Nantonos wrote in an earlier post, The Cult of Epona: Origins and Spread:

It is also frequently assumed that this deity must predate the Roman conquest of Gaul in the 50s bce. So far I am unaware of any evidence for this, however. Furthermore, one of the earliest Epona representations is from Pompei in Southern Italy [Jordan], and is clearly very securely dated as it cannot be later than the volcanic eruptions of 79ce. Another early example is an inscription from Entrains-sur-Nohain, which is dated on epigraphic grounds (by its use of "good first century characters") to the end of the first century or the start of the second [Magnen & Thevenot, pp.39-40].

L. S. Oaks, in her 1986 article "The Goddess Epona: Concepts of Sovereignty in a Changing Landscape," notes possible pre-conquest Epona iconography in the Gundestrup cauldron.

The Gundestrup cauldron is a silver-gilt vessel found in a bog near Gundestrup, Jutland in 1891. It measure 14 inches (35.6 cm) high by 25.5 inches (64.8 cm) wide and weighs about 20 pounds (~9 km). Olmsted and Green both date the cauldron to the 1st C. BCE.

Oaks relies on Garrett Olmsted's 1979 work to hypothesize that the cauldron panel showing a woman on a chariot flanked by an elephant, horse, dog, and other animals could be a reference to Epona, and indeed to the Irish goddesses Medb and Mach, and the Welsh Rhiannon.

According to Oaks, Olmsted theorizes that the representation shown above can be connected through iconography and mythological parallels with a primarily Celtic goddess of sovereignty who was also a divine consort of the king. She goes on to say that Olmsted is the first focus the meaning of the cauldron and related history in pre-conquest Gaul.

Among the cluster of known Celtic goddesses connected with this tradition of sovereignty are Irish Medb and Macha, Welsh Rhiannon, and Gaulish Epona -- that is, Epona as an earlier and more powerful avatar than the benign lady of the popular Gallo-Roman cult. Epona was most closely link with Rhiannon, who in turn has clear parallels with Macha; Medb may be regarded as the strongest and most complete representation of the feminine principle of sovereignty in question.

Olmsted builds his case using images from coins to show how the convention of a horse evolved into an elephant as a symbol of power. This iconographic transition possibly started after the Gauls in the south first witnessed Hannibal crossing into their territory with elephants.

There is a lot more information in Oak's article, several points with which I don't agree (which will appear in another post.)

Note: I've requested Olmsted's work through interlibrary loan.

References

Green, M. (1986). The Gods of the Celts. Gloucestershire, Sutton Publishing Limited.

Linduff, K. (1979). Epona: a Celt among the Romans. Collection Latomus. 38: 817-837.

Magnen, R. and E. Thevenot (1953). Épona : déesse Gauloise des chevaux protectrice des cavaliers. Bordeaux, Delmas.

Oaks, L. S. (1986). The Goddess Epona: concepts of sovereignty in a changing landscape. Pagan Gods and Shrines of the Roman Empire. M. Henig and A. King, Oxford University Committee for Archaeology. Monograph no. 8: 77-84.

Olmsted, G. S. (1979). The Gundestrup cauldron : its archaeological context, the style and iconography of its portrayed motifs and their narration of a Gaulish version of Tâain Bâo Câuailnge. Bruxelles, Latomus.