Inflammation, Acupuncture, and HPA axis: Faulty Science Clouds Understanding

In the wake of my recent critique of acupuncture being touted as a remedy for allergic rhinitis, I was pointed (via a Twitter comment) towards a 2013 review in Evidence Based Complementary and Alternative Medicine, which purported to propose a mechanism for the much-claimed anti-inflammatory effects of acupuncture. There are several putative mechanisms, discussing all of which will make this post gargantuan. Therefore, I shall focus on the explanation involving the hypothalamus-pituitary-adrenal (HPA) axis.

The review, by a group of Australian authors, states in the relevant section unequivocally: “Acupuncture effects on the inflammatory response have been shown to be modulated by the HPA axis in a number of animal studies.” To provide a bit of background, the HPA axis is a term for a functional, interdependent relationship between parts of the brain (the hypothalamus and the pituitary) and the adrenal gland. When soluble mediators of stress (physical stress or immune-induced stress, as in inflammation) stimulate the HPA axis, the adrenal gland releases certain steroid hormones, called corticosteroids, which have potent anti-inflammatory and immunosuppressive properties, and exert a protective effect.

How’s this relevant to acupuncture and animal studies? The claim is that acupuncture stimulates the HPA axis. Corticosterone is the primary form of corticosteroids in rats and other non-human animals; in rat models, inflammation is experimentally induced, and HPA-axis stimulation via acupuncture releases corticosterone, which ameliorates the inflammation. For evidence, the review cites two sequential studies (dated 2005 and 2007) from the same related group, involving researchers from the Center for Integrative Medicine at the University of Maryland in Baltimore, and the Shanghai University of Traditional Chinese Medicine.

It is to be noted, however, that in both studies, the “acupuncture” mentioned in the review is actually done via a procedure called “electroacupuncture” (EA), in which mild electric current is passed through the acupuncture needles following insertion.

Let’s consider the 2007 study¹ in BMC Complementary and Alternative Medicine, in which immune-mediated inflammation was induced in the paws of adult Harlan Sprague Dawley (SD) rats via injection of complete Freund’s adjuvant (CFA, a non-specific inflammatory mixture of antigens derived from the tuberculosis bacteria). EA (10Hz at 3mA and 0.1 ms pulse) administered twice (immediately after and 2h after CFA administration) increased plasma corticosterone levels, compared to controls, and reduced edema from the inflammation. Preventing endogenous corticosterone secretion (by surgical removal of the adrenal glands) or action (by a drug that prevents corticosterone from binding) blocked EA’s edema-reducing action. Therefore, the study concluded that EA stimulates the HPA axis to exert its anti-inflammatory action.

It is a quite a bit of painstaking research work done, funded by an NIH grant, but do the observations really say what the investigators think they say?

Let me discuss some of my major objections. First, I am of two minds about rodent studies in acupuncture. As we know, this alternative medicine modality is based on traditional Chinese medicine (TCM). Using rats in experiments mostly takes away the touchy-feely component of alt-med, and forces the investigator to focus on largely objective measurements instead – which is good. On the other hand, however, the disease conditions experimentally induced in animals may never exactly approximate, or be analogous to, the corresponding human disease.

Besides, more importantly, how do acupuncturists find and use specific “acu-points” on the body of the animals, especially when the acu-points have no anatomical correlates? In the study under discussion, as well as in similar other studies, the positions of the acu-points are approximated based on human acu-points from TCM texts; the rodent acu-points are taken to correspondingly map on the bilateral physical symmetry. But are they really equivalent across species? Where is the evidence?

Also, in clinical studies, the needles are often inserted to the depth where the patient verbalizes feeling a prick or a burning sensation – which is interpreted as the patient’s achievement of ‘de-qi’ or the mystical life force energy envisaged in TCM. How do the rodent subjects indicate accurately – beyond the scope of investigator bias – that they have attained qi?

I am equally irritated by the newfangled modality of EA. The use of electricity – of a specific frequency, strength, and pulse time – is most certainly not derived from the classical texts of acupuncture. So, where did it exactly spring forth from?

Any study using EA, to my mind, absolutely needs to have several rigorous controls, in addition to the EA-recipient test group. There has to be control groups – (a) getting EA in random, and/or unrelated points in the body (to account for non-specific effects); (b) getting regular non-electric acupuncture in the same acu-points as the test subjects, as well as (c) same amount of electricity being delivered – without the needles – to both so-called specific acu-points and non-specific areas; b and c would control for the general effects of electricity on the body. And we do know that electricity does have an effect on stimulating muscles and nerve endings; since neuromuscular transmission activity is electrical in nature, this is hardly surprising. Sadly, most studies – including the one in discussion – fail to include the necessary controls.

Let’s address some specific methodological issues with the study.

