Technology Helps Explain Medication Failure

In almost every area of medicine, doctors can order tests to provide objective physical data to guide their medication selection. However, the practice of psychiatry is most often based on observation, self-report and psychological testing. It appears that we are better at measuring impairment than we are at identifying the source and prescribing an effective medication. Is there a way we can do better?

The director of the National Institute of Mental Health, Tomas Insel, suggests there are many medicines, but they are not working adequately. This is because the symptoms of mental illness are too illusive and are shared by many diagnoses. Insel (2012) says, “It’s much harder to fix something if you don’t know what is going wrong.” Medications are being prescribed to treat a set of symptoms suggestive of a specific disorder without any objective evidence of the cause. Additionally, the practice of polypharmacy has become way too common in children and adolescents.

Pharmaceutical industry advertising promotes adding a medication when the first medication fails to produce the desired results (i.e., adding Abilify to your antidepressant). The message is that when one medication fails, keep adding more in an effort to address the additional symptoms. Each additional medication increases the risk of side effects. It is not uncommon for children to come to us with several medications prescribed. Last month, for example, we saw a 9-year-old female with prescriptions for Olanzapine three times a day, Lithium Carbonate daily and Amphetamine Salts three times a day. Also, a 10-year-old male came to us on Focolin three times a day, Seroquel twice a day, Lexipro daily and Zyprexa daily. If there was a way to determine why a medication failed, would it not be prudent to investigate why? If current technology could help?


Electroencephalograms (EEG) and quantitative EEG (qEEG) have been studied with regard to their application of predicting response of psychotropic drug treatment and monitoring drug toxicity. Pharmaco-EEG is a scientific discipline that started in the 1960s (Mucci et al). These two lofty goals grew out of over 600 papers, published between 1952 and 1961, on the EEG effects of psychoactive drugs (Galderisi & Waters, 2006). However, individual drug profiles and the limited number of parameters used to characterize the EEG signals were just two factors that thwarted development.

In the early 1970s, technological developments enabled analog EEG data to be transformed into digitized data. Two pioneers, E. Roy John and Robert W. Thatcher, created the first normative database to allow Z-Score comparison; thus, quantitative EEG (qEEG) brain mapping was born. Their seminal work provided a detailed description of how behaviorally relevant functions are mediated by the neural system. Since then, each psychiatric medication has had to be tested on normal’s to see how it would affect their qEEGs.


Suffin and Emory (1995) published the first large-scale QEEG study delineating each psychoactive drug’s effect on normal controls. Johnstone and Gunkelman (2005) documented “clusters” of qEEG features in psychiatric populations in order to empirically define clusters of individuals who may be responsive to specific psychoactive medication. These “clusters” of qEEG profiles became known as intermediate phenotypes (i.e., manifestations seen between the genome and behavior). “These intermediate phenotypes are highly heritable and are reliable indices of brain function, are not isomorphic with DSM categories and have implications for therapeutic intervention” (Johnstone & Gunkelman, 2005, p. 102).

The following table lists the qEEG patterns/phenotypes and their medication implications:

QEEG Profile Description of Pattern Medication
Diffused slow activity, with or without low frequency alpha Increased (1-7 Hz) with or without slow posterior dominant rhythm Stimulant
Focal abnormalities, not epileptiform Focal slow activity or focal lack of activity Unknown
Mixed fast and slow Increased activity below 8 Hz, lack of alpha, increased beta frequency activity Combined across classes, e.g., stimulant + anticonvulsant
Frontal lobe disturbances Frontally dominant excess theta or alpha frequency activity Antidepressant, stimulant
Frontal asymmetries Variable asymmetry L>R or R>L, primarily at F3, F4 Antidepressant
Excess temporal lobe alpha Increased alpha activity generated in temporal lobe Stimulant
Epileptiform Transient spike/wave, sharp waves, paroxysmal EEG Anticonvulsant
Faster alpha variants, not low voltage Alpha frequency greater than 12 Hz over posterior cortex Unknown
Spindling excessive beta High-frequency beta with a spindle morphology, often with an anterior emphasis Anticonvulsant
Generally low magnitudes (fast or slow) Low voltage EEG overall Metabolic support neutraceuticles
Persistent alpha with eyes open Lack of appreciable alpha blocking with eyes open Unknown

Johnstone and Gunkelman (2005) suggest that the data generated by the EEG/QEEG to predict medication response, besides being more objective, may help to reduce side effects and give insight into the dynamic balance required when polypharmacy is indicated. However, in 2005, they had limited experience with predicting medication response.


When starting this venture, we had hoped that EEG/qEEG data would help us with medication selection. In many cases, it did. However, the greatest utility of EEG/qEEG data has been to help us identify why medication fails. Although on the surface this seems disappointing, it has led to some significant positive outcomes.

In our center, the primary reason someone is referred for an EEG/QEEG is because medications either fail to work and/or the side effects are intolerable. In the past 5 years, we have done over 350 studies and each one patterned for medication. The analysis of data from these studies revealed a common failure theme.

Our research suggests that medication most often fails when the abnormal brain pattern does not match the medication normally prescribed for the diagnosis. For instance, multiple studies suggest that frontal alpha patterns that are found in a large number of depressed people respond best to SSRIs. However, a smaller number of depressed people (e.g., refractory cases) have focal slowing in the left anterior temporal lobe and they do not respond to any medication. These are also the ones that do respond to transcranial magnetic stimulation and transcranial direct current stimulation. These are neuromodulation techniques that use electrical fields or current to stimulate focally slowed areas. Four distinct EEG phenotypes account for a majority of medication failure, as described in the following sections.


