That is a good question. Digital cameras cannot effectively support ISOs more than about two stops below the rated value, and even then, it risks image quality. When you switched from (say) ASA 160 to ASA 20 film, that was like changing to a sensor with three stops lower native sensitivity. But today, we can't change the camera sensor.
We can select a lower ISO, within limits. But the sensor has a single fixed native sensitivity, with the exception of dual-gain sensors, which have two. Some cameras let you go (say) two stops below the lowest rated ISO, but there is always an asterisk or warning. That is because the sensor is really still at the lowest base ISO, say 800. When you select ISO 200, the sensor is still exposed at 800 from the standpoint of "full well capacity." It may show you a darker LCD image and lower waveform or histogram, but you have no more real headroom than before.
The camera may try to hide this by digitally manipulating the sensor data or in some cases using negative analog gain, but this cannot hide the lower dynamic range.
In the film days the closest similarity would be shooting ASA 160 film, setting the camera ASA dial to 20 (which overexposes three stops), then pull-processing that by three stops during chemical film development. This works better than digital because film has better, smoother highlight roll-off characteristics. But that still negatively affects the overall film image quality. With digital, it's worse -- once it's clipped, it's gone.
In broad conceptual terms, pull-processing film was more like shooting a log profile in digital. Both methods capture a wider dynamic range in a lower-contrast format but require reintroducing contrast later in post.
There were no dual-native sensors until the late 2000s. Aptina Imaging Corp developed and patented this around 2006 (US7075049B2). The dual-gain approach was marketed as “DR-Pix” (Dynamic Response Pixel). Their patent was later acquired by Sony. Canon's simultaneous dual-gain approach was separately developed and has different characteristics.
Despite these advances, I'm not aware of any sensor-based tricks that eliminate the need for ND filters. There are apparently fundamental limitations that currently prevent multi-gain sensors from having multiple widely-spaced native sensitivity levels of (say) ISO 50, 800, 5000 and 12,800. Each level would require additional circuitry for *each* pixel photosite and would probably entail other adverse impacts.
The problem with screw-on ND filters is that they must be removed to get back to zero attenuation, and they often mechanically interfere with the lens hood. Even the highest-quality front-mounted variable NDs can introduce artifacts in certain lighting situations. There are variable ND adapters, but they often require even more juggling to remove them for a zero-attenuation shot.