In total, we included 1723 anticipated in-person company attendees in this analysis. We calculated total carbon emissions to be 3,262,574 kgCO2e (Fig. 2a). Travel accounted for 91–96% of the total carbon emissions per congress, and intercity travel (particularly long-distance and business-class air travel) accounted for 90–94% (Fig. 2a). Indeed, published emissions rates are 2.9-fold higher for long-haul business-class flights (0.4239 kgCO2e per p.km) than for short-haul economy-class flights (0.1462 kgCO2e per p.km) [29]. In a sensitivity analysis, if company attendees modelled as having travelled by business-class flights had in fact travelled by economy-class flights, total air-travel emissions were reduced by approximately 50% for each congress (emissions [kgCO2e] per attendee: ASCO, from 906 to 518; ENETS, from 3935 to 1466; ESMO, from 1337 to 672; WCNR, from 1084 to 532).

Total carbon emissions associated with in-person congress attendance by a contributing activity and b congress, for pharmaceutical company attendees. *The average annual carbon footprint of a UK resident is 5600 kgCO2e [38]. Values are subject to rounding. ASCO American Society of Clinical Oncology, ENETS European Neuroendocrine Tumor Society, ESMO European Society for Medical Oncology, kgCO2e kilograms of carbon dioxide equivalent, WCNR World Congress for Neurorehabilitation

Hotel accommodation was the next largest contributor to total carbon emissions per congress, accounting for 4–9%. Taxis and cooling of company attendees at the congress venue accounted for approximately 1% and 0.01%, respectively, of total carbon emissions per congress. Even if the cooling loads were increased 100-fold (sensitivity analysis), the impact on total emissions was still < 1%.

Carbon emissions per company attendee ranged from 1009 to 1479 kgCO2e for the European congresses, representing 18–26% of the average UK resident’s annual carbon footprint (Fig. 2b). Considering intercontinental travel, emissions per company attendee at ASCO represented 75% of a UK resident’s annual carbon footprint.

Of the four congresses examined, ASCO had the highest total carbon emissions (1,664,787 kgCO2e), largely owing to the impact of intercontinental travel (1,569,937 kgCO2e) [Fig. 2a]. For ASCO, p.km for intercity travel were up to 5.5 times higher (3,981,302 p.km) than for ENETS (2,209,672 p.km), ESMO (2,293,997 p.km), and WCNR (722,219 p.km).

Intercity travel emissions were lowest for WCNR, to which a substantial proportion of company attendees (76/181) were modelled as travelling by train (Fig. 3a). Most intercity flights to European congresses were economy class because most company attendees lived in Europe (Fig. 3a); however, the total distances travelled by business class were greater than those travelled by economy class for ESMO and WCNR (Fig. 3b). At ENETS, distances travelled were lower for business class than for economy class. However, because of the substantial carbon emissions associated with this form of transport, business-class travel was associated with the highest levels of carbon emissions for all three European congresses (Fig. 3c). Overall, carbon emissions per p.km travelled were similar across all three European congresses (range 0.263–0.293 kgCO2e/p.km) (Fig. 3c).

Intercity travel associated with in-person attendance at European congresses by a number of attendees, b distance travelled, and c carbon emissions, for pharmaceutical company attendees. Carbon emissions from train travel are so low that they are not always clearly visible on the scale in the graph. Electricity grid emissions for Chicago (0.390 kgCO2e/kWh) are double those for Barcelona and Madrid (0.220 kgCO2e/kWh). The orange line depicts an average across all travel types. ENETS European Neuroendocrine Tumor Society, ESMO European Society for Medical Oncology, kgCO2e kilograms of carbon dioxide equivalent, p.km passenger-kilometres, WCNR World Congress for Neurorehabilitation

Emissions rates (kgCO2e per attendee) associated with taxis were similar across all four congresses (Fig. 4a). Distances (p.km) travelled by taxi for WCNR were approximately 2.5 times lower than for other congresses but this was largely owing to fewer company attendees. Emissions per attendee for intracity travel were about the same for all congresses, except ENETS for which the majority of company attendees were assumed to have walked to the venue.

