Medicine prices in Latin America's online pharmacies: 24-country benchmark

1. Introduction

Medicines are a major component of household health spending in Latin America and the Caribbean. A private retail price is therefore a practical access indicator, although it is not the same as the price paid by every patient or the cost borne by a health system (Wirtz et al., 2016; Emmerick et al., 2015).

The WHO and Health Action International approach remains the most useful foundation for a standard medicine basket, explicit product matching, availability, and affordability in days of the lowest wage (World Health Organization & Health Action International, 2024; Cameron et al., 2009; Niens et al., 2012). Its strength is comparability of method. Its historical international supplier price series is not used in this report because the last edition is too old to provide a credible 2026 benchmark. Applying consumer inflation would change the date, but would not make the underlying procurement prices current or comparable with retail prices.

Web-scraped prices provide broad, repeatable coverage but measure published catalogues rather than transactions. Online and physical store prices often coincide for large multichannel retailers, yet coverage can favour chains and urban markets, and online pricing can itself be more dispersed (Cavallo & Rigobón, 2016; Cavallo, 2016; Aparicio et al., 2023). These limitations determine the language used throughout: listings, catalogue composition, and observed prices, never sales, utilization, or market share.

Why the comparison method matters

Several approaches were considered. External price referencing can compare the same product across jurisdictions, but it is sensitive to product identity, price type, confidential discounts, and country selection (Vogler et al., 2019; Kanavos et al., 2019). A conventional fixed-weight consumer price index requires quantities or defensible weights, which these catalogues do not provide. A simple average of prices would also let medicines with high unit prices dominate. The selected method instead compares each country-medicine median with the regional median for that same medicine, then takes the median of those relatives within each country. It controls medicine identity, does not invent sales weights, and is robust to extreme values.

What can be learned from product names

An originator is a country-specific regulatory status, not a visual property of a product title. Regional brand aliases, manufacturer names, and substring matches cannot establish which presentation was first authorized in each country. The report therefore does not estimate an originator premium. It uses only the observable naming distinction between a product sold under a trade name and one sold under its international non-proprietary name. This is narrower than the branded-generic typology in the literature, but it is supportable from the collected evidence (Kaplan et al., 2013).

Objective

The objective is to establish a transparent and repeatable benchmark of published private retail medicine prices across the 33 countries of Latin America and the Caribbean. The primary questions are how prices for the same medicine vary across countries and how much of a lowest daily wage a standard treatment costs. Secondary analyses test currency sensitivity, observable product naming, manufacturer attribution, and associations with health-financing indicators.

2. Methods

Study frame and data source

The study frame is the 33 sovereign countries of Latin America and the Caribbean. Included sources are commercial pharmacy or pharmacy-adjacent websites that expose product-level prices without bypassing access controls. Prices were retrieved between 1 and 3 September 2026. One listing is one captured product-price observation from one online catalogue. It is not a transaction.

Table 1. The 16-medicine basketOne strength and form is fixed for each medicine before comparison. Fourteen items follow the WHO and Health Action International global core list; two common solid oral presentations complement surveyed suspensions.Table: Decilion. Data: Decilion.

From a listing to a comparison

A listing must match the basket medicine, strength, and dosage form; combination products are excluded unless the basket item is itself a fixed combination. Pack size is resolved before price is divided by tablets, capsules, millilitres, vials, or inhaler doses. The published list price is preferred, with a promotional price used only when no list price is available.

The matching gates leave 3,521 eligible listings from 16,423 retrieved. Comparisons with fewer than three listings remain visible for transparency but are excluded from price rankings, dispersion, affordability, and the matched product-naming analysis.

The median is preferred to the average because a few unusually high listings can pull an average upward. The primary median still gives every eligible listing equal weight, so a pharmacy with a wider online assortment contributes more observations. A sensitivity analysis first takes the median within each pharmacy and then the median across pharmacies. This gives each observed pharmacy one value for a medicine. The lower quartile describes the more affordable part of the catalogue while reducing reliance on a single minimum.

Price and currency comparisons

The market-rate country index divides each country-medicine median by the regional median for the same medicine and multiplies by 100. The country result is the median across its supported medicines. An index of 100 therefore means the regional median for the same set of medicine-level comparisons, not an absolute price level.

All market exchange rates are dated 1 September 2026. The same medicine-level index is recomputed with purchasing power parity factors over the countries for which those factors are available. Market rates answer what prices convert to in traded United States dollars; purchasing power parity better reflects domestic purchasing power and local retail inputs. Neither is treated as uniquely correct. The report displays both and interprets only results that are robust to the choice.

