If you work in Neonatology you talk about caffeine endlessly whether addressing your own consumption or that of your patients. Our unit like many others typically starts caffeine early in a prophylactic fashion for babies under 32 weeks. The reason for this is due to the high incidence of apnea (AOP) in this population. The incidence of AOP follows an inverse relationship with GA, affecting 50% of those born at 30 weeks versus 7% at 34-35 weeks. Using caffeine in a more therapeutic approach is certainly sensible for babies older than 32 weeks given the lower likelihood of events. People who argue for a prophylactic strategy cite lower incidence of BPD and PDA from the CAP study while those on the side of reserving treatment for those who really need it put more emphasis on potential adverse effects such as polyuria, tachycardia and electrolyte disturbances. I have already revealed my bias as our centre takes a prophylactic approach but the question is whether it is still right given that the CAP study was many years ago now. A new study emerged this week that was shared on Twitter and generated some interesting discussion. So interesting in fact that I thought there was no better place to discuss this paper than here.
Head to Head Comparison
Elmowafi M et al published Prophylactic versus therapeutic caffeine for apnea of prematurity: a randomized controlled trial this month. This trial included 90 infants in the prophylactic arm (caffeine started < 72 hours of age) vs 91 in the therapeutic arm (only started in presence of apnea or if a baby needed mechanical ventilation and all infants were under 32 weeks of age at birth. Unfortunately, while the study was prospecitve it was not blinded although the authors state that in the therapeutic arm a neonatologist who was not part of the study team made the diagnosis of AOP. Caffeine dosing was standardized using the citrate form at 20 mg/kg as a load and with 10 mg/kg maintenance dosing (divide by 2 for base equivalents).
The primary outcome was total duration of oxygen support and a sample size was chosen based on a previous pilot RCT and the expectation that they could show a reduction in O2 requirement of 7 days in the prophylactic group. Looking at the baseline characteristics of the groups the only difference was the median day of starting caffeine which for the prophylactic group was day 1 and for therapeutic day 3.
What happened in both groups and the reasons for starting caffeine are shown in teh following diagram.
As you can see, 13 patients were started due to need for initial mechanical ventilation, 28 for late mechanical ventilation with the rest being apnea as a reason except for 18 patients who never needed caffeine. When looking at the whole patient population the primary outcome was no different between the groups using an intention to treat analysis.
What the authors did next is where the controversy comes from but has some interesting data nonetheless. They did a “per-protocol analysis” on the whole group and found the following:
Per protocol analysis showed that prophylactic caffeine therapy was associated with a significant reduction in the duration of oxygen therapy compared to therapeutic initiation among included preterm infants [median and IQR of 28 (18–36) days versus 34 (23–51) days, p¼.005 respectively].
Essentially what this means is they said let’s exclude those babies who never needed caffeine and just compare babies that all received caffeine in either arm. Now they see a difference of 6 days which is pretty close to the 7 days expected . Purests will note though that by doing so you can’t claim that the groups were balanced in terms of risk factors since the presented baseline characteristics are not necessarily still the same.
They went further though and asked the question. “What about a subgroup of <29 weeks?”. These babies are undoubedly higher risk for AOP and ventilation. Is the problem that the inclusion of babies 30-31 weeks diluted a potential benefit? Here is what they found.
The main two things to draw your attention to are the marked reduction in mild to moderate BPD and shorter length of stay in hospital. Moreover when they looked at risk of BPD for the whole group they found a reduction in risk. Importantly there was no difference in adverse effects from caffeine use in either arm of concern.
What Can We Take From This?
Any discussion about adverse effects needs to acknowledge that these are short term outcomes we are looking at. Do babies who have never received caffeine do better in the long run? The CAP study did not suggest that was the case but it was a different time with different babies. Do I think caffeine is harmful? No I don’t, especially in the babies who would otherwise experience multiple apneic events or prolonged ventilation.
What this study does raise in my mind though is whether a change in approach to prophylactic caffeine is warranted. The authors don’t specify (I don’t think) but I suspect the bulk of the babies who did not need caffeine were in the 30-31 week group. We know that positive pressure ventilation is harmful to the developing preterm lung. Providing caffeine in the first few hours of life in the hopes of avoiding invasive ventilation or shortening the course of ventilation is the main reason that we likely see consistent reductions in BPD with its use. Under 29 weeks I would certainly favour giving caffeine early but above that as the risk of BPD drops considerably is a watch a wait approach a reasonable one to take? As this study was unblinded it is possible the babies were treated differently with respect to ventilation and apnea managment but I do think the risk is low. If the attendings caring for the babies were not part of the study and knew they were free to give caffeine if needed there may be less risk of bias.