• Inadequate controls: only EA and sham EA (defined as non-electric acupuncture) tested; there was no customary saline control for CFA injection.
• According to their observation, CFA-induced inflammation appeared within 2h of injection and peaked between 6-24h. — this, therefore, means that the EA was applied before & around appearance of inflammation.
• Blood from corticosterone measurement was collected before CFA (baseline), at 2h and 5h after CFA — which means, plasma corticosterone estimates were made before reaching peak inflammation. So, it is not surprising that the CFA only control didn’t show appreciable corticosterone release; also possible that in presence of CFA, EA caused extra stress leading to excessive corticosterone release. Interesting to note that EA only caused the next highest corticosterone release at 5h, an observation that hasn’t been adequately highlighted.
• Presentation of corticosterone data as percentage of baseline hides the actual levels which might be high but nonetheless clinically insignificant – we have no way of knowing.
• Sham EA (plain acupuncture) after CFA showed corticosterone levels equivalent to CFA only; is this inadvertent proof that acupuncture – the plain, garden variety without fancy electricity – just doesn’t work?

Another major methodological issue has to do with the choice of the model. The study (as with many other similar studies) used outbred SD rats from Harlan. The rat is a common rodent model used to examine various aspects of the HPA axis and its effects on the immune system; however, the HPA axis response to stress is subject to the genetic background of the specific rodent strain used. For example, Fischer 344 and Lewis rats, both inbred, nonetheless show remarkable differences in their adrenocorticotropic hormone (ACTH) and corticosterone response to immune challenges and other stressors.

Several studies (for instance, see these studies from 1993, 1999, 2004) have investigated in the outbred SD – the maternal rat strain for these inbred ones – the HPA axis response to physical stress (immobilization), mild electric shock (0.5mA, 6s bursts/90s, 30m), inflammatory stimuli (such as, intravenously-given bacterial endotoxin/LPS or cell wall mixture, or interleukin-1; intramuscularly-delivered turpentine), and have made some interesting and persistent observations.

1. Stress- or shock-induced plasma corticosterone levels returned to baseline values in 1-2h; inflammatory stimuli-induced levels may take a little longer (~3-4h).
2. Baseline and stress-induced levels were significantly lower in the morning hours than in the evening hours.
3. Evening baseline plasma corticosterone levels fluctuate greatly from hour to hour.

As can be easily understood, these quirks of the Harlan SD rat model may put a big question mark on the proper interpretation of the observations made in respect of EA’s effect on corticosterone levels, and consequently, of the putative anti-inflammatory effects. To account for these confounding factors, such experiments need to be thought out in great detail and carried out with meticulous care. Sadly, the study (in fact, such studies) using this model do(es) not provide evidence of any awareness of these issues.

Most interestingly, studies have failed to find any long-lasting effect of a single exposure to the stressor. It is also known that in case of chronic stress, rats may become desensitized to the same stress, eliciting a reduced HPA axis response. Therefore, how exactly two (or more) doses of EA would achieve the claimed benefits of corticosterone secretion remains unclear.

It is also problematic to base an entire set of conclusions on a single parameter, such as plasma corticosterone measurement, measured in a single experimental model, such as the Harlan SD rats. Investigators researching HPA axis response to stress have reported wide variations in corticosterone output within experimental groups, within SD rats obtained from different vendors, and within SD rats of different batches from the same vendor.

The problem is accentuated by the observation that immune-challenge mediated inflammation may be resolved as a result of acute stress even without the involvement of the HPA axis; acutely stressed (via mild electric shock) SD rats, in which dead bacteria were injected under the skin to generate an inflammatory immune response, showed lower degree of inflammation and cleared the inflammation sooner compared to non-stressed controls. Interestingly, this effect was seen equally even in adrenalectomized (ADX) rats.

In the present study, therefore, it would have been interesting to see a side-by-side comparison of health and ADX rats’ response to CFA-inducted inflammation and effect of EA. Unfortunately, without providing this dataset, the authors referred to their previous paper (2005) for the non-ADX data. However, visually comparing the relevant figures (only baseline and 5h points were common), it appears that at baseline the ADX rats already have higher edema as measured by hindpaw size, compared to non-ADX rats, but the rise in edema was greater in the latter. It is difficult to make sense of this data coming from two separate experiments, both with relatively small sample sizes.

As is now hopefully clear, the research endeavor that has gone into this work appear full of serious lacunae, which undermine the validity of the conclusions drawn, as well as the overall believability and plausibility of acupuncture as a scientific modality.

That, of course, yet again brings to my mind what Dr. Harriet Hall, physician/science communicator, once referred to as Tooth Fairy Science.

You could measure how much money the Tooth Fairy leaves under the pillow, whether she leaves more cash for the first or last tooth, whether the payoff is greater if you leave the tooth in a plastic baggie versus wrapped in Kleenex. You can get all kinds of good data that is reproducible and statistically significant. Yes, you have learned something. But you haven’t learned what you think you’ve learned, because you haven’t bothered to establish whether the Tooth Fairy really exists.

Aficionados of Tooth Fairy Science suffer from cognitive biases; if their observations appear consistent with their initial hypothesis, they tend to stop looking and don’t bother satisfying the principle of falsifiability required by the scientific method. The resultant work may be grandiose wish-fulfillment, but it certainly is not science.