Transients are seen as a sharp wave or spike and wave discharges that are often paroxysmal in nature. These bursts are often seen focally; however, they can occur more diffusely. Whether epileptiform (resembling epileptic discharges) or not, this activity indicates an unstable brain. Although considered a normal variant, we have found a high correlation between location of the discharges and idiosyncratic pathology. For instance, left posterior temporal transients (Wernicke’s area) are highly correlated with language issues and right anterior temporal transients are highly correlated with impulsivity and explosive episodes in children. This pattern is seen in over 40% of children on the Autistic Spectrum and 25% of children with ADHD.

Implications for Intervention

The first consideration in these cases is to reconsider any medication that reduces seizure threshold. The data show these medications are likely to make the discharges worse and have a negative impact on function. These brains are in need of stabilization by an anticonvulsant. The background frequency helps to guide which anticonvulsant is preferable. An abnormally slow background frequency would be helped more with Trileptal (tends to speed up the background). With normal to fast background frequency brains, Lamictal (more for generalized activity) and Keppra (more for temporal lobe discharges) may be good choices.

The first application of Neurofeedback (NF) in the 1970s was to treat seizure disorders. To date, it is the most thoroughly validated neurofeedback application, with well-designed, randomized trials and double-blinded studies (Sterman, 2000). QEEG-guided NF protocols identify the target location of the transient activity and specify the frequency to be altered in training.


Beta waves have a frequency from 13 to 30 Hz and are normally found when you are alert. Spindling beta is defined as higher voltage (magnitude) beta that is distinctly seen in waxing and waning spindles, primarily in the anterior cortex, and is best considered a non-specific sign of dysfunction or encephalopathy. It is associated with ‘cortical irritability,’ viral or toxic encephalopathies and in epilepsy. This pattern is seen in about 10% of the ADD/ADHD cases and mood disordered population. Other causes of spindling beta would be sedative abuse, particularly the benzodiazepines class of medications. In cases of benzodiazepine abuse, I have seen this pattern remain a year after last dose. Iatrogenic (caused by medical intervention) causes must be ruled out when spindling beta is observed.

Implications for Intervention

Medications that increase beta activity would make spindling beta worse. These include any medications that increase norepinephrine and dopamine. These brains are in need of GABA to calm the spindles. Medications that have been found effective with this pattern are those that increase GABA (Gabapentin, Lyrica), channel blockers (Clonidine and Intuniv) and anticonvulsants, including Depakote.

Beta suppression directly in the center of the area of spindles responds well to neurofeedback and has shown good clinical response. This is especially true when the QEEG is used to identify the target location (center of the beta spindles) and the specific frequency band of the beta spindles.


Encephalopathy is defined as diseased, damaged or malfunction of the brain that is manifested as an altered state. In the EEG, this is seen as a “diffuse low voltage slow” pattern. There are 185 listed causes of encephalopathy; but, what we see most often are caused by metabolic issues, toxic exposure, anoxic (oxygen deprivation), electrolytic imbalances or traumatic injury.

Implications for Intervention

Diffuse low voltage slow brains do not respond well to medication. In fact, stimulants are often used to help energize the underaroused individual with very limited results and often accompanied with unacceptable side effects. In these cases, identifying the cause of the encephalopathy is primary. Here, we may refer to an endocrinologist or an internist can be an invaluable referral. Those who have malfunctioning thyroids can be effectively treated medically. Other interventions that have shown promise are hyperbaric oxygen treatment (Rossignol, 2006) and Interactive Metronome (our findings).


Focal slowing is simply a predominance of slow wave activity in a specific area, as opposed to a more diffuse (generalized) finding. There are two types of focal slowing. Focal rhythmic or semi-rhythmic slowing into the delta range (1-4 Hz) has been closely associated with deeper white matter issues, such as tumors or lesions. Ones we commonly see are associated with more severe head injuries. Slowing in the high theta, slow alpha range (6 to 9 Hz) is associated with gray matter disturbances. We commonly see these pattern months after the impact of a head injury (as opposed to a blast injury). This is also a common pattern found in those with refractory depression, when the focal slowing is located in the left anterior temporal lobe.

Implications for Intervention

Medication acts upon the whole brain and, if you add enough medication to speed up the slowed area, you make the rest of the brain too fast. And any medication used to treat the anxiety will further slow the focal slow area, thereby, increasing pathology. Maybe this is why they have yet to find a medication to treat psychologically disturbed individuals with head injuries.

QEEG-guided NF targeting the range and location of the slowing has shown great promise, especially in gray matter slowing, and is being used extensively in military populations. White matter slowing (1-4 Hz) is proving to be more difficult to treat. However, transcranial direct current stimulation is showing promise as an emerging non-invasive neuromodulation technique, especially when it is QEEG-guided.


What we have learned in the past five years is that QEEG is a valuable tool for an experienced psychiatrist, who knows how to use QEEG to firstly avoid using medications that are likely to make their patients worse; and secondly, to give insight into which medications may be worth an empirical trial. The search for the answer why medications fail is the impetus of our work.

In almost every case where two or more medicines have failed, we have identified transient discharges, spindling beta, encephalopathy or focal slowing as the reason. Knowing why a medication has failed is vital because sometimes the best we can do is to keep from making the patient worse and to have a non-medication intervention option. The QEEG, however, is just a tool to provide information and was never intended to replace an experienced psychiatrist’s wisdom and judgment. Psychiatrists who know how to use QEEG data have a distinct advantage.

Ron Swatzyna Phd, LCSW
Tarnow Center

Leave a Comment