In the end the results of this study in essence are supportive of my practice. I suspect they support many others as well. To answer the question really about side effects I suspect we would need much larger sample sizes than this with adequate power to detect differences but for now I will continue to say “give it and give it early!”
Use of caffeine in the NICU as a treatment for apnea of prematurity is a topic that has certainly seen it’s fair share of coverage on this blog. Just when you think there is an aspect of treatment with caffeine that hasn’t been covered before, along comes a new paper to change my mind.
The Caffeine for Apnea of Prematurity study or CAP, demonstrated that caffeine given between 3-10 days of age reduced the incidence of BPD in those treated compared to those receiving placebo. As an added benefit, in follow-up studies of these patients there appeared to be a benefit to neurodevelopmental outcomes as well at 18-21 months but this was lost by school age with groups being equivalent. In recent years evidence has mounted that starting caffeine earlier in the time course (<3 days and in many cases in the first hour after birth) has led to less need for intubation and BPD. What has really not been known though is whether the use of caffeine in this way might have any long term benefits aside from these short term outcomes.
Dr. Abhay Lodha from Calgary and a group of researchers led by Prakesh Shah from the Canadian Neonatal Network using our robust Canadian network data have tried to answer this with their paper Early Caffeine Administrationand Neurodevelopmental Outcomes in Preterm Infants
The group studied were <29 weeks’ gestation born between April 2009 and September 2011 and admitted to Canadian Neonatal Network centres. As defined in the paper “Neonates who received caffeine were divided into early- (received within 2 days of birth) and late-caffeine (received after 2 days of birth) groups. The primary outcome was significant neurodevelopmental impairment, defined as cerebral palsy, or a Bayley Scales of Infant and Toddler Development, Third Edition composite score of <70 on any component, hearing aid or cochlear implant, or bilateral visual impairment at 18 to 24 months’ corrected age.”
There were 2018 neonates included in the analysis with 1545 in the early group and 563 in the late. It is worth noting that there were 473 infants lost to follow-up meaning that there was about an 80% follow-up rate. Looking at the characteristics of those infants lost to follow-up there were no striking differences that one would expect between them and the group followed.
What did they find?
The odds of BPD (aOR 0.61; 95% CI 0.45–0.81), PDA (aOR 0.46; 95% CI 0.34–0.62), and Severe Neurologic Injury – parenchymal injury or GR III/IV IVH or PVL (aOR 0.66; 95% CI 0.45–0.97) were reduced in the early- caffeine group. The primary outcome was also found to be significantly different as per the table below demonstrating the odds after logistic regression analysis.
So early caffeine seems to be good. Is that all then?
I am very happy to see these results but a few questions remain. Before we get too enthusiastic, I find myself thinking back to the early 2000s after the initial CAP results showed an apparent difference in outcome. The question is whether the reduction in odds seen here for the primary outcome will persist as these children age. Will we see a tendency for the differences to vanish as these children enter school age? I suspect we might but that doesn’t mean all is lost here. What the authors have demonstrated clearly is that early caffeine is not harmful as there is no suggestion of those infants exposed to caffeine so shortly after birth fare worse than those treated later.
Also as the authors state, what isn’t clear is how caffeine works to decrease the risk of developmental impairment. In the discussion they offer some insightful thoughts as to what may be at play and I agree that certainly an anti-inflammatory effect may be responsible for some of the effect. I do wonder though if one could tie the reductions to the lower likelihood of BPD. Development of BPD has been shown many times over to be associated with worse developmental outcomes. Aside from the anti-inflammatory effect mentioned, could the avoidance of early intubation and therefore reduced risk of BPD from positive pressure ventilation be the reason?
In the end if the results persistent into school age, the reason won’t really matter and I hope it does. Will see what happens when we revisit this cohort in a few years but in the meantime I think this paper certainly confirms in my mind the need to give caffeine and make sure it’s provided early!
