A recent exchange on the Twitter reminded me of an old paper from 1968.
— Drug Monkey (@drugmonkeyblog) December 11, 2015
The paper in question is
Scheving LE, Vedral DF, Pauly JE. Daily circadian rhythm in rats to D-amphetamine sulphate: effect of blinding and continuous illumination on the rhythm. Nature. 1968 Aug 10;219(5154):621-2. [PubMed]
The key takeaway message for me is captured in the first figure (click to embiggen), which represents the percentage of rats that died within 24 h of being injected with either 26 mg/kg (darker line) or 30 mg/kg (dotted line) of amphetamine. The X axis depicts the time of day at which the groups were injected and the bar that forms the X axis indicates when the lights were on (6 am to 6 pm) and off.
As you are aware, rats are a nocturnal species and the wiggle trace just above the X-axis confirms this with activity patterns based on noise recording of the colony.
So, back to the point. The only difference across points within a single amphetamine dose is the time of day at which the drug was administered. Mortality rates change from 20% to nearly 80% with the lowest observed during the inactive part of the rats' day.
Light cycle and circadian phase matter. A lot.
This brings me to a second example, which is from one of the papers in a series of investigations by Dave Roberts at Wake Forest. In
Roberts DC1, Brebner K, Vincler M, Lynch WJ. Patterns of cocaine self-administration in rats produced by various access conditions under a discrete trials procedure. Drug Alcohol Depend. 2002 Aug 1;67(3):291-9. [PubMed]
the authors use a procedure in which rats are allowed to self-administer cocaine 24 h per day. The one major difference from the usual 1-2 h per day type of model is that the number of opportunities for cocaine were limited. These "discrete trial" opportunities ranged from 2-5 per hour and each time the animal was permitted 10 min to make a response once the lever was extended. Each response terminated the discrete trial so animals could only take 2-5 infusion per hour.
The figure that continues the point most effectively is from a set of manipulations in which the discrete trial was set to 3 per hour and the per-infusion dose was varied. The data represent the total cocaine intake per hour so look at the 1.0 and 2.0 mg/kg/infusion doses if you want to figure out how many responses out of the 3 opportunities per hour were being made.
The point is again obvious, namely that circadian factors and light cycle matter a lot to the outcome. Imagine the more typical 1 h or 2 h operant self-administration session for cocaine being placed at various points across the rat's light cycle. On average, you might expect different mean intakes.
This is going to contribute to replication and reliability issues, particularly if you expect a given mean amount of drug intake.
It gets even tricker if you want to start exploring the effect of different interventions on cocaine self-administration. Who knows if they themselves have circadian-dependent effects or if the interaction with cocaine taking does? Who knows which direction it takes? We don't know until someone does the study.
And we can all see how much exacting work with light cycles there will be to satisfy ourselves that we know what the influence is. Work that, should it turn out negative, will be nigh on unpublishable.
And to be clear, there are hard practicalities of research that make us ignore these factors at times. Mostly across studies, but sometimes within them. Take the big issue of running behavior in the light or dark cycle of a rat (or mouse). This depends on University Facilities level decision making. Can the rooms be reverse-cycled (technically or at the whim of the animal care department)? Can you get access to the right light-cycle room for your animals for your experiments if you are low on the totem pole (as a lab or within a lab)?
Then there are within-lab factors. Limited numbers of operant boxes and limited numbers of hands. You cannot necessarily squeeze all of your animal work into the prime window of 6 h into dark to 12 h identified in the Roberts paper, above. Maybe this function changes depending on your procedures and you have an even narrower stability window.
So there will be compromises.
But these compromises will most assuredly affect the perceived replicability (aka generalization) of the work.