Carbon emissions associated with in-person congress attendance by a taxi, b hotel accommodation, and c congress venue, for pharmaceutical company attendees. Distance travelled by taxis for ENETS in panel a included 8230 p.km by shuttle bus. ASCO American Society of Clinical Oncology, ENETS European Neuroendocrine Tumor Society, ESMO European Society for Medical Oncology, kgCO2e kilograms of carbon dioxide equivalent, p.km passenger-kilometres, WCNR World Congress for Neurorehabilitation

Total carbon emissions associated with hotel accommodation were higher at ASCO (217.4 kgCO2e per attendee) than at the European congresses (50.3–123.2 kgCO2e per attendee), owing to the fact that hotels in Chicago are typically more carbon intensive than those in Europe (Fig. 2a). Normalised hotel emissions per room-night, based on differences in electricity grid efficiencies between cities, reinforced the difference above, with hotel emissions at 7.2 kgCO2e per room-night in France and 17.6 kgCO2e per room-night in Spain, compared with 36.2 kgCO2e per room-night in the US (Fig. 4b).

Emissions rates (kgCO2e per kWh) associated with cooling loads at congress venues varied across congresses, with up to a tenfold difference between the highest (ASCO, 0.390) and lowest values (WCNR, 0.039) [Fig. 4c]. These differences were due to the numbers of company attendees and local electricity grid emissions factors.

In total, there were 1839 virtual company attendees. We calculated total carbon emissions to be 19,095 kgCO2e (Fig. 5a). Internet energy use accounted for 93–94% of the total carbon emissions per congress, with office energy accounting for the remaining emissions.

Total carbon emissions associated with virtual congress attendance by a contributing activity and b congress, for pharmaceutical company attendees. *The average annual carbon footprint of a UK resident is 5600 kgCO2e [38]. Values are subject to rounding. ASCO American Society of Clinical Oncology, ENETS European Neuroendocrine Tumor Society, ESMO European Society for Medical Oncology, kgCO2e kilograms of carbon dioxide equivalent, WCNR World Congress for Neurorehabilitation

Carbon emissions per company attendee ranged from 7 to 14 kgCO2e across all four congresses, representing 0.12–0.25% of the average UK resident’s annual carbon footprint (Fig. 5b). Emissions per attendee were driven largely by the volume of content delivered via the congress platform (i.e. the total number of hours delivered, evaluated using the programmes for each congress). In a sensitivity analysis, changing virtual congresses from a high bandwidth (24,511 kgCO2e) to a low bandwidth (13,680 kgCO2e) would result in a 44% reduction in carbon emissions. While low bandwidth presents a carbon saving compared with high bandwidth, the impact may not be worth the drop in transmission quality.

Virtual congress attendance was associated with substantially lower carbon emissions per company attendee (mean 10.4 kgCO2e) than in-person attendance (mean 1894 kgCO2e) (Fig. 6). Per-attendee emissions associated with virtual attendance accounted for approximately 1% of those associated with in-person attendance across all four congresses (ASCO, 0.3%; ENETS, 0.6%; ESMO, 0.8%; WCNR, 1.1%). Per-person taxi emissions alone (mean 17.4 kgCO2e) were similar to the total per-person carbon emissions associated with virtual attendance. A total of 3,243,478 kgCO2e was saved by company attendees moving from in-person to virtual attendance at just these four congresses.

Comparison between in-person and virtual congress emissions for pharmaceutical company attendees. *The average annual carbon footprint of a UK resident is 5600 kgCO2e [38]. Data are averaged across all four congresses. Values are subject to rounding. Circle diameter depicts average per-person carbon emissions. kgCO2e kilograms of carbon dioxide equivalent

In a separate whole-congress analysis, with emissions considered for 67,300 anticipated attendees across all four congresses, we calculated whole-congress rough order of magnitude (ROM) emissions to be 127,515,000 kgCO2e for in-person attendance and 640,700 kgCO2e for virtual attendance (Online Resource 3). For virtual attendance, ROM emissions per attendee (mean 9.5 kgCO2e) were equivalent to 0.5% of those for in-person attendance (Online Resource 3). The carbon footprint of holding the ASCO, ENETS, ESMO, and WCNR congresses in person is equivalent to the yearly emissions for 14,448, 541, 6766, and 1014 UK residents, respectively. The carbon footprint of holding these congresses virtually is equivalent to the yearly emissions for 89, 1, 22, and 2 UK residents, respectively. Similar to the main analysis of company attendees, the main driver of in-person emissions was air travel. The main driver of virtual emissions was webhosting, accounting for 71–88% of the total ROM emissions. If green webhosting (assumes a zero-carbon footprint) was assumed, we calculated whole-congress ROM emissions for virtual attendance to be 94,350 kgCO2e, representing an 85% reduction.