Affordability is the cost of a standard 7-day acute or 30-day chronic course divided by the lowest statutory daily wage. Because the course price and wage begin in the same local currency, the exchange-rate conversion cancels exactly (Cameron et al., 2009; Niens et al., 2012).

Product naming and manufacturers

Product naming has two observable categories. Named product means the listing carries a trade name; INN-only means it is presented under the international non-proprietary name. No inference of originator status is made. Prices are compared only where both categories have at least three listings for the same medicine and country.

Manufacturer is retained as the analytic term because online-pharmacy price studies use it for the company attached to a listed brand (Negi et al., 2021). In this report it means the company or laboratory identified by the pharmacy catalogue or the shared product master. It does not assert which plant made a specific pack. WHO terminology itself allows manufacturer to overlap with a marketing-authorization applicant or holder, while treating legal and contract manufacturing as distinct roles (World Health Organization, n.d.).

Resolution prefers the product registry, then the listing field. A value identical to a product brand is not accepted as a manufacturer. The direct catalogue label, researched corporate group, home country, and source are kept in separate fields. Physical manufacturing sites are omitted because establishing them would require product-by-product verification in national regulatory records. Counts describe catalogue listings, not sales or market share.

Health-financing analysis

The market-rate price index and median days of wage are tested against six WHO Global Health Expenditure Database variables using Spearman rank correlation. All twelve tests are reported without a multiplicity adjustment and are treated as exploratory.

3. Results

3.1 Online availability

At least one basket medicine is priced in 24 countries. Table 2 shows both the breadth of the basket found and the number of listings behind it. A country outside the ranking lacks enough supported medicines; it is not classified as cheap or expensive.

Coverage is broadest in Brazil, Guatemala, Peru and Dominican Republic, where 15 of the 16 basket medicines were found. It is sparsest in Guyana, Haiti, Grenada, Trinidad and Tobago and Barbados, with no more than three. These differences describe what the collection could observe online, not medicine availability across all pharmacies or health services.

Table 2. Availability by countryCountries below the ranking threshold remain visible.Table: Decilion. Data: Decilion.

3.2 Pharmacy sources

Table 3 lists 107 online pharmacy sources across 24 countries. The median source prices 8 basket medicines; 60 sources price at least half of the basket, while 13 price only one medicine. This uneven catalogue depth is why the report also tests an equal-pharmacy weighting.

Each source is counted once, regardless of how many physical outlets it represents. The table therefore documents the evidence base rather than the number of pharmacies operating in a country. The frozen first column keeps the pharmacy and country visible while numeric columns scroll.

3.3 Cross-country dispersion

Figure 1 shows every supported country median for each medicine, expressed relative to that medicine's regional median of 100. The dots make the centre, clusters and tails visible; a single average or maximum-to-minimum ratio would conceal that structure.

After excluding salbutamol's mixed-unit artefact, the widest observed span is for bisoprolol 5 mg tablet: the highest country median is 106.1 times the lowest. The chart is descriptive and not sales-weighted, so dense clusters indicate similar posted prices, not where most patients purchased.

Figure 1. Country prices vary within every medicineEach dot is one supported country-medicine median relative to the regional medicine median of 100. Salbutamol is excluded because its mixed units create a known artefact.Chart: Decilion. Data: Decilion.

3.4 Comparability across analytical lenses

Currency and catalogue composition both matter. Figure 2 holds the sample fixed at 17 countries and places three analytical lenses side by side. Purchasing power parity changes the rank of 14 countries: Argentina moves from fifth at market rates to first under purchasing power parity, while Costa Rica moves from first to third. These are different interpretations of the same underlying local prices, not competing claims about which conversion is uniquely correct.

Catalogue composition also matters. The listing median differs by more than 10 percent from the equal-pharmacy median in 52 of 173 supported comparisons with at least two pharmacies. Giving each pharmacy equal weight therefore tests whether one large catalogue is exerting disproportionate influence.

Within 436 pharmacy-medicine catalogues with at least three listings, 76 have a median at least 50 percent above the lower quartile. Cheaper listings can therefore coexist with a higher typical listing. This is consistent with a catalogue containing both affordable and high-priced options, but the data do not show which option patients purchase or whether competitors react to one another.

