in the draw. If you invent tidy independent normals because they look neat, you are no longer studying the published system. You are studying your drawing habits.

Useful discipline looks like this:

• cite the IIV source,
• keep covariate rules explicit,
• record the random seed and the sample size,
• say which population the parameters came from (ICU, healthy volunteers, dialysis, and so on).

PTA: fix the MIC, count the successes

Probability of target attainment (PTA) answers a conditional question:

If MIC equals this value, what fraction of virtual subjects meet the PK/PD target?

You usually compute PTA across a grid of MIC values and plot a declining curve. The interesting part is often the cliff: the MIC region where success falls apart.

CFR: bring the local microbiology back in

Cumulative fraction of response (CFR) folds PTA over an MIC distribution from surveillance or local epidemiology:

[ = _i (_i)P(_i) ]

PTA asks “what if MIC is X?”
CFR asks “given the MICs we actually see around here, how often do we succeed?”

That second question is only as honest as the MIC weights. A distribution from another country, another year, or a mixed species group can make CFR look decisive while answering the wrong question.

A minimal workflow you can defend

1. Define regimen and PK model, with provenance.
   
2. Define free-drug target or targets.
   
3. Simulate N virtual subjects.
   
4. Compute PTA on an MIC grid.
   
5. Optionally weight by an MIC distribution to get CFR.
   
6. Stress-test the fragile inputs: binding, IIV, renal strata, MIC source.

If the target is joint (two drugs, two conditions), success is the intersection inside each virtual subject. Averaging two separate PTA values is not the same thing. A later lesson walks through that case.

Takeaway

Virtual populations turn “does this look fine on paper?” into “how often does this pass under realistic variability?” PTA keeps MIC fixed. CFR brings epidemiology back. Neither repairs a bad target choice or a mismatched population.

Previous: PK/PD targets and the MIC · Next: Resistance, tolerance, and persistence.

Start with the awkward object: the MIC

The minimum inhibitory concentration is the lowest concentration that stops visible growth in a defined lab assay. That sentence already hides three compromises.

First, “visible growth” is an assay endpoint, not a full story of killing. Second, the assay has a floor and a set of conditions. Change those conditions and the number can move. Third, the pathogen in a patient is not living in Mueller-Hinton broth with perfect mixing and a fixed inoculum.

Still, MIC is useful because it compresses susceptibility into something a simulator can compare with an exposure curve. The healthy habit is to treat MIC as a scenario input with uncertainty, not as a moral fact about the organism.

Three index families, said simply

Most antibacterial PK/PD targets fall into one of three families. The leading f means free (unbound) drug.

If your model outputs total plasma concentration but the target is written for free drug, the units are already arguing with each other. Fix that before you admire the PTA curve.

Choosing an index without mystique

Ask three questions before you code a target.

1. What does the drug class literature actually use for this organism and indication style: time above threshold, integrated exposure, or peak?
2. Are you modeling free or total concentrations, and does the cited target match that choice?
3. Is success one condition or two? Combinations may need joint rules.

A published target such as “100% fT>MIC” or “fAUC/MIC at least 125” is a decision rule for simulation. It is not a promise that every patient above the line is cured and every patient below it fails.

A working habit for modelers

Write the index, the free/total state, and the MIC source in one line whenever you start a run. If you cannot fill that line, the simulation is not ready for interpretation yet.

Tolerance and persistence can still spoil a “good” MIC story. We will separate those ideas in a later lesson. For now, keep the simpler rule: pick the index that matches the pharmacology, then keep the units honest.

Next: PTA, CFR, and virtual populations.