Skip to main content

User login

What is not?

 Impossible is not a fact, its an opinion 

Galactorrhea + Amenorrhea

June 1, 2010 by Dr.Inayath

Case Sheet

 

A 30 year old female patient came with complaints of amenorrhea since 3 months and milky breast discharge since 2 weeks. The LMP of the patient is (etc, more facts about amenorrhea from obs/gyn)

The milky discharge is intermittent, not associated with pain or discharge of blood or swelling. Patient is unmarried and is not sexually active. Patient has no complaints of cold intolerance, loss of hair, excessive sleepiness, weight gain, swelling over the neck or (hypothyroid symptoms). Patient is not taking any drugs like antipsychotics or antidepressants. On physical examination, no significant findings found. Vitals: BP – 120/80 mm Hg, PR – 90/min, Temp – 98.6 F

 

Dear Dr,Inayath, your case is published, however it is not complete, please fill in the rules too

 

Case Discussion: 

<!-- @page { margin: 2cm } TD P { margin-bottom: 0cm } P { margin-bottom: 0.21cm } -->

The presumptive diagnosis is that of hyperprolactinemia. The DDX include pregnancy, lactation, hypothyroidism, medications.

The patient says she is not sexually active, so pregnancy is ruled out – had it been otherwise no matter what, we would have a Beta HCG done. If the patient has not been pregnant in recent times, lactation is ruled out, how will we teach this to system? We can add a rule that if patient is not sexually active, pregnancy is strongly negative (-100), if pregnancy is negative , then lactation is negative too. This avoids the system from asking the question “were you recently pregnant” if the patient is not sexually active.

 

The system when given the presenting symptoms i.e galactorhea + amenorrhea fires FC chain DDX i.e Pregnancy (1.4), Lactation (1.4), Hypothyroidism (1.3), Drug reaction (1.3), hyperprolactinemia (1.2).

 

 

The questions of all these diseases are loaded multiplied by the Inc Score given in brackets, so the questions can be sorted and most important ones asked first. This is important because there might be dozens of question with each disease and we cannot ask each and every question. Other diseases where amenorrhea and galactorhea are present may also be triggered and if the user wishes to probe deeper he can ask those questions too.

 


QUESTION

Rule score * Inc

Final score

are you sexually active?”

(1) * 1.4

1.4

do you have fullness of breast?”

(0.4) * 1.4

0.56

were you recently pregnant?”

(1) *1.4

1.4

presence of thyroid swelling”?

(0.4) * 1.3

0.52

intolerance to cold?”

(0.6) * 1.3

0.78

excessive sleepiness?”

(0.6) * 1.3

0.78

 

 

 

Important note to myself – we should include a system where diseases with all symptoms included should be tried first for eg. If there are three symptoms x,y,z – instead of firing all rules which contain individually x,y,z we must first fire rules which contain all three x+y+z. This is important since we are starting with canned questions, and will have lots of facts to start with.

 

Another important point is that of question scoring, if disease A has three facts – x (0.8),y (0.7), z(0.5) and the disease A is triggered it loads all the rules with their scores i.e question weight = question score / total score

x = 0.8/2 = 0.40, y = 0.7/2 = 0.35, z = 0.5/ 2 = 0.25

 

But suppose question x is rejected, the other questions will still be in the list with same scores, we want this to be reduced – when the king falls the soldiers are frightened – the hypothesis is shaken, it has lower value – the formula is

 

question weight = question score / total score + square root of summation function of previous questions scores

 

so if x falls, the score of y will be 0.7/(2+ s| 0.4) = 0.7/2.2 = 0.31

if y too falls the score of z will be 0.5/(2+ s| 0.4+0.35)=0.5/2.8 = 0.175

 

So making this change makes the values fall – y from 0.35 to 0.31 and z from 0.25 to 0.175

 

 

 

 

 

Getting back to the case, we will exhuast all the questions pertaining to the diseases, we find diseases with total scores – pregnancy (-0.7), lactation(-0.7), hypothyroidism (0.3), drugs (-0.8), hyperprolactinemia(0.5). So in this case hyperprolactinemia is the first to be worked out and will be the main diagnosis.

 

Suppose if the patient had answered “active sexual life” as “yes”, pregnancy would have been (0.6), lactation (0.2 or negative depending on how other pregnancy related questions have been answered), hypothyroidism(0.3), drugs (-0.8) and hyperprolactinemia (0.5). So the first to be worked up would have been pregnancy by doing a hcg test.

 

The manipulation and fine tuning of scores require real time testing, so we leave it for now. Then there is another aspect i want to write down – the ability to define the situation completely beforehand, for eg. In this case of amenorrhea + galactorhea syndrome i would like to define a protocol irrespective of which diagnosis is considered, the system needs to switch from normal mode to protocol mode if told.