Figure 2. One sample, three price-index lensesThe same 17 countries appear in every column. Values and ranks are shown together; 100 is the regional medicine reference. None of the columns is weighted by sales.Chart: Decilion. Data: Decilion.

3.5 Affordability

Figure 3 compares the typical medicine with the least affordable medicine inside each ranked country. In 8 of 18 countries, even the median standard course costs more than one day of the lowest statutory wage. In the other ten, a relatively low typical burden can coexist with one course that is much less affordable.

The largest supported burden is ceftriaxone 1 g injection in Argentina, at 22.0 days of the lowest wage for one standard course. Ceftriaxone is also the least affordable medicine in the five countries with the highest worst-case burden. The distance between the two ends of each line, rather than the country median alone, is the central result.

Figure 3. Treatment burden is not the price rankingDays of the lowest wage is independent of exchange-rate conversion. The adult dose used for paracetamol suspension is an upper bound.Chart: Decilion. Data: Decilion.

3.6 Exact-product pharmacy prices

Holding the country, medicine and a valid product identifier constant narrows the product-mix problem. Brazil has 136 interpretable products seen at three or more pharmacies, with a median highest-to-lowest ratio of 1.13. Colombia has 47, with a median ratio of 1.30.

Figure 4 shows the full distributions rather than only those two medians. In Brazil, 77.9 percent of exact products are priced within 25 percent of the lowest observed pharmacy price, compared with 40.4 percent in Colombia. Within a 50 percent band, however, the shares are similar at 84.6 and 80.9 percent. Brazil has the tighter centre but also the longer upper tail.

The audit output retains 1 extreme comparison for source review rather than silently treating them as price findings. No sales, stock, discounts after login or competitor response data are available, so clustering cannot establish coordination.

Figure 4. Exact-product price spreads in Brazil and ColombiaEach dot is one product observed at three or more pharmacies. Only these two countries meet the 20-product display threshold.Chart: Decilion. Data: Decilion.

3.7 Product naming

Figure 5 expresses the named-product median as a percentage of the INN-only median for the same country and medicine. A bar at 100 percent means equal prices, a value above 100 means the named product is dearer, and a value below 100 means it is cheaper. The number in brackets is the count of matched country comparisons, which is essential for judging how stable each medicine result is.

The medicine-level median is above 100 percent for 11 of the 15 medicines and below it for 4. The most extreme bar, co-trimoxazole tablets at 252 percent, rests on only three country comparisons. Results with deeper support are more moderate: paracetamol tablets are at 110 percent across 15 comparisons, ciprofloxacin at 106 percent across 13, and omeprazole at 87 percent across 11.

Across all 93 matched country-medicine comparisons, named products are dearer 54 times and cheaper 39 times; the overall median is 106 percent of the INN-only price. The mixed direction is why the report does not treat a trade name as a general price premium, and neither naming category establishes originator status. Salbutamol remains in Figure 5 because each ratio is formed within one country before aggregation; the cross-country unit-scale mismatch that excludes it from Figure 1 does not directly enter this calculation.

Figure 5. A trade name does not consistently predict priceNamed-product price divided by INN-only price within the same country and medicine. Naming does not establish originator status.Chart: Decilion. Data: Decilion.

3.8 Manufacturers identified in pharmacy catalogues

The observation field identifies 2,361 usable manufacturer labels and the registry recovers 507 more; 653 listings remain unresolved. In this report, manufacturer means the company or laboratory identified by the pharmacy catalogue or product master. It does not identify the physical plant that made a pack.

Table 4 researches the corporate identity of the 15 leading resolved labels. 4 are brands or operating identities rather than unambiguous legal-company names. An official corporate group was identified for 12; no parent was found in official materials for 3. Medley remains under Sanofi because its announced sale to Grupo EMS was awaiting regulatory approval and closing on the evidence date. Merck is resolved to Merck KGaA, Darmstadt, Germany, from the Concor and Glucophage products in the underlying rows, not from the name alone.

The largest listed entity, EMS, represents only 4.6 percent of eligible listings, while the unresolved group represents 18.6 percent. Corporate groups are linked to official evidence but are not substituted for the published labels in the counts. With no sales volumes, the table cannot estimate market shares or concentration.

Table 4. Leading manufacturer or company labels in the online catalogueThe listed entity's home country appears beneath its name. Counts are catalogue listings, not sales; markets observed are countries in this dataset, not corporate operating reach.Table: Decilion. Corporate identities: official company disclosures, checked 4 September 2026.