Much has been written about methylxanthines over the years with the main questions initially being, “should we use them?”, “how big a dose should we use” and of course “theophylline vs caffeine”. At least in our units and in most others I know of caffeine seems to reign supreme and while there remains some discussion about whether dosing for maintenance of 2.5 -5 mg/kg/d of caffeine base or 5 – 10 mg/kg/d is the right way to go I think most favour the lower dose. We also know from the CAP study that not only does caffeine work to treat apnea of prematurity but it also appears to reduce the risk of BPD, PDA and duration of oxygen therapy to name a few benefits. Although initially promising as providing a benefit by improving neurodevelopmental outcomes in those who received it, by 5 and 11 years these benefits seem to disappear with only mild motor differences being seen.
Turning to a new question
The new query though is how long to treat? Many units will typically stop caffeine somewhere between 33-35 weeks PMA on the grounds that most babies by then should have outgrown their irregular respiration patterns and have enough pulmonary reserve to withstand a little periodic breathing. Certainly there are those who prove that they truly still need their caffeine and on occasion I have sent some babies home with caffeine when they are fully fed and otherwise able to go home but just can’t seem to stabilize their breathing enough to be off a monitor without caffeine. Then there is also more recent data suggesting that due to intermittent hypoxic episodes in the smallest of infants at term equivalent age, a longer duration of therapy might be advisable for these ELBWs. What really hasn’t been looked at well though is what duration of caffeine might be associated with the best neurodevelopmental outcomes. While I would love to see a prospective study to tackle this question for now we will have to do with one that while retrospective does an admirable job of searching for an answer.
The Calgary Neonatal Group May Have The Answer
Lodha A et al recently published the paper Does duration of caffeine therapy in preterm infants born ≤1250 g at birth influence neurodevelopmental (ND) outcomes at 3 years of
age? This retrospective study looked at infants under 1250g at birth who were treated within one week of age with caffeine and divided them into three categories based on duration of caffeine therapy. The groups were as follows, early cessation of caffeine ≤ 14 days (ECC), intermediate cessation of caffeine 15–30 days (ICC), and late cessation of
caffeine >30 days (LCC). In total there were 508 eligible infants with 448 (88%) seen at 3 years CA at follow-up. ECC (n = 139), ICC (n = 122) and LCC (n = 187). The primary outcome here was ND at 3 years of age while a host of secondary outcomes were also examined such as RDS, PDA, BPD, ROP as typical morbidities. It made sense to look at these since provision of caffeine had previously been shown to modify such outcomes.
Did they find a benefit?
Sadly there did not appear to be any benefit regardless of which group infants fell in with respect to duration of caffeine when it came to ND. When looking at secondary outcomes there were a few key differences found which favoured the ICC group. These infants had the lowest days of supplemental oxygen, hospital stay ROP and total days of ventilation. This middle group also had a median GA 1 week older at 27 weeks than the other two groups. The authors however did a logistic regression and ruled out the improvement based on the advanced GA. The group with the lowest use of caffeine had higher number of days on supplemental oxygen and higher days of ventilation on average than the middle but not the high caffeine group. It is tempting to blame the result for the longer caffeine group on these being babies that were just sicker and therefore needed caffeine longer. On the other hand the babies that were treated with caffeine for less than two weeks appear to have likely needed it longer as they needed longer durations of oxygen and were ventilated longer so perhaps were under treated. What is fair to say though is that the short and long groups having longer median days of ventilation were more likey to have morbidities associated with that being worse ROP and need for O2. In short they likely had more lung damage. What is really puzzling to me is that with a median GA of 27-28 weeks some of these kids were off caffeine before 30 weeks PMA and in the middle group for the most part before 32 weeks! If they were in need of O2 and ventilation for at least two weeks maybe they needed more caffeine or perhaps the babies in these groups were just less sick?
What is missing?
There is another potential answer to why the middle group did the best. In the methods section the authors acknowledge that for each infant caffeine was loaded at 10 mg/kg/d. What we don’t know though is what the cumulative dose was for the different groups. The range of dosing was from 2.5-5 mg/kg/d for maintenance. Lets say there was an over representation of babies on 2.5 mg/kg/d in the short and long duration groups compared to the middle group. Could this actually be the reason behind the difference in outcomes? If for example the dosing on average was lower in these two groups might it be that with less respiratory drive the babies in those groups needed faster ventilator rates with longer durations of support leading to more lung damage and with it the rest of the morbidities that followed?
It would be interesting to see such data to determine if the two groups were indeed dosed on average lower by looking at median doses and total cumulative doses including miniloads along the way. We know that duration may need to be prolonged in some patients but we also know that dose matters and without knowing this piece of information it is tough to come to a conclusion about how long exactly to treat.
What this study does though is beg for a prospective study to determine when one should stop caffeine as that answer eludes us!