 

 

The canned questions for this case are :

 

a. Galactorhea – duration? is the milky discharge associated with blood/swelliing? Is there amenorrhea? Sexually active?

b. Amenorrhea – duration? LMP? Sexually active?

Or if the patient said “Menstrual abnormalities” the canned questions would be duration? heavy/less?

 

We dont have to worry about repeated questions as Pyke understands and records what has been answered previously.

 

 

P.S – It will be interesting (and fun!) to build the case sheet generator as the case sheet generator has to know how to group answers and present it in a doctor-readable form.

 

When does the system know to put a break to symptoms and proceed for examination? Its like this – all questions are sorted with scores,if we take the top 2/3rd of the questions as a simple way to decide, we may lose important questions which are at the bottom if the case is a wide DDX case - there might be strong contenders at the bottom of the list, moreover the list is dynamic and being sorted actively. We need a formula here too to mark the threshold, which i will add later.

 

When the system switches from symptoms to signs – first we do a routine “canned” questions of Vitals, pallor, cyanosis, edema etc.

Then we sort the signs derived by the expert system into their respective organ system groups. Once this is done, we are actively sorting all the three lists – we move from signs to investigations.

Investigations require a different set of rules, these rules are again already answered by canned questions i.e is this a rural setup? A PHC? A UHC? Tertiary care center? The investigations that cannot be done in that setup or already marked as “cannot be done, refer to higher center”. The remaining investigations are then used to proceed.

 

 

In this case, the investigations that come up are TSH (0.5), HCG (0.2), S. Prolactin (0.7), Drug levels (0.1), MRI (0.1) etc.

 

Remember at every point the user has total control, we at any point can look at the disease scores and finalize. Suppose he/she stops at examination and the scores at that time are hyperprolactinemia (70%), Hypothyroidism(20%), we may aswell proceed to treat hyperprolactinemia.

 

Important note: What i have illustrated are just 3-4 diseases, we may have many other diseases which can pop up, for eg. Am writing a rule Diagnosis : Chronic Renal Failure if symptoms – Galactorhea + Amenorrhea (20).

(dont worry there will be other ruleset for CRF which will be needed to fulfill the diagnosis, however this makes the disease to be considered)

One good thing about the system is that you can define rules anywhere in any number of pieces the system will string together all and makes scores and weights.

 

Another interesting point is that we can write rules the way we want to eg - “ Chronic renal failure if diagnosis hyperprolactinemia (20).” The system will automatically bring all hyperprolactinemia findings into question stack to test, and if yes, CRF gets a score too. Here the CF of hyperprolactinemia is multiplied with the rule CF to get CRF CF.


I have also read in the book that this syndrome can be caused by seizure, stress, hypoglycemia, sleep, cirrhosis, nipple stimulation, CRF. Is this true? If so, how many episodes or in what conditions should we consider the above conditions for evaluation? Any ideas?

Hyperprolactinemia is associated with Pituatary adenoma which can be micro/macro and which may have visual field defects as a symptom and can be confirmed by MRI. So when do we go for MRI, only if the patient has visual field defects? I think we should do it for every case, as it can be a brain tumor disrupting the dopamine connection and causing hyperprolactinemia.

 

Drugs list are : Phenothiazines, Metochlorpromide, methyldopa, reserpine, TCAs, narcotics, SSRIs, Risperidone.

 

Now there is a twist in the case – the book says it need not be amenorrhea but can be any menstrual abnormality! So do we modify the rules again? No, for this my plan is to have a taxonomy in facts, the facts present in the subcategory automatically activates the fact categories above it. So amenorrhea would automatically lead to activation of menstrual abnormalities which in turn will lead to activation of gynecological symptoms category. Now the system must be made smart enough to know that rules have to use tokens/$ to derive from the taxonomy. Correction – the correct way would be to make system smart enough to generalize rules for eg. If i give a rule menstrual abnormalities + galactorhea = Hyperprolactinemia, the system automatically generates rules like amenorrhea + g = H, oligomenorrhea + g = H etc etc.

 

Another twist – the rules differ in males and females!

 

 

 

 

 

 

 

The management guideline rules are as follows :

 

 

Rules: 

To be added from discussion

Questions: 

What is the diagnosis? What is the best next step in managing this patient? <!-- @page { margin: 2cm } P { margin-bottom: 0.21cm } -->

What examination is to be done? What are we trying to find here?

Hypothyroid findings?

 

What is the diagnosis? What is the DDX, and how does the questions relate to that?

 

0
Your rating: None
Premium Drupal Themes by Adaptivethemes