3.9 Prices and health financing

Table 5 reports all twelve exploratory rank correlations between the two price outcomes and six WHO health-financing indicators. Only one of twelve exploratory correlations is nominally significant: the market-rate index and out of pocket spending per person (rho 0.672, p 0.0023, n 18). It may reflect income, currency conversion, or the mechanical effect of higher medicine prices on household spending.

A conservative Bonferroni sensitivity across the twelve tests gives an adjusted p value of 0.0276, so the association remains statistically detectable. That correction does not solve the more important interpretive problems: the cross-section has only 18 countries, spending and prices may affect one another, and the online retail channel is only part of total medicine access. The other eleven coefficients do not provide evidence of an association in this collection.

Table 5. Price outcomes and health-financing indicatorsSpearman rank correlations over ranked countries. The p values shown are unadjusted; the pharmaceutical-spending rows have only seven countries and therefore the weakest support.Table: Decilion. Data: Decilion; WHO Global Health Expenditure Database.

4. Discussion

The strongest result is not a single league table. It is the combination of large within-medicine dispersion, visibly different market-rate and purchasing-power-parity rankings, and an exchange-rate-invariant affordability measure. Together they show why a cross-country price statement must name its currency lens.

The internal median-of-relatives index is better suited to this catalogue than an old procurement benchmark or a quantity-weighted price index. It holds the medicine constant, tolerates incomplete country baskets, and does not invent purchase weights. Its limit is equally important: countries price different subsets of the basket. The five-medicine ranking threshold and the displayed interquartile ranges reduce, but do not eliminate, that composition problem.

The pharmacy-balanced sensitivity makes a second composition problem visible: a large online assortment otherwise gives one pharmacy more influence than a smaller competitor. The resulting changes are large enough that the primary ranking should be read together with this sensitivity, particularly in the middle of the distribution.

The exact-product comparison removes one important source of product mix and shows that posted prices often cluster, particularly in Brazil. It cannot show that competitors agreed on a ceiling. The catalogues do not reveal transactions, sales, stock, costs or patient choice. Testing competition would require repeated same-product observations over time and evidence of how pharmacies respond to each other's price changes.

The product-naming result is deliberately narrower than an originator analysis. A trade name is observable in a catalogue; first authorization and reference-product status require country-specific regulatory evidence. Collapsing these concepts would create precise-looking premiums from an unreliable label. The named-versus-INN comparison can describe catalogue presentation and price, but it cannot estimate a causal brand effect.

Manufacturer normalization and the researched corporate crosswalk improve the leading-label analysis but do not solve entity resolution for the full catalogue. Product registries recover many missing fields, yet manufacturers, marketing authorization holders, distributors, brands, and corporate parents can all appear in the same catalogue slot. Keeping the published label, provenance, corporate group, and home country separate is more informative than forcing every label into one owner or claiming a physical manufacturing site.

Online catalogues remain a selected view of retail markets. They do not reveal stock at the moment of purchase, discounts after login, insurance reimbursement, public provision, sales volume, or which listed product patients choose. The next collection should preserve the same basket and rules, deepen sparse Caribbean coverage, and add country-specific regulatory identifiers before reconsidering any originator question.

5. Conclusions

This first collection demonstrates that a regional online pharmacy benchmark is feasible and that currency choice materially affects cross-country interpretation. Market-rate dollars, purchasing power parity, and days of wage answer different questions and should remain visible together.

The defensible substantive findings are that prices for the same medicine vary widely, exact-product posted prices cluster more tightly in Brazil than Colombia within the limited matched sample, affordability does not simply reproduce the market-rate ranking, and named products are not consistently dearer than INN-only products in matched comparisons. Manufacturer metadata remains too incomplete for market-concentration claims.

A repeat collection using the frozen matching rules would convert this cross-section into a short panel. That is the most valuable next step because it would distinguish persistent price levels from temporary exchange-rate movements and catalogue changes.

Appendix

Reproducibility

The report was built from the frozen extractor package frozen-wave3-2026-09-04 and analysis commit b895c18. Input files are checked against package manifests, exclusions are recorded, and every literal covered by the report checks is recomputed before the HTML is written.

The independent reproduction check confirms listing counts and unit price medians within the one percent acceptance gate across 280 extractor country-medicine summaries. Historical reference-price ratios may remain in internal compatibility files, but they are not an analytical result or a reported comparison.

Table 6. Manufacturer attribution provenanceUnresolved brand values are kept separate from truly empty or unusable manufacturer fields.Table: Decilion. Data: Decilion.