Caffeine seems to be good for preterm infants. We know that it reduces the frequency of apnea in the this population and moreover facilitates weaning off the ventilator in a shorter time frame than if one never received it at all. The earlier you give it also seems to make a difference as shown in the Cochrane review on prophylactic caffeine. When given in such a fashion the chances of successful extubation increase. Less time on the ventilator not surprisingly leads to less chronic lung disease which is also a good thing.
I have written about caffeine more than once though so why is this post different? The question now seems to be how much caffeine is enough to get the best outcomes for our infants. Last month I wrote about the fact that as the half life of caffeine in the growing preterm infant shortens, our strategy in the NICU might be to change the dosing of caffeine as the patient ages. Some time ago though I wrote about the use of higher doses of caffeine and in the study analyzed warned that there had been a finding of increased cerebellar hemorrhage in the group randomized to receive the higher dosing. I don’t know about where you work but we are starting to see a trend towards using higher caffeine base dosing above 5 mg/kg/d. Essentially, we are at times “titrating to effect” with dosing being as high as 8-10 mg/kg/d of caffeine base.
Does it work to improve meaningful outcomes?
This month Vliegenthart R et al published a systematic review of all RCTs that compared a high vs low dosing strategy for caffeine in infants under 32 weeks at birth; High versus standard dose caffeine for apnoea: a systematic review. All told there were 6 studies that met the criteria for inclusion. Low dosing (all in caffeine base) was considered to be 5- 15 mg/kg with a maintenance dose of 2.5 mg/kg to 5 mg/kg. High dosing was a load of 5 mg/kg to 40 mg/kg with a maintenance of 2.5 mg/kg to 15 mg/kg. The variability in the dosing (some of which I would not consider high at all) makes the quality of the included studies questionable so a word of warning that the results may not truly be “high” vs “low” but rather “inconsistently high” vs. “inconsistently low”.
The results though may show some interesting findings that I think provide some reassurance that higher dosing can allow us to sleep at night.
On the positive front, while there was no benefit to BPD and mortality at 36 weeks PMA they did find if they looked only at those babies who were treated with caffeine greater than 14 days there was a statistically significant difference in both reduction of BPD and decreased risk of BPD and mortality. This makes quite a bit of sense if you think about it for a moment. If we know that caffeine improves the chances of successful extubation and we also know it reduces apnea, then who might be on caffeine for less than 2 weeks? The most stable of babies I would expect! These babies were all < 32 weeks at birth. What the review suggests is that those babies who needed caffeine for longer durations benefit the most from the higher dose. I think I can buy that.
On the adverse event side, I suppose it shouldn’t surprise many that the risk of tachycardia was statistically increased with an RR of 3.4. Anyone who has explored higher dosing would certainly buy that as a side effect that we probably didn’t need an RCT to prove to us. Never mind that, have you ever taken your own pulse after a couple strong coffees in the morning?
What did it not show?
It’s what the study didn’t show that is almost equally interesting. The cerebellar hemorrhages seen in the study I previously wrote about were not seen at all in the other studies. There could be a lesson in there about taking too much stock in secondary outcomes in small studies…
Also of note, looking at longer term outcome measures there appears to be no evidence of harm when the patients are all pooled together. The total number of patients in all of these studies was 620 which for a neonatal systematic review is not bad. A larger RCT may be needed to conclusively tell us what to do with a high and low dosing strategy that we can all agree on. What do we do though in the here and now? More specifically, if you are on call tomorrow and a baby is on 5 mg/kg/d of caffeine already, will you intubate them if they are having copious apneic events or give them a higher dose of caffeine when CPAP or NIPPV that they are already on isn’t cutting it? That is where the truth about how you feel about the evidence really comes out. These decisions are never easy but unfortunately you sometimes have to make a decision and the perfect study hasn’t been done yet. I am not sure where you sit on this but I think this study while certainly flawed gives me some comfort that nothing is truly standing out especially given the fact that some of the “high dose” studies were truly high. Will see what happens with my next patient!
This has been a question that has befuddled Neonatologists for years. Get ten of us in a room and you will get a variety of responses ranging from (talking about caffeine base) 2.5 mg/kg/day to 10 mg/kg/day. We will espouse all of our reasons and question the issue of safety at higher doses but in the end do we really know? As I was speaking to a colleague in Calgary yesterday we talked about how convinced we are of our current management strategies but how we both recognize that half of what we think we know today we will be questioning in 10 years. So how convinced should we really be about caffeine?
