There are many topologies that can be used to power LEDs.  But as you probably already know, it all starts with having the design requirements in place before you start.  Without this, you may end up with a design that is less than optimal, or worse, one that may not always operate properly.  For example, when driving one or more LEDs, the LED’s minimum and maximum forward drop, current level and operating temperature determine what the converters output voltage range needs to be.  In looking at a typical datasheet for a red LED for example, I find there can be a 35% variation in its forward voltage drop at its optimal drive current.  This variation drops to a more reasonable 6% if the parts are “grouped” or “binned” by the LED manufacturer for forward voltage.  Additionally, the relative forward voltage drop varies 13% over operating temperature and selecting the LEDs set current can contribute another 16%. 

Okay, so what does this all mean?  Armed with this information, determine what the converter's minimum and maximum output voltage will need to be.  This is simply the sum of all the LED forward drops plus the sense resistor voltage.  Based on the converters' input voltage range, determine if the output voltage is always going to be greater, less, or somewhere in-between.
 

Four common types of topologies to power LEDs
       
Boost
As its name implies, the boost converter output is always greater than its input voltage and is often the most efficient of the converters shown above.  Advantages include a clamped FET voltage (which minimizes ringing and noise) and low input ripple current (for a small input cap). Disadvantages are higher (than output) currents in the inductor, FET and diode, potentially higher losses in the FET (due to high voltage switching and higher current) and pulsating currents in the output cap.  For an output of much less than 1A, these disadvantages are often minimal. For a boost design example, see here. 

Buck
When driving a small number of LEDs from a higher input voltage, go with the buck converter.  It is generally quite efficient with the smallest overall footprint.  Advantages include component stress levels of Vin max or less and Iout or less, and low output ripple current (for a small output cap).  Disadvantages are pulsating input currents and potential ringing on the FET switch voltage.  A cool thing to do is to put the output cap in parallel with the LEDs (rather than to ground) to simplify the control loop. For a buck design example, see here.

Inverting Buck-Boost
Inverting buck-boost is one of the “somewhere in between” topology options, meaning that the converter acts as a buck or boost.  Even though the output is negative, the LEDs don’t really care so don’t let that stop you.  Because the controller is referenced to –Vout, the current sense resistor feedback is connected like a standard buck.  In fact, it’s often hard to spot differences between it and a buck because many of the connections are the same, but the operation of the circuit is not.  Advantages include using a buck (or sync-buck) controller, but getting the buck-boost function!  Disadvantages are lower efficiency, bigger current levels similar to a boost, and level-shifting is required if access to the controller enable is needed. For a inverting buck-boost design example , see here. 

SEPIC
SEPIC is a non-inverting buck-boost topology.  The inductors can either be coupled (like a transformer) or completely separate.  Advantages include clamped switching operation like a boost (great for low noise and high switching frequency) and low input ripple current (for a small input cap).  Disadvantages are maximal use of components, lower efficiency, highest current stress levels, and least understood with a complex control loop (but generally not an issue at lower bandwidths). For a SEPIC design example , see here.