Even the Cochrane Review Suggests There Is Something Amiss
Back in 2010 the Cochrane Collaboration examining 6 trials on caffeine for treating apnea of prematurity concluded “Methylxanthine is effective in reducing the number of apnoeic attacks and the use of mechanical ventilation in the two to seven days after starting treatment.” Notice the bolded section. Two to seven days. Interesting that we don’t see the effect last in perpetuity. Why might that be? Do babies become resistant with time or is there a change in the way these infants metabolize the drug such that levels in the bloodstream drop after that time point. It is almost certainly the latter and in the last 7 years have we really seen any response to this finding? I would say no for the most part although I don’t work in your unit so hard to say for sure. At least where I practice we pick a dose somewhere between 2.5-5 mg/kg/day and give a load of 10 mg/kg when we start the drug. From time to time we give a miniload of 5 mg/kg and may or may not increase the dose of maintenance based on the number of apneic events the babies are having. What if we could be proactive instead of reactive though. Do the babies need to have multiple events before we act or could we prevent the events from happening at all?
Proactive Treatment With Caffeine
We have known that caffeine clearance increases with postnatal age. The half-life of the drug shortens from about a week at the earliest gestational ages to 2-2.5 days by term equivalent age. For those infants who are older such as 32 weeks and above we expect them to be off caffeine (if they need it) within 2-3 weeks so I am not really talking about them but what about the babies born earlier than that or certainly MUCH earlier at 23 and 24 weeks who will be on caffeine possibly till term. Should one size (dose) fit all? No it really shouldn’t and some crafty researchers led by Koch G have published a paper that demonstrates why entitled Caffeine Citrate Dosing Adjustments to Assure Stable Caffeine Concentrations in Preterm Neonates.
In this paper the authors armed with knowledge of the half life of caffeine at different gestational ages were able to calculate the clearance of the drug at different postnatal ages to demonstrate in a model of a 28 week male infant weighing 1150g. The authors further took into account predicted weight changes and were able to calculate what the expected caffeine levels would be in the fictional infant at various time points. The target caffeine levels for this patient were a trough level of 15 -20 mg/L which are the currently acceptable ranges in the literature. The testing was first done using a standard load of 10 mg/kg (base) followed by 2.5 mg/kg/d (base) and demonstrated levels which yielded the following graph over time. What this demonstrates is that if the dose is unchanged over the first 7 weeks, this hypothetical infant will only achieve effective concentrations for the first week. Interesting isn’t it that the Cochrane review found clinical effect over the first 2-7 days? What if you were to double the dose to really “hit” the infant with a good dose of caffeine from the start and maintain at that level based on their weight gain as shown next. Well, you will get what you are hoping for and keep the trough level above 15 mg/L but you will hit 30 mg/L that some have said is too high and can lead to adverse effects (ever seen SVT with these high doses? I have). Like Goldilocks and the Three Bears could there be a dosing strategy that might be just right? The authors put in another model based on the knowledge of caffeine clearance over time and suggested a strategy in which after the first week the adjusted maintenance doses would be 3 mg/kg/day and 3.5 mg/kg/day in the third to fourth weeks and lastly 4 mg/kg/d in the 5th to 8th week. Using that dosing schedule the model produced this curve. As you can see, the infant would have a therapeutic target without reaching levels above 30 mg/L and potential for side effects. As many of you read this however you may ask the obvious question. Each of us have seen infants who require higher doses than this to rid themselves of significant apnea and escape reintubation. Given that this is a mathematical model it assumes that this fictional infant will respond beautifully to a trough level of 15 to 20 mg/L but some will not. Even in the curve shown it is clear that there is some room to go higher in the dosing as the curve is just touching 20 mg/L.
A Suggestion For The Future
What grabbed my attention here is the possibility that we could take a proactive rather than reactive approach to these infants. Once a small baby is controlled on their dose of caffeine whether it is 2.5, 3, 5 or even 6 mg/kg/d of caffeine should we wait for more events to occur and then react by increasing caffeine? What if we are too late to respond and the patient is intubated. What effect does this have on the developing lung, what about the brain that is subjected to bradycardic events with resultant drops in cardiac output and cerebral perfusion. Perhaps the solution is to work with our pharmacists and plan to increase dosing at several time points in the infants journey through the NICU even if they aren’t showing symptoms yet. No doubt this is a change in approach at least for the unit I work in but one that should start with a conversation!